Journal of Geophysical Research: Biogeosciences

Linkages between denitrification and dissolved organic matter quality, Boulder Creek watershed, Colorado

Authors


Abstract

[1] Dissolved organic matter (DOM) fuels the majority of in-stream microbial processes, including the removal of nitrate via denitrification. However, little is known about how the chemical composition of DOM influences denitrification rates. Water and sediment samples were collected across an ecosystem gradient, spanning the alpine to plains, in central Colorado to determine whether the chemical composition of DOM was related to denitrification rates. Laboratory bioassays measured denitrification potentials using the acetylene block technique and carbon mineralization via aerobic bioassays, while organic matter characteristics were evaluated using spectroscopic and fractionation methods. Denitrification potentials under ambient and elevated nitrate concentrations were strongly correlated with aerobic respiration rates and the percent mineralized carbon, suggesting that information about the aerobic metabolism of a system can provide valuable insight regarding the ability of the system to additionally reduce nitrate. Multiple linear regressions (MLR) revealed that under elevated nitrate concentrations denitrification potentials were positively related to the presence of protein-like fluorophores and negatively related to more aromatic and oxidized fractions of the DOM pool. Using MLR, the chemical composition of DOM, carbon, and nitrate concentrations explained 70% and 78% of the observed variability in denitrification potential under elevated and ambient nitrate conditions, respectively. Thus, it seems likely that DOM optical properties could help to improve predictions of nitrate removal in the environment. Finally, fluorescence measurements revealed that bacteria used both protein and humic-like organic molecules during denitrification providing further evidence that larger, more aromatic molecules are not necessarily recalcitrant in the environment.

1. Introduction

[2] Humans have significantly altered reactive nitrogen cycling within the biosphere [Galloway et al., 2008]. Anthropogenic addition of excess reactive nitrogen to the environment has several unintended consequences that directly affect our society, including contamination of groundwater and estuarine eutrophication [e.g., National Research Council, 2000]. Despite high anthropogenic loadings, the results from mass balance studies show that on average only 10%–30% of added nitrogen is transported to coastal environments [Howarth et al., 2006; Seitzinger et al., 2006]. These same mass balance studies point to denitrification, the microbial reduction of nitrate (NO3) to gaseous nitrogen (N2O and N2), as one of the processes responsible for removal of the “missing” fixed nitrogen [van Breemen et al., 2002]. However, denitrification remains a poorly characterized process due to measurement difficulties and environmental heterogeneities [Groffman et al., 2006], and it is challenging to accurately assess the importance of the process on a watershed, regional, or global scale [Johnes and Butterfield, 2002]. Recently, model estimates of whole stream network denitrification have been improved through the addition of reaction rate constants aimed at incorporating the nonlinear nature of biological activity [Alexander et al., 2009; Mulholland et al., 2008; Wollheim et al., 2008]. By increasing our understanding of fundamental controls on denitrification we can continue to improve model estimates enabling better predictions for how watersheds will respond to global change pressures such as enhanced nitrogen deposition and land use change.

[3] Organic carbon fuels the majority of microbial processes, including those that regulate in-stream nitrogen constituents via denitrification [Seitzinger, 1994], and is an important link between energy and nutrient dynamics within streams. Dissolved organic carbon (DOC) influences stream ecosystem functions by controlling biogeochemical reactions and microbial food webs. Extensive work has been done to understand the factors controlling dissolved organic matter (DOM) quantity and quality (i.e., chemical composition) and the flux of DOM within streams. DOM is made up of a range of organic compounds, each with varying degrees of reactivity and ecological significance. The chemical nature of DOM is highly dependent on its source [McKnight et al., 2001] and in situ transformations by microbes [Cole et al., 2007], resulting in both spatial and seasonal variability within a watershed. The bioavailability of DOM, and hence its microbial utilization, is correlated with its chemical nature. For example, organic acids (a dominant component of freshwater DOM) with lower C:N ratios are generally more bioavailable to microbial communities [e.g., Hunt et al., 2000]. Thus by measuring the quantity and the quality of DOM, researchers may be able to better constrain the spatiotemporal patterns of in-stream nitrogen processes.

[4] The reliance of many inorganic nitrogen transformations on the availability of oxidizable organic carbon ultimately links the removal of nitrate (NO3) to carbon cycling. Several studies in riparian and hyporheic sediments [e.g., Hedin et al., 1998; Sobczak et al., 1998] and in streams [Inwood et al., 2005] have linked denitrification rates and NO3 concentration to DOC availability. It should be noted that several studies have found no significant relationship between DOC concentrations and denitrification rates [e.g., Arango and Tank, 2008; Bernhardt and Likens, 2002]. In general, these studies examined how denitrification rates responded to predefined carbon additions or DOC concentrations. Fewer studies have measured denitrification rates and characterized the chemical nature of the in situ organic matter pool. An exception is work by Baker and Vervier [2004], who found a positive relationship between the fractional contribution of acetate, formate, lactate, and butyrate to total DOC and rates of denitrification within the hyporheic zone of the Garonne River. By examining a natural gradient of DOM quality and quantifying denitrification potentials in relation to seasonal and spatial changes in the ambient organic matter pool this study provides an important link between these previous field experiments and observations.

[5] The Colorado Front Range receives elevated inorganic nitrogen (N) deposition due to its proximity to the Denver metropolitan area and agricultural activities in eastern Colorado, with N deposition increasing with elevation [e.g., Williams et al., 1996]. High elevation watersheds are particularly sensitive to increased N deposition because of their short growing seasons, poorly developed soils, and sparse vegetation. In response to this chronic deposition, the structure and function of both aquatic and terrestrial ecosystems have shifted [e.g., Baron et al., 2000; Williams and Tonnessen, 2000]. The Boulder Creek watershed (Figure 1) has been subjected to this human perturbation and spans an elevation gradient that produces multiple ecosystem types with varied nutrient characteristics. Specifically, as vegetative cover increases at decreasing elevation, soil C:N ratios increase and stream DOC concentrations increase, while stream NO3 concentrations decrease [Hood et al., 2003]. The chemical composition of DOM in streams also changes in space and time. For example, Hood et al. [2005] found that DOM sources changed significantly throughout the year within the alpine and subalpine regions, dominated by allochthonous sources during snowmelt and algal production from upstream lakes in the late summer.

Figure 1.

Map of the Boulder Creek, Colorado, watershed. Sampling took place at 7 sites throughout the watershed. Additional information about the sites can be found in the text and in Table 1. Site BE is a first-order stream entering the Boulder Creek main stem from the north, below the ORO sampling location.

[6] Here we utilize the nitrogen and carbon gradients created through the nexus of atmospheric deposition and elevation-controlled ecosystems in the Boulder Creek watershed to explore the relationships between organic matter characteristics and denitrification. We measured the denitrification potential, DOM quality, and stream chemistry at seven stream sites along an elevation gradient in the Boulder Creek watershed. Laboratory bioassay experiments were repeated 2–3 times at each site in an effort to determine if DOM characteristics, quantity, and denitrification potential changed through time. Our primary objective was to determine if DOM quality is a significant control over denitrification rates and, if so, what DOM quality measures provide the greatest explanatory power within a predictive model.

2. Methods

2.1. Site Description

[7] The Boulder Creek watershed is located within the Front Range of the Rocky Mountains and drains 1160 km2. The watershed encompasses a large altitudinal gradient, ranging from 4120 m (Continental Divide) to 1480 m (eastern plains) (Figure 1). The large elevation gradient within the watershed results in five climatic zones: alpine, subalpine, montane, foothills, and plains. The sampling locations span this gradient allowing us to sample across a small region yet capture a range of ecosystem types (Table 1, Figure 1). Sampling events also spanned a range of discharge conditions (3 L s−1 at GG in July 2009 to 2689 L s−1 at ORO in July 2009) given the variation in sampling date and stream order (Table 1).

Table 1. Boulder Creek Watershed Sampling Sites and Collection Datesa
SiteCollection DatesElevationClimatic ZoneDescriptionDischarge (L s−1)
  • a

    Sites GL4, ALB, and WF are located within the Niwot Long-Term Ecological Research (Niwot LTER) site, and sites GL4, GG, and BE are part of the Boulder Creek Critical Zone Observatory (BcCZO).

  • b

    The gauge was malfunctioning on the date of sampling, this value is interpolated from surrounding dates; na indicates that discharge measurements are not available for all dates.

GL46/09, 8/093535 malpine1st-order tributary of Boulder Creek155 / 109
ALB6/09, 8/093250 msubalpine1st-order tributary of Boulder Creek, downstream from GL4267 / 200
WF6/09, 8/092963 msubalpine2nd-order stream, North Boulder Creek1644b / 475
GG11/08, 3/09, 7/092680 mmontane1st-order tributary of Boulder Creekna / 5 / 3
ORO11/08, 2/09, 7/091775 mfoothills4th-order stream, Boulder Creek main stem396 / 425 / 2689
BE5/091934 mfoothills1st-order tributary of Boulder Creekna
BCA11/08, 2/091562 mplains4th-order stream, Boulder Creek main stem226 / 453

2.2. Sample Collection

[8] Water and sediment samples for experiments were collected from stream channels at all 7 sampling sites in 2008 and 2009 (Table 1). Stream water and sediment samples were collected in 1 L HDPE bottles and baked (475°C) mason jars, respectively. Surface sediment (approximately the top 5 cm) was collected from multiple locations within the stream with a trowel to account for the heterogeneity of material in the channel, i.e., each jar consisted of sediment from multiple locations within the stream. Samples were transported back to the laboratory on ice. Water for chemical analysis was filtered through a Gelman 0.45 μm capsule filter, and preserved by freezing (anions), chilling at 4°C (organic carbon analyses, pH, alkalinity), and acidification with H2SO4 (pH ≈ 2) and chilling at 4°C (cations). Unfiltered water and sediment were stored at 4°C until the incubation experiments commenced (<2 days).

2.3. Analysis of Stream Solutes

[9] Anions (including NO3 − N) and cations (including NH4+ − N) were analyzed by ion chromatography [Smith et al., 2006]. Dissolved organic carbon (DOC) was analyzed with an O.I. Analytical Model 700 TOC Analyzer (O.I. Analytical, College Station, Tex.) via the platinum catalyzed persulfate wet oxidation method [Aiken et al., 1992], except in the case of samples collected at BE, which were analyzed via high temperature combustion with a Shimadzu TOC analyzer (Shimadzu, Columbia, Md.). Quality assurance and control included analyzing samples in duplicate in addition to using instrument replication, reference standards, and blanks.

2.4. Dissolved Organic Matter Characterization

[10] Following the methods of Aiken et al. [1992], one liter of filtered stream water DOM was fractionated using resin columns into three groups: larger molecular weight hydrophobic acids (HPOA), smaller molecular weight hydrophilic molecules (HPI), and transphilic acids (TPIA) using Amberlite XAD-8 and XAD-4 resins. The amount of organic matter within each fraction was calculated using the DOC concentration and the sample mass of each fraction and are presented as the percentage of total DOC. Stream water samples were fractionated in duplicate and average values are presented. The standard deviation of these fractions is ≤2%.

[11] Stream water and sediment extract samples were analyzed for UV-Vis absorption using a Hewlett Packard Model 8453 photo-diode array spectrophotometer (λ= 200–800 nm) and a 1-cm path length quartz cell. The instrument blank consisted of deionized water. Samples were diluted by weight to be within the range of the instrument. Specific UV absorbance (SUVA254) was determined by dividing the decadal absorption coefficients (m−1) measured at λ = 254 nm by DOC concentration (mg C L−1). SUVA254 has been used as a measure of DOC aromaticity [Weishaar et al., 2003]; replicate SUVA254 measurements indicated measurement uncertainty (one standard deviation) of 0.1 L mg C−1 m−1.

[12] Stream water and sediment extract samples were also characterized by three-dimensional (3-D) fluorescence, a measure of the fluorescing portion of the DOM pool, using a Fluoromax-3TM fluorometer as described byCory et al. [2010]. Briefly, the resulting 3-D fluorescence intensity results, excitation-emission matrices (EEMs), must be corrected for a number of factors including: blank subtraction, Raman normalization, and instrument specific corrections. The EEMs were subsequently analyzed using three methods: calculation of the fluorescence index, “peak picking,” and Parallel Factor Analysis (PARAFAC). The fluorescence index (FI), introduced byMcKnight et al. [2001], provides information about the source of the organic material by analyzing the slope of an emission curve at an excitation wavelength of 370 nm; with steeper slopes (>1.8) correlated with microbial or algal end-members and lower FIs (1.3–1.4) correlated with terrestrial sources. The FI ratio was calculated as the ratio of emissions intensities at excitation wavelength of 370 nm and emission wavelengths of 470 and 520 nm [Cory et al., 2010]; the standard deviation of replicate measures and subsequent calculated FIs was ±0.004. Different regions of an EEM have been linked to various pools of DOM (e.g., humics) and thus by determining the fluorescence intensity within each region or “peak” the 3-D data can provide compositional information of the fluorescing DOM pool [Coble, 1996, 2007; Kraus et al., 2008; Stedmon et al., 2003]. Peaks were determined from the literature [e.g., Kraus et al., 2008, and references therein] and the fluorescence intensity at that location was determined using MATLAB 7.9 (The Mathworks Inc., 2009, Natick, Mass.). Finally, the EEMs were analyzed via comparison to the PARAFAC model developed by Cory and McKnight [2005]. This model attempts to predict the EEMs of DOM on the basis of 13 components, 7 of which have spectra similar to known quinones and 2 of which are similar to amino acid spectra. The results of this analysis are reported as the percent of total fluorescence explained by each component, e.g., the weight or loading of each component to the total measured fluorescence. The residuals of this comparison were always less than 5% of the total fluorescence for all the stream samples, indicating that the thirteen components successfully modeled the majority of fluorophores in the sample.

2.5. Stream Sediment Characteristics

[13] Duplicate stream sediment samples were weighed and dried in a 60°C oven to determine the percent dry weight by mass. After drying, sediment was sieved through a 2 mm screen and the <2 mm size fraction was ground in a shatter box for 20 s. The ground sediment was analyzed for carbon and nitrogen content using an Exeter Analytical CE-440 CHN elemental analyzer; duplicate samples indicate an average standard deviation of 0.04% and 0.01% for the percent carbon and nitrogen measures, respectively.

2.6. Denitrification Assays

[14] The denitrification potential of the stream sediments was determined using the acetylene block technique [Yoshinari and Knowles, 1976] by measuring N2O accumulation over ∼24 h in the headspace of sealed anoxic bottles. While there are limitations with the acetylene block technique [see Groffman et al., 2006, for discussion], this approach allowed us to examine the spatial and temporal variability of denitrification potentials throughout the watershed. The experiments consisted of two treatments (± added NO3) done in triplicate at ambient stream temperature (∼4°C) with no organic matter additions. The November 2008 bioassays did not include an ambient treatment. Briefly, approximately 1 L of unfiltered river water was deoxygenated in a large, baked beaker on ice via equilibration with the atmosphere of an anoxic glove box for 24 h. Bottles for denitrification assays were prepared within the glove box to ensure an anoxic environment. Sediment was pooled from multiple (2–3) mason jars within the glove box for the analytical triplicates in order to minimize sediment heterogeneity. It should be noted that, despite the compositing of sediment samples, in the field and in lab, it is unlikely that the sediment in each replicate was exactly the same. Approximately 30 g of wet sediment and 30 g of deoxygenated unfiltered river water were added to six 60 mL baked serum bottles, which were then sealed with autoclaved rubber stoppers and removed from the glove box. Bottles were flushed with He for 20 min on ice, after which 5 mL of calcium carbide generated C2H2 and 15 mL of He were added to each bottle. After the bottles rotated for 1 h in an incubator (∼4°C), 0.75 mL of 5 mM anoxic NaNO3 was added to three of the six bottles, bringing the [NO3] in the water to ∼100 μM. The headspace of each bottle was then immediately analyzed for N2O on a gas chromatograph (HNU GC301) equipped with an electron capture detector and a back-flush valve and then reanalyzed at 30–60 min intervals. The experiments continued until it appeared that the entire nitrate pool had been reduced to N2O, usually between 24 and 36 h. Due to repeated sampling, we accounted for changes in headspace pressure when calculating the N2O concentrations in the bottles. The reported denitrification rates were determined using a linear regression of N2O produced over the first 4 h of the experiment to minimize bottle effects and the ability of the microbes to adjust to the anoxic environment. In several cases the rate of N2O production in the ambient denitrification potential assays became nonlinear after 2–3 h; in these instances rates were determined over a shorter period of time (as noted in Table 3). Rates of N2O production are expressed as denitrification potentials by converting N2O to N and normalizing by sediment dry mass (μmoles N (g DM)−1 h−1).

[15] Incubations to assess changes in the fluorescing DOM pool during the denitrification experiments were conducted in July and August 2009. Following the same procedure as above, 125 mL baked serum bottles were filled with approximately 60 mL of sediment and deoxygenated unfiltered river water. In this case, 10 mL of C2H2 and 30 mL of He were added to each bottle, with 1.25 mL of 5 mM NaNO3 added for the NO3 treatment. After the first N2O measurement (T0) three bottles were sacrificed; the sediment slurry was centrifuged at 10,000 RPM (Beckman Model J2–21 centrifuge, Beckman Coulter, Brea, Calif.) for 15 min and the supernatant filtered (0.45 μm) for organic matter analyses. This measurement was made to determine if the DOM pool changed substantially due to increased contact with sediment in the bottles. Similarly, after N2O was no longer being produced (TF) triplicate bottles were sacrificed for organic matter concentration and characterization measurements.

2.7. Dissolved Organic Matter Changes and Denitrification

[16] Using the fluorescence EEMs in conjunction with the denitrification experiments, we characterized the types of organic matter used by the bacteria in the reduction of added nitrate. The utilized DOM (DOMdenit) was characterized by comparing the EEMs measured on sediment water slurries from the beginning of the assays (T0) and at the end of N2O production in the assays (TF) from both the ambient and elevated nitrate experiments (equation (1)).

equation image

where AM and NO3 represent the EEMs collected from the ambient nitrate concentration and 100 μM nitrate denitrification assays, respectively. The fluorescing pool of DOM within the experimental bottles could change for three reasons: (1) use by microbes, (2) production by microbes, and (3) leaching or sorption of DOM from/to the sediment. By subtracting T0 EEMs from TF EEMs changes associated with sediment-DOM interactions should be minimized.

2.8. Aerobic Metabolism Bioassays

[17] The bioavailability of the organic matter in the stream water and sediment was determined by measuring the evolution of CO2 in sealed bottles over time. The aerobic incubations were done in conjunction with all 2009 denitrification assays. For each site there were three treatments, river water (two replicates), river water and sediment (three replicates), and deionized water. Baked 60 mL serum bottles were used for these assays, with 45 mL of unfiltered water added to the river water only treatment and 30 g of sediment and 30 mL of unfiltered water added to the water and sediment treatment. All bottles were sealed with autoclaved stoppers and 25 mL of air was added via syringe to ensure an aerobic environment. The bottles were then incubated for five to seven days on a rotator at 4°C and the headspace was periodically sampled for CO2using a LICOR Infrared detector (Model LI-6262) after an equilibration period of 5–6 h. Rates of CO2 evolution were determined using linear regression. Rates of sediment CO2 production were determined by subtracting the blank normalized results of the river water only treatment from the sediment and river water treatment.

[18] Measurements of oxygen concentrations in replicate bottles for February through June 2009 experiments provided evidence that the bottles remained aerobic in all cases with the exception of the GL4 in June 2009. Rates of O2 consumption averaged 0.07 μmoles O2 (g DW)−1 h−1, ranging from 0.01 μmoles O2 (g DW)−1 h−1 (ORO) to 0.35 μmoles O2 (g DW)−1 h−1 (GL4). Given that the bottles start with oxygen levels ranging from 300 to 320 μmoles L−1, only the GL4 experiment went anoxic over the 5–7 day experiment.

2.9. Data Analyses

[19] Pearson's correlation (r) statistics were reported for all correlations, along with the associated p-value. In cases where data were not normally distributed it was transformed to meet the requirements of the statistical test, in all cases using a logarithm (base 10). Analysis of Variance (ANOVA) with a Tukey's Family Error rate of 5% was used to determine statistical differences across the watershed, i.e., to test differences between average site values. Two-tailed t-tests were used to calculate whether experimental rates differed significantly through time at a given site. Given that many of the stream nitrogen and carbon chemistry variables were correlated, multiple linear regressions (MLR) were used to determine the relative predictive importance of each characteristic; i.e., by using statistics associated with MLR analysis the relative strength of each relationship can be ascertained because the model accounts for the variation of each predictor. Best subsets regression was used to explore which predictors should be entered into the MLRs. Best subsets determines the “best” set of predictors by using a variety of statistics including adjusted R2, Mallows' CP (which should be close to the number of predictors in the model plus the constant), and S (standard deviation of the error term in the model) on four different groups of variables, each group contained log (DOC) concentration, log (NO3) concentration and log sediment C content: (1) summary spectral characteristics (log (SUVA), log(FI), and log(absorbance at 254 nm)), (2) fractionation data (%HPOA, %TPIA, and %HPI), (3) PARAFAC components, and (4) log of the peak intensities for both ambient and nitrate denitrification potentials. The predictors chosen by best subsets were then put into a MLR and only kept if their coefficients were statistically significant. In addition, when an outlier was present the analysis was repeated without the outlier data, when present, differences are noted in the text. Relationships were significant if the p-value was <0.05.

3. Results

3.1. Water Chemistry

[20] GL4 had the greatest (p< 0.001) nitrate-nitrogen concentrations throughout the sampling period (21.9μM ± 9.5 μM, Table 2). The remaining sites, all below the tree line, had average nitrate concentrations of 5.3 μM ± 3.2 μM with concentrations increasing at the onset of snowmelt. At the time of sampling, all the sites had nondetectable ammonium concentrations (detection limit ≈2 μM). Dissolved organic carbon concentrations were greatest at GG (550 μM ± 100 μM, p < 0.001) and lowest at GL4, ALB, and WSF (191.7 μM ± 58.3 μM, p < 0.001). Given the frequency of sampling, we compared our results with data obtained from the Boulder Creek Critical Zone Observatory (BcCZO, http://czo.colorado.edu/) and Niwot Long-term Ecological Research (Niwot LTER,http://culter.colorado.edu/NWT/index.html) programs. When comparisons were possible (GL4, ALB, WF, GG, and ORO) the average measured nitrate, ammonium, and DOC concentrations of our samples were always within one standard deviation of 2008 and 2009 means. Seasonal variability of DOC and inorganic nitrogen is well documented in the alpine and subalpine sites, with the greatest concentrations coinciding with the snowmelt peak and lowest concentrations occurring during the recession limb in late summer [Hood et al., 2005; Williams et al., 2001]. Organic carbon and inorganic nitrogen seem to follow similar patterns within the montane and foothills with the exception of increases in NO3 concentration following spring precipitation events (unpublished data, BcCZO). While sampling did not capture peak snowmelt, samples were taken on both the ascending (e.g., February 2009 ORO and June 2009 GL4) and descending limbs of hydrographs (e.g., Nov 2008 ORO and August 2009 GL4). This seasonal variation in sampling adds to the variability in inorganic nitrogen and DOC conditions captured by the laboratory bioassay experiments.

Table 2. Water Chemistry and Sediment Characteristics in the Stream at the Time of Sampling
SiteMonthWhole Water, FilteredStream Sediment
NO3 (μM)DOC (μM)Avg %CAvg %NC:N
GL4June 0921.411426.2870.51914.1
 Aug 0935.36838.7490.71514.3
ALBJune 091.882502.3040.11723.0
 Aug 094.142080.6090.03023.4
WFJune 091.612580.3340.01525.7
 Aug 095.901670.3070.02713.3
GGNov 082.386502.2150.11921.7
 Mar 0911.345502.7620.13424.1
 July 094.034580.4790.03615.3
BEMay 090.743671.3440.04535.1
ORONov 084.741750.1510.01214.5
 Feb 0911.621500.1600.00823.4
 July 095.841920.2080.01219.9
BCANov 083.292750.5330.03020.9
 Feb 0910.302080.5910.02626.9
 May 090.483750.5170.01345.0

3.2. Dissolved Organic Matter Characteristics

[21] Organic matter characteristics varied with both space and time within the watershed (see Tables S1 and S2 in the auxiliary material). SUVA254 was greatest at GG (4.4 L mg−1 m−1 ± 1.3 L mg−1 m−1, p < 0.01), and tended to be lowest at GL4 (2.3 L mg−1 m−1 ± 0.5 L mg−1 m−1) though there was no statistical difference between sites other than GG (Table S1). The average percent of HPOA and TPIA at each site did not vary significantly across the watershed; HPOA ranged from 28% at GL4 in August 2009 to 52% at GG in July 2009 and TPIA ranged from 13% in GG in March 2009 to 24% at ALB in August 2009 (Table S1). Average FI, calculated from the 3-D excitation-emission matrices, was greatest (p < 0.05) at BE (1.5) and BCA (1.43 ± 0.06) and ranged from 1.27 at ALB in June 2009 to 1.51 at BCA in February 2009 (Table S1). In general, FI data indicated that the majority of DOM originated in the terrestrial environment (FI < 1.4), with the exception of four sampling events: BE 5/09, GL4 8/09, and BCA 11/08 and 2/09 (Table S1). Absorbance at λ = 254 nm was highly correlated with the amount of HPOA in the sample (R = 0.672, p = 0.004), and there was an inverse relationship between SUVA and FI (R = −0.553, p = 0.026) reflecting the aromatic nature of terrestrial sources of DOM, similar to what has been observed in other studies [e.g., Hood et al., 2005] (Table S3).

[22] Examination of different regions or peaks within the fluorescence EEMs provided further information about the chemical composition of the stream DOM pool. The maximum intensity of the “A” peak, associated with humic acids [Kraus et al., 2008, and references therein], showed the greatest variability between samples (0.13 to 2.36) as well as the highest intensity at all sites over the sampling period contributing between 12% and 15% of the total fluorescence in the samples (Table S1). The maximum intensity of the “B” peak, often associated with amino acids [Coble, 1996], ranged from 0.05 to 1.07 between sites, however, it contributed minimally to the total DOM fluorescence, between 0.3% to 1.8% (Table S1). Seven of the thirteen PARAFAC components identified by the Cory and McKnight [2005] model varied significantly (p < 0.05) between sites (Table S2). In particular, DOM at GG had significantly different component loadings compared to the other sites; suggesting it has a greater amount of reduced (C4) and less oxidized (C11) quinones [Cory and McKnight, 2005] in the fluorophore pool (Table S2). Component 4, identified as a reduced quinone [Cory and McKnight, 2005], had the greatest loading of the thirteen components, contributing between 19% and 36% to the modeled fluorophore pool (Table S2). Components 8 and 13, with spectra similar to amino acids [Cory and McKnight, 2005], did not vary significantly between sampling sites and had combined loadings ranging from 3% (GG, July 2009) to 10% (GL4, August 2009, Table S2).

3.3. Sediment Characteristics

[23] The carbon and nitrogen contents of the stream sediment did not vary significantly across the watershed with the exception of GL4 (Table 2). The sediment collected at GL4 had significantly (p < 0.001) more carbon (7.5% ± 1.7%) and nitrogen (0.6% ± 0.1%) than the sites at lower elevations. The C:N of sediment ranged from 13.3 (WF, August 2009) to 45.0 (BCA, May 2009). Temporal variation at GG and ALB in sediment carbon and nitrogen content (Table 2) may be due to bed sediment movement during high flow events or sampling differences, despite efforts to remove variability in sediment at a given site.

3.4. Stream Chemistry Patterns

[24] Stream water nitrogen and carbon characteristics co-varied throughout the watershed. In general, watershed sites with greater amounts of nitrate had lower amounts of organic carbon (e.g., GL4). Pearson correlations (Table S3) on transformed variables revealed statistically significant negative correlations between nitrate concentrations and DOC concentrations (R = −0.595, p = 0.015) and %HPOA (R = −0.526, p = 0.036). Dissolved organic carbon concentrations were significantly related to a number of organic matter characteristics, namely %HPOA (R = 0.659, p = 0.005) and absorbance at 254 nm (R = 0.929, p < 0.0001). Furthermore, the various PARAFAC components and fluorescence peak intensities were significantly correlated with each other as well as with other organic matter descriptors (Table S3). Two PARAFAC components (C11 and C2), as defined by Cory and McKnight [2005], were negatively correlated with peak intensities indicative of amino acids (B and T peaks), while these same peak intensities were positively correlated with another PARAFAC component (C4). Similarly, Component 8 from the PARAFAC model, associated with trypotophan-like molecules, was positively correlated with FI, reflecting the association between autochthonous produced DOM with more labile molecules such as amino acids.

3.5. Denitrification Potentials

[25] The average denitrification potentials at ambient nitrate concentrations (here after referred to as ambient denitrification potentials) did not vary significantly across sites, though there were significant differences at ALB (p = 0.03) and BCA (p = 0.003) between experiments (Table 3). Ambient denitrification potentials were greatest at GL4 (0.23 μmoles N2O − N (g DM)−1 h−1 ± 0.25 μmoles N2O − N (g DM)−1 h−1) and lowest at BE (3.4 × 10−5 μmoles N2O − N (g DM)−1 h−1 ± 7 × 10−5 μmoles N2O − N (g DM)−1 h−1). Average denitrification potentials measured under conditions of elevated nitrate concentrations (100 μM, hereafter referred to as nitrate denitrification potentials) varied significantly in space (GL4 had the highest rates, p < 0.001) as well as at a given site through time (GL4, ALB, WF and GG, p < 0.05) (Table 3). Denitrification potentials measured under the different nitrate conditions were significantly correlated with each other (R = 0.765, p = 0.002); with nitrate denitrification potentials 0.393 μmoles N2O − N (g DM)−1 h−1 higher, on average, than ambient denitrification rates. The November 2008 experiments (at GG, ORO, and BCA) included replicates without sediment (i.e., water only) instead of ambient nitrate concentration sediment slurries. Denitrification potentials in the stream water were indistinguishable from zero (data not shown).

Table 3. Denitrification Potentials as Determined by N2O Production Rates Measured in Laboratory Bioassays After 4 h of Incubation, Except Where Otherwise Noted
SiteMonthN2O Production Rates (μmoles N (g DM)−1 h−1)
@ Ambient NO3 Concentrations@ 100 uM NO3 Concentrations
AverageSDAverageSD
  • a

    Rates calculated based on first 2–3 h of experiment. All weights are normalized to dry mass (DM) of sediment. See text for further details.

  • b

    Average rate = 5.37 × 10−6 μmoles N (g DM)−1 h−1.

  • c

    Average rate = 3.46 × 10−5 μmoles N (g DM)−1 h−1.

  • d

    Rate calculated based on 6 h of experiment.

GL4June 090.06010.21790.72370.1460
 Aug 090.41170.09091.59430.2173
ALBJune 090.01320.00380.10730.0255
 Aug 090.0003a0.00020.20480.0395
WFJune 090.0109a0.00390.07340.0152
 Aug 090.0093a0.00050.14540.0116
GGNov 080.90780.0688
 Mar 090.0063a0.00520.91660.0533
 July 090.0000b0.00020.12670.0092
BEMay 090.0000c0.00010.11540.0273
ORONov 080.1192d0.0175
 Feb 090.1118a0.18300.24650.0724
 July 090.0001a0.00010.10510.0113
BCANov 080.42370.1576
 Feb 090.02140.00210.62680.0476
 May 090.00040.00050.64190.0462

[26] Ambient and nitrate denitrification potentials were exponentially related to stream nitrate concentrations (adj R2 = 0.939, p < 0.0001 and adj R2 = 0.567, p = 0.0005, respectively) (Figure 2). However, these relationships were driven in large part by one experiment (GL4, August 2009). Consequently, the relationships were reanalyzed treating this result as an outlier. Note that GL4 has significantly higher nitrate concentrations compared to the other sites (Table 2). This approach does not suggest that these results are in error, but rather is taken to determine the robustness of relationships at lower nitrate concentrations. Repeated analyses without this experiment indicate that while the trends remain, in the case of nitrate denitrification potentials, the relationship is no longer significant (Figure 2b). The chemical nature of the DOM was significantly related to both ambient and nitrate denitrification potentials. Relationships between nitrate and carbon stream chemistry and ambient denitrification potentials (see Table S4) were driven by the results of the August 2009 GL4 experiment, with the exception of a negative relationship with DOC which remained significant after the outlier was removed (R = −0.579, p = 0.048). Nitrate denitrification potentials were significantly correlated with four fluorophore components identified by the Cory and McKnight PARAFAC model. There was a positive relationship with component 8 (R = 0.626, p = 0.009) and negative correlations with loadings on component 2 (R = −0.593, p = 0.016), component 6 (R = −0.681, p = 0.004), and component 11 (R = −0.562, p = 0.023) (Figure 3). Nitrate denitrification potentials were positively related to the percent of carbon in the stream sediment (R = 0.726, p = 0.001); the removal of the GL4 August 2009 data resulted in a weaker yet still significant relationship (R = 0.602, p = 0.018) (Table S4). Similarly, ambient and nitrate denitrification potentials were positively related to the percent of nitrogen in the sediment (R = 0.558, p = 0.047 and R = 0.709, p = 0.002, respectively). Interestingly, ambient and nitrate denitrification potentials were not significantly correlated with sediment C:N.

Figure 2.

Stream nitrate concentrations were exponentially related to denitrification potentials as measured in lab under both (a) ambient and (b) elevated nitrate concentrations. Each data point represents an average rate as calculated by linear regression for triplicate experiments over 4 h (6–8 time points), with error bars representing the standard deviation between the calculated rates of the three replicates. The fitted lines represent the exponential fit with (solid line) and without (dashed line) the July 2009 GL4 denitrification potential experiment results. The result from this experiment (gray circle at 35 μM NO3) was identified as an outlier, and the analysis was repeated without this data point. The relationship between stream nitrate concentrations and ambient denitrification potentials (Figure 2a) remains significant without the outlier; however, the relationship is no longer significant in the case of the nitrate denitrification potentials (Figure 2b).

Figure 3.

Denitrification potential rates under elevated nitrate concentrations (100 μM NO3) were significantly related to PARAFAC components as defined by the model described in Cory and McKnight [2005]. Component 2 (C2) and Component 11 (C11) are both spectrally similar to oxidized quinones, while Component 8 (C8) is similar to tryptophan.

[27] Best subsets regression indicates that the best multiple linear regression (MLR) model to explain ambient denitrification potential variability includes DOC concentrations and the percent of DOC as HPI (adj R2 = 78.1%, S = 0.0182 μmoles N (g DM)−1 h−1, Table 4). In addition, basic spectral data (SUVA, FI, absorbance at 254 nm) and DOC concentrations predicted 84.4% (adj R2 = 0.84, S = 0.045 μmoles N (g DM)−1 h−1) of the variance in ambient denitrification potentials, however, this relationship was heavily dependent on the data from the GL4 August 2009 experiment and when it was removed the adjusted R2 fell to 44.3% and none of the coefficients remained significant. Best subsets regression demonstrates that while nitrate denitrification potentials are better explained by sediment carbon content (adj. R2 = 0.494, p = 0.001) than stream DOC concentrations, the chemical characteristics of the DOC are able to explain a greater amount of variability than sediment carbon content (adj. R2 = 0.676, Table 4). Furthermore, the inclusion of the August 2009 GL4 data does not change the significance of various PARAFAC components or peak intensities into a MLR when predicting nitrate denitrification potentials. The inclusion of PARAFAC components 8 and 11 along with the stream DOC concentration predicted 67.6% of variability of the nitrate denitrification potential (adj. R2 = 0.676, Table 4); removal of the outlier resulted in a slightly weaker predictive relationship (adj. R2 = 0.538, Table 4). It should be noted that the amount of carbon in the stream sediment was significantly negatively related to %C11 (R = −0.502, p = 0.047), this relationship could potentially be driving the significance of C11 in the MLR. Fluorescence EEM peak intensities (E, D, A and B) in combination with stream nitrate concentrations predicted 70.4% of the variability of nitrate denitrification potential (adj R2 = 0.704, S = 0.1566 μmoles N g−1 h−1) (Table 4) and was not affected by the inclusion of the GL4 August 2009 data.

Table 4. Multiple Linear Regressions Using Ambient Stream Nitrate and Carbon Chemistry to Explain Potential Rates of Denitrification Measured in the Laboratorya
Multiple Linear Regression EquationR2Adj. R2S
  • a

    Explanatory variables are included in equation only if their coefficients were significant (p ≤ 0.05) with the exception of the constant. Further, equations are included only if they remained significant without the inclusion of the GL4 July 2009 outlier; changes in coefficient values are noted when applicable. Nitrate and DOC concentrations are in μM, component loadings are entered as percents, and peak intensities are in Raman units. R2 describes the amount of variation in the observed response variable that is explained by the predictor(s), adjusted (adj.) R2 is a modified R2 that has been adjusted for the number of terms in the model, and S represents the standard distance data fall from the regression line (in units of the response variable, μmoles N g−1 h−1); thus the better the equation predicts the response, the lower the value of S.

Ambient
= 0.972 − 0.122 * log DOC − 1.21*%HPI84.4%78.1%0.018
 
+100 μM Nitrate
= 5.17 − 0.428*logDOC + 13.4*%C8 − 38.1*%C1174.1%67.6%0.245
= 2.72 − 0.138*logDOC + 10.5*%C8 − 25.8*%C11 (without outlier)63.7%53.8%0.212
= −0.167 + 0.146*logNO3 − 3.93*logE + 7.88*logD − 4.10*logA + 0.525*logB82.8%70.4%0.156

3.6. Dissolved Organic Matter Changes and Denitrification

[28] Any positive values in the resulting EEM (DOMdenit) from equation (1) can be attributed to use by the microbes in the presence of elevated nitrate concentrations while any negative fluorescence intensity values are due to the production of DOM (Figure 4). Examination of EEMs using 2-D spectra at an excitation wavelength of 270 nm (Figure 5) provides information about the region commonly associated with amino acid like molecules (i.e., peaks “B” and “T” and components 8 and 13 in the Cory and McKnight [2005] PARAFAC model) as well as phenols (ex: 280 nm; em: 400 nm) [Huguet et al., 2010]. There is a measurable difference between the 2-D spectra for the 5 sites (Figure 5), indicating consumption of amino acid-like molecules at GL4, ALB, and WF (em: 300–340 nm). Additional changes in the fluorescence intensity are observed at 400–480 nm emissions for ALB, WSF, GG, and ORO. This region of the EEM is often associated with humic acids and recently created DOM [Huguet et al., 2010, and references therein]. Comparison of the DOMdenit EEMs with nitrate denitrification potentials reveals a positive trend between the intensity of the “B” and “N” peaks and denitrification rates (R = 0.686 and R = 0.761, respectively), however neither trend is significant due to the significantly higher (p < 0.0001) denitrification potentials for the GL4 site. Removal of GL4 from the data set results in significant relationships between the fluorescence intensities of the DOMdenit EEMs in the “B,” “T,” and “A” regions and nitrate denitrification potentials (R = 0.977, 0.982, and 0.940, respectively; p < 0.05).

Figure 4.

The excitation-emission matrices illustrating the changes in the fluorophore pool during the anaerobic denitrification assays, i.e., the result ofequation (1). (a) GL4 August 2009, (b) ALB August 2009, (c) WF August 2009, (d) GG July 2009, and (e) ORO July 2009. The peaks appearing in Figures 4a–4c in the region defined by ex 270 nm and em 300–350 nm are indicative of protein-like molecules being consumed over the course of the experiment. The smaller peaks in Figures 4d and 4e are indicative of the use of humic and recently degraded material; the ex 340 nm, em 450 nm peak in (d) has been correlated with humic materials [Coble, 1996] and the ex 270 nm, em 450 nm in Figure 4e has been correlated with humic and recent materials [Coble, 1996]. Note that the color bar scale changes with each plot to maximize the resolution of the images.

Figure 5.

The 2-D spectra of emissions at an excitation of 270 nm for the DOMdenit EEMs, i.e., the difference in fluorophore pools attributed to the denitrification of the added nitrate within the bioassays, see equation (1). These 2-D spectra highlight the region of the EEM commonly associated with amino-acid-like or protein-like fluorophores (ex: 270 nm; em: 300–340 nm). In a comparison of the 5 sites, GL4, ALB, and WSF all show greater consumption within this range, indicating use of these molecules during the denitrification of the added nitrate.

3.7. Aerobic Incubations

[29] Average aerobic sediment CO2 production did not vary significantly across the Boulder Creek watershed sites (p = 0.474). The highest rates were measured during March 2009 at GG (1.38 μmoles CO2 g(DM)−1 h−1 ± 0.18 μmoles CO2 g(DM)−1 h−1) and at BCA in February 2009 (1.22 μmoles CO2 g(DM)−1 h−1 ± 0.36 μmoles CO2 g(DM)−1 h−1). The average amount of CO2 produced over the course of the incubation and the average percent mineralized carbon, defined as percent of DOC converted to CO2 over the course of the incubation, also did not vary significantly between sites (p = 0.182, 0.442, respectively). While the total amount of CO2 produced at each site mirrored the CO2 production rates, the percent mineralized carbon was greatest at GL4 in August 2009 (16%). Aerobic respiration, as measured by CO2 production rate, was positively correlated with stream nitrate concentrations (R = 0.621, p = 0.031), the amount of carbon (R = 0.672, p = 0.017) and nitrogen (R = 0.642, p= 0.024) in the sediment, as well as the contribution of protein-like fluorophores in the DOM pool as determined by the Cory and McKnight PARAFAC model (%C8, R = 0.642,p = 0.02). Ambient denitrification rates were significantly related to the percent mineralized carbon (R = 0.690, p = 0.013), while nitrate denitrification rates were significantly related both to the percent mineralized carbon (R = 0.767, p = 0.004) (Figure 6) and to CO2 production rates (R = 0.648, p = 0.023). It should be noted that predictor variables were logarithmically transformed to meet the normal distribution requirement of a linear correlation with the exception of component 8 (%C8).

Figure 6.

Relationship between aerobic respiration as measured by the amount of CO2 produced relative to the total DOC present (percent mineralized carbon) and nitrate denitrification rates (i.e., denitrification potentials under elevated nitrate concentrations). The solid line represents the regression line calculated with all of the experimental results included, while the dashed line represents the regression without the inclusion of the outlier (GG July 2009). It should be noted that the aerobic respiration experiments took place over the course of a week, while the potential rates of denitrification were calculated based on 4 h of data.

4. Discussion

4.1. Anaerobic and Aerobic Respiration and Dissolved Organic Matter

[30] Past research has shown that heterotrophic microbial activity within the hyporheic zone is limited by the bioavailability of organic matter [Jones, 1995] and that the availability of organic substrate may limit denitrification rates in streams with high NO3 concentrations [Arango et al., 2007]. In Boulder Creek, rates of nitrate denitrification potential were strongly correlated with aerobic CO2 production rates and the percent of mineralized carbon, suggesting, that with greater rates of sediment respiration, the potential for NO3 reduction also increases. It is important to note that the CO2 evolution and denitrification experiments were done under different oxygen conditions and thus calculated respiration rates cannot be attributed to denitrification. However, this positive relationship does imply that microbial community functions, including both anaerobic and aerobic respiration, are dependent on similar characteristics of the substrates available within the stream environment. This finding is supported by whole stream ecosystem work conducted by Mulholland et al. [2008], who found that ecosystem respiration was positively correlated with denitrification rates. Furthermore, the significant positive relationships between the amounts of protein-like fluorescence and both aerobic respiration rates and denitrification potentials under elevated nitrate conditions are consistent with past work examining the aerobic biodegradability of DOM [Fellman et al., 2009]. These parallels suggest our understanding of organic matter availability in aerobic environments could be applied to anaerobic situations when denitrification is active.

4.2. Denitrification Potentials

[31] We compared our measured ambient denitrification potentials to other studies which use the acetylene-block method and report rates normalized to sediment mass. In general, our ambient denitrification potentials (average 0.0497μmoles N (g DM)−1 h−1) fall within ranges reported in the literature. Our results are similar to two studies from agricultural streams in the Midwest; Arango et al. [2007] report rates ranging from 0.0007 to 0.79 μmoles N (g DM)−1 h−1 for sediment, coarse and fine benthic organic matter, biofilm and sand; and Schaller et al. [2004] reported an average sediment denitrification potential of 0.027 μmoles N (g DM)−1 h−1. Our results were also within the range reported for four streams in Maryland (0.0002 to 0.3539 μmoles N (g DM)−1 h−1) [Groffman et al., 2005], and four streams in Illinois (0 to 0.2357 μmoles N (g DM)−1 h−1) [Opdyke and David, 2007]. In contrast our nitrate denitrification potential rates (average 6.50 ± 6.11 μmoles N (g DM)−1 h−1) generally fall above the reported rates, though this is not surprising given that the previously mentioned rates were measured at ambient stream nitrate concentrations. Further, the increased denitrification potentials measured in the NO3 amended bottles reflect the findings of Opdyke and David [2007], who also observed an increase in nitrate denitrification potential in incubation experiments with fine grained sediments. While a grain size analysis was not performed on our sediments, based on visual inspection, sites with finer grained sediment (GL4, GG, and BCA) generally showed higher nitrate denitrification potentials when compared to sites with coarser sediment (e.g., ORO, BE).

[32] A positive relationship between nitrate concentrations and denitrification potentials has been found across different ecosystem and land use types [e.g., Inwood et al., 2005; Mulholland et al., 2008] and our results reflect this relationship (Figure 2, Table S4). However, as previously noted this relationship was driven in large part by the results from one set of experiments, GL4 August 2009. For example, ambient denitrification potentials were positively related to stream nitrate concentrations, however, below 22 μM nitrate concentrations only explained 25% of the variability of the measured denitrification potential in the laboratory (Figure 2a).

[33] Examining the relationships between denitrification potentials and stream carbon chemistry provides insight as to the control of organic carbon quantity and quality on denitrification rates. Using information about the organic carbon pool (concentration and percent hydrophilic organic acids) increased the predictive power in the MLR to 78.1% (Table 4). Interestingly, the relationship between denitrification potentials and DOC concentrations is consistently negative in the multiple linear regressions. This relationship could be due to increased competition for available nitrate by heterotrophs [Taylor and Townsend, 2010]. The result could be due to a spurious relationship, specifically the inverse relationship between NO3 and DOC concentrations across the Boulder Creek watershed (Table S3). However, when NO3 is included as an explanatory variable, while not significant on its own, the relationship between DOC and nitrate denitrification potential remains negative. A similar result was found by Schaller et al. [2004], who report a negative relationship between denitrification potentials and DOC concentrations in NO3 rich streams of the Midwest.

[34] Given that nitrate denitrification potentials were significantly related to the chemical characteristics of the DOM pool as defined by fluorescence (Figure 3, Table S4), a significant amount of variability (up to 70%) in nitrate denitrification potentials could be explained by including peak intensities or PARAFAC components into a MLR (Table 4). Furthermore, nitrate denitrification potentials were positively related with protein-like fluorophores (e.g., C8, as identified by theCory and McKnight [2005] PARAFAC model, and peak B). In comparison, past studies have shown mixed results concerning the relationship between stream DOC and denitrification rates. Inwood et al. [2005] determined that DOC was a secondary influence on measured denitrification rates, though they postulate that this relationship could be an indicator of sediment carbon's influence on denitrification, a relationship established in this (Table S4) and other studies [e.g., Arango et al., 2007]. Others have found no relationship between denitrification rates and stream carbon concentration [e.g., Herrman et al., 2008] nor a measured response to the addition of a labile carbon pool [e.g., Bernhardt and Likens, 2002]. However, as previously noted, most researchers have examined the relationship between DOC concentrations and denitrification rates, as opposed to the chemistry of the organic matter. Further, several studies have shown that DOC concentration is not related to the measures of carbon quality [e.g., Jaffé et al., 2008], suggesting that our findings are not necessarily in conflict with these previous studies.

[35] The link between denitrification rates and ambient carbon quality metrics, coupled with consistently higher denitrification potentials under elevated nitrate concentrations, suggests that the rate of denitrification could be governed by the quality of the organic matter pool, especially at higher nitrate concentrations when carbon may be limiting. It further suggests that if the headwaters of Boulder Creek continue to receive chronic inputs of N deposition and thus remain nitrogen saturated [Williams et al., 1996], the reduction of added nitrate could be mitigated by the documented seasonal changes in the composition of the DOM pool [Hood et al., 2005; Miller and McKnight, 2010; this study].

[36] The results of the multiple linear regression analyses suggest that significant amounts of the variability in nitrate denitrification potentials could be explained using ambient stream nitrate concentrations and carbon chemistry. Regressions were able to explain up to 70.4% and 78.1% of the observed variability in the measured nitrate denitrification potentials and ambient denitrification potentials, respectively (Table 4). Admittedly, these relationships are from laboratory incubations measuring potential reaction rates, however they suggest that spectral measurements can be used to incorporate DOM reactivity into river network models [e.g., Böhlke et al., 2009; Mulholland et al., 2008; Seitzinger et al., 2006; Wollheim et al., 2006] in an effort to improve nitrate removal predictions.

4.3. Utilization of Protein-Like DOM by Denitrifiers

[37] Using the fluorescence measurements in conjunction with the denitrification experiments, we characterized the types of organic matter used by the bacteria in the reduction of added nitrate. Results of these calculations revealed that at the 2 sites with significantly higher denitrification potentials (ALB and GL4) the microbial community preferentially consumed protein-like DOM (Figures 4 and 5) during the course of denitrification. Sites WF, GG, and ORO, which are located below tree line, had lower denitrification rates despite stream nitrate concentrations similar to ALB (Table 2). While the bacteria in the WF experiment also appeared to preferentially consume the protein-like fluorophores (Figures 4c and 5) the same was not true for GG and ORO (Figures 4d–4e and Figure 5), where humic-like DOM appeared to be consumed instead. The preference of amino acids and proteins by bacterial populations explains the positive relationships between the contribution of protein-like molecules to the DOMdenitfluorophore pool and nitrate denitrification potentials. The positive relationship between humic-like DOMdenit fluorophores and nitrate denitrification potentials suggests that classes of larger molecular weight DOM are also readily utilized (as reviewed in Findlay and Sinsabaugh [1999]) and cannot be thought of as recalcitrant. It should be noted that DOM is an exceedingly complex group of molecules, of which only a small portion fluoresces. If the assumption is made that changes in the fluorophore pool mirror the overall pool of DOM, these experiments provide evidence that both simple and more complex organic molecules are oxidized during denitrification.

5. Conclusions and Implications

[38] Significant alteration of nitrogen and carbon cycling in response to human activities has occurred worldwide, affecting even relatively remote ecosystems lacking human development. Linkages between the availability of labile organic matter and nitrogen cycling have long been recognized [e.g., Hedin et al., 1998]; however, past work has mostly focused on the response of nitrate removal to the addition of a labile, homogenous pool of organic matter (e.g., acetate). This work presents an important step toward linking the chemical character of the organic matter pool to the removal of nitrate from the aquatic environment. Through a series of experiments we were able to show that organic matter characteristics were important predictors of nitrate removal potential under ambient and elevated nitrate concentrations. In addition, aerobic respiration rates and denitrification rates were strongly correlated with similar characteristics of the DOM pool as well as with each other, indicating that much can be learned about the potential of an ecosystem to reduce and remove nitrate through the examination of aerobic metabolism. Furthermore, the use of ambient stream chemistry and DOM spectral characteristics were able to explain over 70% of the variability observed in laboratory assays. Due to the fact that these results are based on laboratory experiments, an essential next step is to examine these relationships in the field, through both experimental and observational studies.

[39] These findings add to a growing body of evidence that suggest that, as stream nitrate concentrations increase, denitrification will likely become carbon limited. This is compounded by the projected increases in reactive nitrogen worldwide [Galloway et al., 2008], likely leading to more and more ecosystems moving toward nitrogen saturation and thus being carbon limited. This study suggests that the chemical character of organic matter is important in determining the rates of nitrate removal. Importantly, many of the DOM optical properties presented in this paper are relatively easy to measure and require little time and expense, thus they can provide a quick indicator of reactive potential. Further, in situ sensors could be designed to measure only specific excitation/emission pairs, providing real time monitoring of substrate reactivity in a stream environment. Incorporating this growing data set into studies of nitrogen transport and removal in aquatic ecosystems should lead to a greater understanding of where nitrate removal may be enhanced throughout the stream network.

Acknowledgments

[40] We would like to thank D. Repert and J. Crawford for assistance with the laboratory experiments; K. Butler and C. Hart for chemical analyses; and L. Larsen, S. Ewing, and three anonymous reviewers for helpful comments on an earlier versions of this manuscript. This research was done in collaboration with the U.S. Geological Survey's National Research Program, Boulder Creek Critical Zone Observatory (NSF-EAR 0724960), and Niwot Long-Term Ecological Research Site (NSF DEB 0423662) and supported by a National Science Foundation award (NSF-EAR 0814457) to R.T.B. The use of brand names is for identification purposes only and does not imply endorsement by the U.S. Geological Survey.