Triple Nitrate Isotopes Indicate Differing Nitrate Source Contributions to Streams Across a Nitrogen Saturation Gradient

Abstract

Nitrogen (N) deposition affects forest biogeochemical cycles worldwide, often contributing to N saturation. Using long-term (>30-year) records of stream nitrate (NO₃ ⁻) concentrations at Fernow Experimental Forest (West Virginia, USA), we classified four watersheds into N saturation stages ranging from Stage 0 (N-limited) to Stage 3 (N-saturated). We quantified NO₃ ⁻ contributions from atmospheric and microbial sources using δ¹⁵N, δ¹⁸O, and Δ¹⁷O of NO₃ ⁻ and characterized the concentrations and isotopes of NO₃ ⁻ in precipitation. Despite receiving identical atmospheric inputs, the proportions of atmospheric NO₃ ⁻ in streams averaged from 7 to 10% in the hardwood watersheds (stages 1, 2, and 3) and 54% in the conifer watershed (Stage 0). This suggests that the hardwood watersheds may be less responsive to future reductions in N deposition than the conifer watershed, at least in the short term. As shown in other studies, atmospheric NO₃ ⁻ proportions were higher during stormflow. Despite large proportions of atmospheric NO₃ ⁻ in the Stage 0 stream, total atmospheric NO₃ ⁻–N flux from this watershed (2.9 g ha⁻¹) was lower than fluxes in the other watersheds (range = 117.8–338.5 g ha⁻¹). Seasonal patterns of δ¹⁵N–NO₃ ⁻ in the hardwood watersheds suggest enrichment of the soil NO₃ ⁻ pool during the growing season due to plant uptake. In all watersheds, δ¹⁸O-based mixing models over-estimated atmospheric NO₃ ⁻ contributions to streams by up to 12% compared to Δ¹⁷O-based estimates. Our results highlight the importance of atmospheric deposition as a NO₃ ⁻ source in low-concentration streams and demonstrate the advantage of using Δ¹⁷O–NO₃ ⁻ over δ¹⁸O–NO₃ ⁻ for NO₃ ⁻ source apportionment.

Introduction

Long-term elevated deposition of atmospheric nitrate (NO₃ ⁻) and ammonium (NH₄ ⁺) to forests can lead to nitrogen saturation (Peterjohn and others 1996; Gilliam and others 2001; Aber and others 2003), with ecosystem responses including increased nitrification rates (Peterjohn and others 1996; Gilliam and others 2001; Lovett and Goodale 2011), decreased soil fertility (Adams and others 2007), and increased NO₃ ⁻ and base cation leaching from soils leading to soil acidification (Adams and others 2006; Fernandez and others 2010). Nitrate concentrations and fluxes in soil and stream waters have been previously used to assess the N saturation status of forested watersheds (Stoddard 1994; Peterjohn and others 1996; Aber and others 2003; Lovett and Goodale 2011). However, ecosystem responses to atmospheric deposition can vary depending on site factors such as land-use and disturbance histories, species composition, and hydrologic regime (Aber and others 2003; Adams and others 2014). When examined at the watershed scale, stream NO₃ ⁻ dynamics integrate myriad ecosystem processes; it is therefore important to understand the relative contributions of atmospheric and microbial sources to stream NO₃ ⁻ export. As anthropogenic additions to the global N cycle will likely continue unabated for the foreseeable future (Galloway and others 2003), clarifying how post-depositional processes in forests influence the transport and fate of atmospheric NO₃ ⁻ is critical to understanding ecosystem responses to chronic N deposition and the development of N saturation.

Aber and others (1989, 1998) described a conceptual model of ecosystem N saturation consisting of four stages, ranging from N-limited (Stage 0) to severely N-saturated (Stage 3); increasing nitrification rates and long-term increases in stream NO₃ ⁻ concentrations are important indicators of N saturation status in this conceptual model. Although rigorous testing of the N saturation concept requires extensive characterization of ecosystem pools and fluxes (Lovett and Goodale 2011), long-term (that is, decadal) stream chemistry dynamics can provide a convenient framework for investigating ecosystem responses to atmospheric NO₃ ⁻ deposition (Stoddard 1994). When coupled with stable isotope measurements of stream NO₃ ⁻ (δ¹⁵N, δ¹⁸O, and Δ¹⁷O), the relative contributions of atmospheric and microbial sources to stream NO₃ ⁻ across a range of N saturation stages can be elucidated.

Previous studies have used nitrogen and oxygen stable isotopes to track spatio-temporal changes in atmospheric and microbial NO₃ ⁻ pools and to differentiate between them (Durka and others 1994; Burns and Kendall 2002; Pardo and others 2004; Barnes and others 2008; Sebestyen and others 2008, 2014; Goodale and others 2009). For example, temporal patterns in δ¹⁵N values in precipitation and stream NO₃ ⁻ have been used to identify NO₃ ⁻ uptake in forests (Barnes and others 2008). Similarly, δ¹⁸O–NO₃ ⁻ values have been used to distinguish atmospheric and soil sources and to determine their relative contributions to ecosystem NO₃ ⁻ pools (Williard and others 2001; Pardo and others 2004; Goodale and others 2009; Sebestyen and others 2014). However, preferential uptake of lighter isotopes (for example, ¹⁴N and ¹⁶O) during biological processes such as denitrification can alter δ¹⁵N and δ¹⁸O values of residual NO₃ ⁻ pools, leading to imprecise source apportionment in some cases (Michalski and others 2004; Riha and others 2014). Nitrate isotopic source signatures can also result from processes not dependent on mass; these mass-independent signatures—such as Δ¹⁷O–NO₃ ⁻—are not fractionated by biological processes and therefore serve as conservative tracers of atmospheric and microbial NO₃ ⁻ (Michalski and others 2004). Because atmospheric NO₃ ⁻ is characterized by positive Δ¹⁷O values and microbial NO₃ ⁻ by values of zero or less, Δ¹⁷O is a robust tool for quantifying the contributions of atmospheric and microbial NO₃ ⁻ to surface waters.

Nitrogen cycling is an important, but poorly understood phenomenon in forests, and the importance of atmospheric deposition in determining N exports is still not well elucidated. In this study, we quantified atmospheric and microbial source contributions to stream NO₃ ⁻ using triple NO₃ ⁻ isotopes (δ¹⁵N, δ¹⁸O, and Δ¹⁷O) of precipitation and stream water in four nearly adjacent watersheds at Fernow Experimental Forest (Figure 1). These watersheds have historically received some of the highest rates of atmospheric NO₃ ⁻ deposition in the U.S. (Adams and others 2003). Despite receiving identical atmospheric N inputs, these watersheds demonstrate differing long-term (>30-year) patterns in stream NO₃ ⁻ concentrations due to differences in internal N retention (Figure 2). Such stream NO₃ ⁻ concentration dynamics are consistent with the progressive stages of watershed N saturation described by Stoddard (1994) and Aber and others (1989, 1998) and are likely influenced to some extent by differences in tree species composition among the watersheds. These decades-long differences in stream NO₃ ⁻ concentrations served as the framework for the N saturation gradient examined in this study. Using triple NO₃ ⁻ isotope analyses, our study objectives were to (1) characterize and quantify the source contributions to stream NO₃ ⁻ in watersheds receiving identical N deposition inputs but exhibiting long-term differences in stream NO₃ ⁻ concentrations and (2) elucidate the roles of biological and hydrologic drivers in regulating atmospheric NO₃ ⁻ export. Based on the results of prior studies (for example, Durka and others (1994)), we hypothesized that watersheds with higher N saturation status would have greater proportions of atmospheric NO₃ ⁻ in stream water than those with lower N saturation status.

Figure 1

Map of study watersheds within the larger boundary of Fernow Experimental Forest (FEF). Numbers indicate N saturation stage; black dots show weir locations in study watersheds. Contour spacing is 30 m. Inset shows the location of FEF (black dot) in the state of West Virginia (shown in grey).

Figure 2

Monthly stream NO₃ ⁻–N concentrations in the study watersheds at Fernow Experimental Forest averaged from weekly measurements collected from 1983 to 2007. The four watersheds exhibit differing long-term concentrations and seasonal patterns of NO₃ ⁻–N in streams; watershed N saturation stage classification was based on these long-term differences in stream NO₃ ⁻–N dynamics.

Methods

Study Site

Fernow Experimental Forest is located in West Virginia, on the unglaciated Allegheny Plateau (Figure 1). Elevations in the study watersheds range from 720 to 865 m, and slopes average about 20%. Bedrock is composed of hard sandstone and softer shale; little water storage occurs in these strata (Reinhart and others 1963; Kochenderfer 2007). Soils are well-drained silt loams of the Calvin series (loamy-skeletal, mixed active, mesic typic Dystrudept), averaging 1 m in depth (Kochenderfer 2007). Infiltration rates are high and most precipitation reaches the streams via subsurface flow (Reinhart and others 1963). During high-intensity rain events, runoff is high and falls off quickly during periods of low-intensity or no precipitation (Reinhart and others 1963). The growing season extends from late April through October and precipitation is evenly distributed throughout the year, averaging 1450 mm; significant snowpack does not accumulate over long periods. Mixed hardwoods are the dominant forest type at Fernow, with northern red oak (Quercus rubra), sugar maple (Acer saccharum), red maple (Acer rubrum), black cherry (Prunus serotina), and yellow poplar (Liriodendron tulipifera), the most abundant species. As an exception to these native hardwoods, one research watershed was planted to Norway spruce (Picea abies) monoculture in 1973.

The research watersheds for this study were selected based on long-term (>30-year) differences in patterns of mean stream NO₃ ⁻–N concentrations (Figure 2); these long-term differences were used to classify the watersheds into different N saturation stages, ranging from Stage 0 (N-limited) to Stage 3 (severely N-saturated), following the conceptual model of N saturation described by Aber and others (1989). We sampled three mixed hardwood-dominated watersheds (Stages 1, 2, and 3) and one conifer-dominated watershed (Norway spruce; Stage 0) for this study. Several site characteristics are similar among the watersheds (for example atmospheric deposition inputs, climate, discharge patterns, geology, soils, elevation) but the watersheds differ with respect to a number of factors including stream NO₃ ⁻–N concentrations, mean streambed and watershed slope, land-use history, and dominant overstory composition (Table 1). In addition, the riparian zone in the conifer-dominated Stage 0 watershed is characterized by patches of Sphagnum sp. in some areas; such features are absent in any of the hardwood-dominated study watersheds. The presence of Sphagnum sp. patches in the Stage 0 watershed suggests that riparian wetland areas may be starting to form (Tiner 1993).

Table 1 Description of Study Watersheds

Sample Collection

From January through December 2010, 1 L stream samples were collected in acid-washed HDPE bottles from the stream just above a 120° V-notch weir at the base of each watershed. Samples were collected on a weekly basis when streams were flowing. Samples were vacuum-filtered through 0.22 μm polyethersulfone membrane filters to remove suspended solids and biological material. All samples were processed at the US Forest Service Timber and Watershed Laboratory in Parsons, West Virginia within 24 h of collection. Filtered samples were frozen and transported to the University of Pittsburgh, where they remained frozen until isotopic analysis. Weekly samples of wet-only deposition were collected at a National Atmospheric Deposition Program (NADP) National Trends Network site approximately 2 km from the study watersheds from February through December 2010. These samples were filtered through 0.45 μm polyethersulfone membrane filters and archived at 4°C at the NADP Central Analytical Laboratory in Champaign, IL, USA. Precipitation samples were shipped to the University of Pittsburgh and frozen until isotopic analysis.

Isotopic Analysis

Nitrate concentrations were measured by ion chromatography (Dionex ICS-2000) at the University of Pittsburgh. For isotopic analysis, a denitrifying bacteria, Pseudomonas aureofaciens, was used to convert aqueous NO₃ ⁻ into gaseous N₂O which was introduced into the mass spectrometer (Sigman and others 2001; Casciotti and others 2002). For Δ¹⁷O analysis, this N₂O was thermally decomposed at 800°C into N₂ and O₂ prior to isotopic analysis following the method described by Kaiser and others (2007). Duplicate samples were analyzed for δ¹⁵N and δ¹⁸O–NO₃ ⁻ (and separately for Δ¹⁷O–NO₃ ⁻) on an Isoprime Trace Gas and Gilson GX-271 autosampler coupled with an Isoprime Continuous Flow Isotope Ratio Mass Spectrometer (CF-IRMS) at the Regional Stable Isotope Laboratory for Earth and Environmental Science Research at the University of Pittsburgh. Isotope values are reported in parts per thousand relative to NO₃ ⁻ standards as follows:

$$ \delta^{15} {\text{N}}, \delta^{18} {\text{O}}, \quad {\text{and}} \quad \delta^{17} {\text{O}} \, (\permil) = \left[ {\left( {\frac{{R_{\text{sample}} }}{{R_{\text{standard}} }}} \right) - 1} \right] \times 1000 $$

(1)

where R = ¹⁵N/¹⁴N, ¹⁸O/¹⁶O, or ¹⁷O/¹⁶O. The mass-independent oxygen isotope anomaly (Δ¹⁷O–NO₃ ⁻) is likewise reported in parts per thousand and calculated using the equation:

$$ \Delta^{17} {\text{O }}\left( \permil \right) = \delta^{17} {\text{O}} - 0.52 \times \delta^{18} {\text{O}} $$

(2)

Samples with low NO₃ ⁻ concentrations were pre-concentrated prior to bacterial conversion to N₂O. For pre-concentration, we calculated the sample volume necessary to obtain a final NO₃ ⁻ concentration of 20 nmol (for δ¹⁵N and δ¹⁸O analysis) or 200 nmol (for Δ¹⁷O analysis) in a 5 ml sample. Appropriate sample volumes were measured into 10% hydrochloric acid-washed Pyrex or Teflon beakers and placed in a drying oven at 60°C until all liquid was evaporated. The interior of each beaker was then rinsed with 10 ml of 18 MΩ water to reconstitute duplicate samples to the appropriate concentration. Samples were prepared for isotopic analysis following the bacterial denitrifier method as previously described. International reference standards were similarly pre-concentrated for correction of pre-concentrated samples.

δ¹⁵N and δ¹⁸O values were corrected using international reference standards USGS-32, USGS-34, USGS-35, and IAEA-N3; USGS-34 and USGS-35 were used to correct Δ¹⁷O values. These standards were also used to correct for linearity and instrument drift. Standard deviations for international reference standards were 0.2, 0.5, and 0.2‰ for δ¹⁵N, δ¹⁸O, and Δ¹⁷O, respectively.

δ¹⁵N–NO₃ ⁻ values must be corrected for interference from contributions of ¹⁴N¹⁴N ¹⁷O to m/z 45 (Coplen and others 2004). These contributions were evaluated following the relationship described in Coplen and others (2004), where a 1‰ increase in δ¹⁵N corresponds to an 18.8‰ increase in Δ¹⁷O. Corrected δ¹⁵N values in stream samples were zero to 1.3‰ (mean = 0.02‰) lower than uncorrected values, depending on the mass-independent contribution of Δ¹⁷O in the sample. Because Δ¹⁷O–NO₃ ⁻ values were low in most samples and because we could not apply the correction to some samples due to a lack of Δ¹⁷O data (particularly in the Stage 0 watershed), the δ¹⁵N values presented here do not include the mass-independent Δ¹⁷O correction. Although interference-corrected δ¹⁵N data are slightly lower than those presented here, the temporal and spatial patterns in δ¹⁵N should not be strongly influenced by the omission of this correction.

End-Member Mixing Analysis

We used a two-end-member isotope mixing model approach (using both δ¹⁸O and Δ¹⁷O isotope systems) to calculate the fraction of atmospheric NO₃ ⁻ exported from each watershed (equation (3)):

$$ f_{\text{atm}} = \frac{{\chi_{\text{stream}} {-} \chi_{\text{nitrification}} }}{{\chi_{\text{atm}} - \chi_{\text{nitrification}} }} $$

(3)

where χ is δ¹⁸O or Δ¹⁷O of NO₃ ⁻. We used the lowest measured stream NO₃ ⁻ δ¹⁸O (or Δ¹⁷O) value in each watershed as the nitrification end-member (Barnes and others 2008). The lowest Δ¹⁷O–NO₃ ⁻ values in the hardwood watersheds were −1.2, −1.2, and −1.4‰ in the Stage 1, 2, and 3 watersheds, respectively. In the Stage 0 watershed, a value of zero was used for the nitrification end-member Δ¹⁷O value, as all stream samples had positive values. For the atmospheric end-member, we used the volume-weighted mean δ¹⁸O (or Δ¹⁷O) value of all weekly precipitation samples collected during 2010. We used the volume-weighted mean rather than individual precipitation samples collected at the same time as stream samples because hydrologic transit times have not been characterized for most of the study watersheds, and the isotopic composition of atmospheric NO₃ ⁻ in a stream sample is unlikely to correspond to that of the precipitation sample collected during the same week.

Nitrate Flux Calculation

Nitrate–N fluxes were calculated in each watershed on a weekly basis using total weekly discharge (based on continuous flow level measurements taken at the weir) and NO₃ ⁻–N concentrations determined from stream samples collected weekly. Total annual NO₃ ⁻–N flux in each watershed was calculated as the sum of all weekly NO₃ ⁻–N fluxes during the study period. Total annual atmospheric NO₃ ⁻–N fluxes were calculated for the entire study period by multiplying the total annual NO₃ ⁻–N flux in each watershed by the discharge-weighted mean proportion of atmospheric NO₃ ⁻ in the stream (based on Δ¹⁷O–NO₃ ⁻). As the proportion of atmospheric NO₃ ⁻ in the Stage 0 watershed was based on only those stream samples with NO₃ ⁻–N concentrations sufficient for Δ¹⁷O–NO₃ ⁻ analysis (n = 4), the total atmospheric NO₃ ⁻–N flux calculated for this watershed should be interpreted with caution.

Statistical Analysis

We used analysis of variance (ANOVA) to test for significant differences among watersheds. When significant differences were indicated, we applied Tukey’s Honestly Significant Difference test to determine which means were significantly different (α = 0.05). The experiment-wise error rate was held at α = 0.05 for comparisons among watersheds. To test for differences between growing season (May–October) and dormant season (November–April) means within individual watersheds, we used repeated measures ANOVA. All statistical analyses were conducted using volume-weighted concentrations and concentration-weighted isotope values; therefore, all means are reported as weighted values. In order to include samples with NO₃ ⁻ concentrations below the instrument detection limit (0.01 mg NO₃ ⁻ L⁻¹) in statistical analyses, NO₃ ⁻ concentrations for these samples were set at 0.005 mg NO₃ ⁻ L⁻¹ (one-half of the instrument detection limit). All statistical analyses were conducted using SAS statistical software (SAS Institute, Inc. 2011).

Results

Nitrate–N Concentration

Intermittent streamflow in all study watersheds prevented sample collection during parts of July through October in the Stage 0, 1, and 2 watersheds, and August through October in the Stage 3 watershed. In addition, low stream NO₃ ⁻–N concentrations in the Stage 0 watershed precluded isotopic analysis of several samples. Stream NO₃ ⁻–N concentrations across all watersheds ranged from below the instrument detection limit to 1.91 mg L⁻¹ and precipitation NO₃ ⁻–N concentrations ranged from 0.02 to 0.67 mg L⁻¹ (Figure 3; Table 2). Mean NO₃ ⁻–N concentrations were significantly different in all watersheds (P < 0.01; Table 2); this was expected, given the long-term patterns in stream NO₃ ⁻–N concentrations in these watersheds (Figure 2) and the nature of the study design (that is, stream NO₃ ⁻ concentrations were used to define watershed N saturation status). The Stage 1, 2, and 3 (hardwood) watersheds showed significant seasonal differences in stream NO₃ ⁻–N concentrations, with lower mean values during the growing season (May to October) than the dormant season (November to April) (P < 0.01; Table 3).

Figure 3

A Nitrate–N concentration, B δ¹⁵N–NO₃ ⁻, C Daily total discharge, D δ¹⁸O–NO₃ ⁻, and E Δ¹⁷O–NO₃ ⁻ of weekly stream and precipitation samples collected at Fernow during 2010. Months shaded in grey represent the growing season. The dashed circle in E indicates samples collected during the snowmelt event on 23 February.

Table 2 Weighted Means (SE) and Ranges for Stream NO₃ ⁻–N Concentration and Isotope Values

Table 3 Weighted Means (SE) of Growing Season and Dormant Season NO₃ ⁻–N Concentrations and Isotope Values

δ¹⁵N of Nitrate

Across all watersheds, values of δ¹⁵N in stream NO₃ ⁻ ranged from −2.9‰ to +5.9‰, and precipitation values ranged from −6.0‰ to +5.1‰ (Figure 3; Table 2). Mean δ¹⁵N–NO₃ ⁻ values were not significantly different among the watersheds or precipitation (Table 2). Seasonally, mean stream δ¹⁵N–NO₃ ⁻ values were significantly higher during the growing season than the dormant season in the Stage 1, 2, and 3 watersheds (P < 0.01; Table 3). The seasonal variability in δ¹⁵N–NO₃ ⁻ may have contributed to the lower mean δ¹⁵N–NO₃ ⁻ value in the Stage 0 watershed, as streamflow was absent in this watershed during most of the growing season (when δ¹⁵N values were highest in the Stage 1, 2, and 3 watersheds; Figure 3). In contrast to the seasonal patterns in the hardwood watersheds, the mean δ¹⁵N–NO₃ ⁻ value in precipitation was lower during the growing season than the dormant season, but this difference was not statistically significant (Table 3).

δ¹⁸O of Nitrate

Across all watersheds, δ¹⁸O values in stream NO₃ ⁻ ranged from −9.4 to +72.8‰, whereas precipitation δ¹⁸O–NO₃ ⁻ values ranged from +51.1‰ to +81.7‰ (Figure 3; Table 2). The highest mean δ¹⁸O–NO₃ ⁻ value of stream NO₃ ⁻ occurred in the Stage 0 watershed (+40.4‰; Table 2). This mean was significantly higher than those in all other watersheds, but significantly lower than the mean precipitation δ¹⁸O–NO₃ ⁻ value (+65.3‰) (P < 0.01 for all comparisons). In all watersheds and in precipitation, mean δ¹⁸O–NO₃ ⁻ values were significantly higher during the dormant season than the growing season (P < 0.02; Table 3).

Δ¹⁷O of Nitrate

Δ¹⁷O–NO₃ ⁻ values in stream water ranged from −1.4‰ to +23.9‰ across all watersheds (Figure 3; Table 2), with a significantly higher mean value in the Stage 0 watershed than in the other watersheds (P < 0.01 for all comparisons; Table 2). However, mean stream Δ¹⁷O–NO₃ ⁻ values were not significantly different when ANOVA was conducted using only data for the four dates on which Δ¹⁷O values were available in the Stage 0 watershed (in other words, n = 4 for each watershed). While this analysis allows for a more balanced comparison of Δ¹⁷O values across watersheds, the small sample sizes weaken the statistical power of ANOVA, making it difficult to discern whether the mean Δ¹⁷O value in the Stage 0 watershed is truly different from the others. Precipitation Δ¹⁷O–NO₃ ⁻ values ranged from +12.3 to +29.1‰, with a mean of +20.6‰. Mean Δ¹⁷O–NO₃ ⁻ values in the Stage 1 and 2 watersheds and in precipitation were significantly higher during the dormant season than the growing season (P < 0.04; Table 3). In the Stage 0 and 3 watersheds, mean Δ¹⁷O–NO₃ ⁻ values were also higher during the dormant season than the growing season, but seasonal differences were not statistically significant. In all watersheds, the highest Δ¹⁷O–NO₃ ⁻ value measured during 2010 occurred on 23 February, coincident with a snowmelt event (Figure 3).

Proportions and Fluxes of Atmospheric Nitrate in Streams

We calculated the proportions of atmospheric NO₃ ⁻ in streams using both δ¹⁸O–NO₃ ⁻ and Δ¹⁷O–NO₃ ⁻ values to evaluate differences in estimated atmospheric NO₃ ⁻ proportions in streams. Substituting δ¹⁸O–NO₃ ⁻ values into equation (3), the mean proportions of atmospheric NO₃ ⁻ in the Stage 0, 1, 2, and 3 watersheds were 76, 14, 12, and 9%, respectively. When Δ¹⁷O–NO₃ ⁻ values were substituted into equation (3), mean proportions of atmospheric NO₃ ⁻ were 54, 9, 10, and 7% in the Stage 0, 1, 2, and 3 watersheds, respectively. The Δ¹⁷O-based estimate for the Stage 0 watershed (54%) should be interpreted with caution, as samples collected on only four dates during 2010 had adequate NO₃ ⁻–N concentrations for Δ¹⁷O–NO₃ ⁻ analysis (for comparison, mean proportions of atmospheric NO₃ ⁻ in the Stage 1, 2, and 3 watersheds calculated for these four sampling dates were 20, 11, and 10%, respectively). This caveat notwithstanding, both δ¹⁸O- and Δ¹⁷O-based estimates indicate that proportions of atmospheric NO₃ ⁻ in streams were much lower in the Stage 1, 2, and 3 (hardwood) watersheds than in the Stage 0 (conifer) watershed, and mean proportions of atmospheric NO₃ ⁻ tended to decrease with increasing mean stream NO₃ ⁻–N concentrations. For all watersheds, the mean proportions of atmospheric NO₃ ⁻ in streams estimated by δ¹⁸O–NO₃ ⁻ were greater than those based on Δ¹⁷O–NO₃ ⁻.

In contrast to the pattern of atmospheric NO₃ ⁻ proportions in streams, total fluxes of atmospheric NO₃ ⁻–N in streams increased with increasing mean NO₃ ⁻–N concentration (Table 4). During the study period, total stream fluxes of atmospheric NO₃ ⁻–N were 2.9, 117.8, 207.5, and 338.5 g ha⁻¹ in the Stage 0, 1, 2, and 3 watersheds, respectively (Table 4). These Δ¹⁷O-based flux estimates do not account for contributions of atmospheric NH₄ ⁺ that may have been oxidized to NO₃ ⁻ prior to export from the watersheds. Therefore, these estimates represent only that fraction of total stream N export contributed directly from atmospheric NO₃ ⁻. Total atmospheric NO₃ ⁻–N deposition (wet + dry) at Fernow during 2010 was 3.28 kg ha⁻¹ and total inorganic N deposition (wet + dry NO₃ ⁻–N and NH₄ ⁺–N) was 5.25 kg ha⁻¹ (U.S. Environmental Protection Agency 2014).

Table 4 Characteristics of Annual Stream Discharge, NO₃ ⁻–N Concentrations, Stream NO₃ ⁻–N Fluxes, and N Retention in the Study Watersheds

Discussion

Biological and Hydrologic Influences on Stream Nitrate Sources and Export

The patterns of atmospheric NO₃ ⁻ export from the study watersheds at Fernow contradict our initial hypothesis of increasing proportions of atmospheric NO₃ ⁻ in streams with increasing N saturation status. However, on a flux basis, atmospheric NO₃ ⁻ contributions to streams increased with increasing N saturation stage, consistent with the results of Durka and others (1994) who observed the greatest atmospheric NO₃ ⁻ fluxes from forests in a deposition-induced state of decline. Our results also demonstrate a comparatively stronger response of microbial NO₃ ⁻ fluxes to increasing N saturation stage relative to atmospheric NO₃ ⁻ fluxes, suggesting that while the capacity for biological N cycling may not decrease with progressive N saturation, the capacity for N retention within watersheds can diminish.

The study watersheds at Fernow reflect important differences in biological and hydrologic ecosystem attributes. Perhaps the most obvious difference is that of the species composition in the Stage 0 watershed, the Norway spruce monoculture. Many previous studies have demonstrated the influence of species composition on soil characteristics related to N cycling and export (Buldgen and others 1983; Lovett and others 2002, 2004; Templer and others 2005; Kelly and others 2011; Prescott and Grayston 2013). The high Δ¹⁷O–NO₃ ⁻ values in the Stage 0 watershed indicate that atmospheric deposition is a more important source of stream NO₃ ⁻ than microbial nitrification in this conifer-dominated watershed. Although it is possible that the low stream NO₃ ⁻-N concentrations and limited Δ¹⁷O–NO₃ ⁻ data in the Stage 0 watershed resulted in a biased estimate of atmospheric NO₃ ⁻ contributions in this stream, additional data support the observation that microbial nitrification constitutes a smaller proportion of total stream NO₃ ⁻ in the Stage 0 watershed than in the Stage 1, 2, and 3 watersheds. First, although δ¹⁸O is a less conservative tracer of atmospheric NO₃ ⁻ than Δ¹⁷O (Michalski and others 2004; Kendall and others 2007; Riha and others 2014), the larger dataset (n = 12) and high isotope values of δ¹⁸O–NO₃ ⁻ in the Stage 0 watershed (median = +41.7‰) support the conclusion that atmospheric deposition is a major source of stream NO₃ ⁻. Second, the mean A horizon C:N in the Stage 0 watershed is higher than that in the Stage 3 watershed (18 vs. 15 in the Stage 0 and 3 watersheds, respectively; (Kelly and others 2011)), and the Stage 2 watershed (mean C:N = 16; Christ and others 2002). It is possible that A horizon C:N is more spatially variable in the Stage 2 and 3 watersheds due to greater species diversity in these mixed hardwood-dominated watersheds compared to the Norway spruce-dominated Stage 0 watershed, potentially influencing mineralization and nitrification rates (Lovett and others 2004). Typically, litter from conifer species is more recalcitrant (that is, higher C:N and lignin:N) than litter from hardwood species such as maple and tulip poplar (Lovett and others 2004; Kelly 2010), and previous studies have demonstrated significant negative correlations between soil C:N and net nitrification rates (Christ and others 2002; Ross and others 2004; Christenson and others 2009; Kelly and others 2011). In the Stage 0 watershed, the mean annual net nitrification rate of 4.2 kg N ha⁻¹ y⁻¹ (Kelly 2010) is 27 times lower than the rates measured by Gilliam and others (2001) in the Stage 2 and 3 watersheds (114 and 115 kg N ha⁻¹ y⁻¹ respectively). This markedly lower rate of microbial NO₃ ⁻ production in the Stage 0 watershed contributes to both the low proportions of microbial NO₃ ⁻ in the stream and the low total NO₃ ⁻–N flux from this watershed. The stream NO₃ ⁻ source dynamics observed in the study watersheds have important implications for our understanding of the ecosystem effects of N deposition. For example, the higher proportions of atmospheric NO₃ ⁻ in the Stage 0 stream suggest that this watershed might show the strongest immediate response to changing rates of atmospheric N deposition. In contrast, the overwhelming contributions of microbial NO₃ ⁻ to the Stage 1, 2, and 3 streams suggest that a reduction in rates of atmospheric N deposition to these watersheds might yield only minor immediate reductions in stream NO₃ ⁻ concentrations and fluxes. In the longer term, reductions in N deposition would likely reduce the level of N cycling and nitrification in these watersheds, resulting in a decline in the microbial fraction of exported NO₃ ⁻.

Although the relative contributions of NO₃ ⁻ from atmospheric and microbial sources influence stream NO₃ ⁻-N concentrations and fluxes, watershed hydrologic status is an additional key driver of microbial and atmospheric NO₃ ⁻ export to streams (Creed and Band 1998; Adams and others 2014; Sebestyen and others 2014). In all of the study watersheds, δ¹⁸O–NO₃ ⁻ and Δ¹⁷O–NO₃ ⁻ values were generally higher in stream samples collected during stormflow, and both δ¹⁸O–NO₃ ⁻ and Δ¹⁷O–NO₃ ⁻ were positively correlated with discharge (P < 0.05) (except Δ¹⁷O–NO₃ ⁻ in the Stage 0 watershed, due to the limited Δ¹⁷O–NO₃ ⁻ dataset in this watershed). In the Stage 1, 2, and 3 watersheds, 54 to 61% of total NO₃ ⁻ export and 62 to 77% of total atmospheric NO₃ ⁻ export during 2010 occurred on nine sampling dates, coincident with stormflow. This suggests that most of the total NO₃ ⁻ (and atmospheric NO₃ ⁻ in particular) that was exported from these watersheds during 2010 occurred during stormflow. The role of hydrologic drivers in regulating atmospheric NO₃ ⁻ export is further demonstrated by the high Δ¹⁷O–NO₃ ⁻ values measured in all watersheds on 23 February, coincident with the rising limb of a snowmelt event. Using the highest precipitation Δ¹⁷O–NO₃ ⁻ value measured during the study period (+29.1‰) and the lowest Δ¹⁷O–NO₃ ⁻ value measured in each watershed (and 0‰ for the Stage 0 watershed) for the nitrification end-member in equation (3), conservative estimates of atmospheric NO₃ ⁻ in streams were 82, 22, 15, and 15% of total stream NO₃ ⁻ in the Stage 0, 1, 2, and 3 watersheds, respectively, on 23 February. Rapid routing of melt water to streams and decreased biological activity during the winter could have facilitated elevated proportions of atmospheric NO₃ ⁻ in streams on this date. Several studies in forested watersheds have also reported elevated atmospheric NO₃ ⁻ contributions to streams during snowmelt (Burns and Kendall 2002; Sebestyen and others 2008; Goodale and others 2009; Pellerin and others 2012). That the largest proportions of atmospheric NO₃ ⁻ in all streams occurred on the same date suggests that a common factor (for example, hydrologic status) influenced atmospheric NO₃ ⁻ export from the study watersheds during this stormflow event. However, when examined over the entire study period, differing biogeochemical responses among some of the study watersheds indicate that hydrologic and biological drivers may interact to regulate NO₃ ⁻ export dynamics over longer temporal scales.

Although the lowest mean proportion of atmospheric NO₃ ⁻ (7%) occurred in the Stage 3 watershed, total NO₃ ⁻–N concentrations and discharge were highest in this watershed resulting in the greatest fluxes of both total and atmospheric NO₃ ⁻–N (Table 4). In contrast, the Stage 1 watershed had a slightly greater mean proportion of atmospheric NO₃ ⁻ (9%) and nearly identical total annual discharge to that in the Stage 3 watershed, but total and atmospheric NO₃ ⁻–N fluxes in the Stage 1 watershed were around 3 times less than those in the Stage 3 watershed. Thus, despite identical atmospheric deposition inputs and nearly identical patterns of discharge during the study period, these two watersheds exhibited differing total and atmospheric NO₃ ⁻ exports during 2010. The difference in NO₃ ⁻–N fluxes indicates that retention of atmospheric NO₃ ⁻ in the Stage 1 watershed is greater than that in the Stage 3 watershed. This may be due to biological differences (for example, degree of NO₃ ⁻ retention in vegetation and soils, possibly due to lower nitrification rates in the Stage 1 watershed), hydrologic differences (for example, hydrologic transit times along flowpaths), or some combination of these (and possibly additional) factors. Although the predominant hydrologic source areas and flowpaths have not been characterized in these watersheds, we might expect the steeper slopes in the Stage 1 watershed (mean = 30.6% versus 24.8% in the Stage 3 watershed) to facilitate equal or greater fluxes of atmospheric NO₃ ⁻ to streams if hydrology were the primary driver of atmospheric NO₃ ⁻ export. That less atmospheric NO₃ ⁻ was exported from the Stage 1 watershed compared to the Stage 3 watershed despite similarities in discharge and atmospheric inputs suggests that the capacity for retention of atmospheric NO₃ ⁻ may be lower in the Stage 3 watershed. Indeed, the Stage 3 watershed served as a net source of NO₃ ⁻–N relative to the amount of atmospheric NO₃ ⁻–N deposited during 2010 (Table 4). However, quantification of the relative strengths of vegetation and soil N sinks in the two watersheds would be required to validate this speculation.

Although the higher proportions of atmospheric NO₃ ⁻ in streams during stormflow at Fernow are in agreement with the results of other studies that have examined atmospheric NO₃ ⁻ dynamics during hydrologic events (Burns and Kendall 2002; Sebestyen and others 2008, 2014; Pellerin and others 2012; Riha and others 2014), it is important to reiterate that the proportions of atmospheric NO₃ ⁻ we observed in streams are generally small throughout the year (range of median atmospheric NO₃ ⁻ proportions in all streams = 3 to 39%). Indeed, much larger discharges than those observed during the snowmelt event on 23 February were recorded from all of the watersheds during 2010 but these events occurred during the growing season, resulting in only marginal increases in the proportion of atmospheric NO₃ ⁻ in streams. Therefore, although hydrologic status is an important driver of atmospheric NO₃ ⁻ dynamics in these streams, extensive microbial NO₃ ⁻ production—particularly in the Stage 2 and 3 watersheds (Gilliam and others 2001)—also influences NO₃ ⁻ source dynamics and total NO₃ ⁻–N flux.

Seasonal Nitrate Isotope Dynamics

Temporal patterns of triple NO₃ ⁻ isotopes provide insight into the biological N cycling dynamics of the study watersheds. A strongly seasonal pattern of stream δ¹⁵N–NO₃ ⁻ values occurred in the Stage 1, 2, and 3 watersheds (Figure 3). The sharp increase in δ¹⁵N–NO₃ ⁻ values from early April through late August may have resulted from mass-dependent fractionations during nitrification and/or plant uptake. This increase was likely not due to denitrification, as a concomitant increase in δ¹⁸O–NO₃ ⁻ values (characteristic of denitrification activity; Kendall 1998) did not occur in any of the streams. Although mineralization of organic N to NH₄ ⁺ does not result in large isotopic fractionations (Kendall 1998), oxidation of NH₄ ⁺ to NO₃ ⁻ can yield fractionations ranging from −12‰ to −29‰ (Shearer and Kohl 1986). Thus, as the ratio of net nitrification to net mineralization increases, the δ¹⁵N–NO₃ ⁻ in soils should approach that of the original soil NH₄ ⁺ pool (Spoelstra and others 2007). Net nitrification rates in the Stage 2 and 3 watersheds are high during the growing season—reaching nearly 100% of net N mineralization rates—and soil N pools remain well above 0 g NO₃ ⁻–N m⁻² throughout the year, indicating that microbial NO₃ ⁻–N production exceeds plant uptake in these watersheds (Gilliam and others 2001). Preferential uptake of the lighter ¹⁴N isotope during the growing season may have yielded an isotopically-enriched residual soil NO₃ ⁻ pool (Högberg 1997; Kendall and others 2007; Templer and others 2007) from which stream NO₃ ⁻ was derived. Burns and Kendall (2002) reported δ¹⁵N–NO₃ ⁻ values ranging from +9 to +16‰ in O horizon soils incubated at 22°C compared to +1.5‰ in soils incubated at 4°C. The authors suggested that greater cycling of N through mineralization, nitrification, and immobilization processes may have yielded higher isotope values in the soils incubated at higher temperatures. Similar processes could have contributed to the enrichment in δ¹⁵N values observed during the growing season in the Stage 1, 2, and 3 watersheds at Fernow. Although plant N demand decreases sharply at the end of the growing season, soil microbial processes such as nitrification can continue throughout the dormant season under an insulating snowpack (Campbell and others 2005). Decreased plant N demand at the end of the growing season and throughout the dormant season may similarly explain the lower δ¹⁵N–NO₃ ⁻ values observed in streams during this time, as isotopic fractionation in the residual soil NO₃ ⁻ pool would have been less pronounced. The low δ¹⁵N–NO₃ ⁻ values observed in the Stage 0 watershed (median = +0.7‰), their similarity to precipitation δ¹⁵N–NO₃ ⁻ values (median = +0.3‰), and the absence of seasonal isotopic enrichment in this watershed suggest that biologically mediated isotopic fractionations in the soil NO₃ ⁻ pool were less extensive in this watershed. This is in agreement with previous studies at Fernow that reported much lower rates of net nitrification in the Stage 0 watershed compared to the Stage 2 and 3 watersheds (Gilliam and others 2001; Kelly 2010).

In contrast to the seasonal enrichment of stream δ¹⁵N–NO₃ ⁻ values in the Stage 1, 2, and 3 watersheds, stream δ¹⁸O–NO₃ ⁻ values were generally lower during the growing season in all watersheds (Figure 3; Table 3). The absence of a seasonal enrichment in stream δ¹⁸O–NO₃ ⁻ values similar to that observed for δ¹⁵N–NO₃ ⁻ is unexpected, as both are subject to mass-dependent fractionation during biological processes and exhibit similar isotope effects during biological assimilation (Granger and others 2004; Deutsch and others 2009). It is possible that a seasonal enrichment in δ¹⁸O–NO₃ ⁻ values did occur similar to that observed in δ¹⁵N–NO₃ ⁻ values, but the effect was obscured by the concomitant decline in proportions of atmospheric NO₃ ⁻ in streams during the growing season. Due to the greater disparity in δ¹⁸O–NO₃ ⁻ values between atmospheric and microbial sources relative to the differences in δ¹⁵N–NO₃ ⁻ values, it is possible that decreasing proportions of atmospheric NO₃ ⁻ in streams obscured any uptake-induced enrichment that may have occurred in δ¹⁸O–NO₃ ⁻ values. In addition, Burns and Kendall (2002) attributed greater variation in δ¹⁵N values compared to δ¹⁸O values of NO₃ ⁻ in a soil incubation experiment to differences in reaction pathways of nitrogen and oxygen during microbial NO₃ ⁻ processing. Whereas the original isotopic signature of N is partially preserved during N cycling processes, the isotopic composition of oxygen is reset with each nitrification cycle (Burns and Kendall 2002). The oxygen isotopic composition of microbial NO₃ ⁻ is reset through variable contributions of oxygen atoms from soil water and O₂ in soils during nitrification (Buchwald and Casciotti 2010; Snider and others 2010); such exchange processes may have also contributed to the divergent seasonal patterns in δ¹⁵N and δ¹⁸O of NO₃ ⁻ observed in the study watersheds.

In contrast to the seasonal pattern in δ¹⁸O–NO₃ ⁻ values, Δ¹⁷O–NO₃ ⁻ values in weekly stream samples from the Stage 1, 2, and 3 watersheds were less variable from January to June (Figure 3), although significant seasonal differences were observed in the Stage 1 and 2 watersheds (P < 0.05; Table 3). Consequently, estimates of atmospheric NO₃ ⁻ proportions in streams calculated using δ¹⁸O–NO₃ ⁻ were higher during the dormant season than proportions calculated using Δ¹⁷O–NO₃ ⁻ values (Figure 4). The greatest differences between δ¹⁸O- and Δ¹⁷O-based estimates occurred in January and February, when δ¹⁸O–NO₃ ⁻ values were highest in streams; δ¹⁸O-based estimates were up to 12% higher than Δ¹⁷O-based estimates during this time. This may be attributable to seasonal differences in the δ¹⁸O values of water used by soil microbes during nitrification, as previously discussed. If seasonal variations in the oxygen isotopic composition of atmospheric NO₃ ⁻ alone were responsible for the higher dormant season δ¹⁸O–NO₃ ⁻ values observed in streams, this would have also been reflected in the pattern of Δ¹⁷O–NO₃ ⁻ values of stream samples in all of the watersheds. That the δ¹⁸O-based estimates of atmospheric NO₃ ⁻ in streams are greater than those based on Δ¹⁷O–NO₃ ⁻ and show a more distinctly seasonal pattern suggests that biologically mediated isotope fractionations resulted in the biased δ¹⁸O-based estimates of atmospheric NO₃ ⁻. Previous studies have also attributed higher δ¹⁸O-based estimates of atmospheric NO₃ ⁻ in streams (compared to Δ¹⁷O–NO₃ ⁻-based estimates) to the influence of microbial activity (Riha and others 2014). These results demonstrate the advantage of using Δ¹⁷O–NO₃ ⁻ over δ¹⁸O–NO₃ ⁻ for more precise apportionment of atmospheric and biological NO₃ ⁻ contributions to streams.

Figure 4

Estimated percent atmospheric NO3⁻ in weekly stream samples in hardwood watersheds at Fernow using Δ¹⁷O and δ¹⁸O of NO₃ ⁻ values in a two end-member isotope mixing model. δ¹⁸O-based percentages over-estimate atmospheric NO₃ ⁻ contributions to streams during the dormant season (shown by circles), whereas percent estimates in growing season samples (shown by crosses) plot closer to the 1:1 line (shown by the dotted line) in all watersheds.

Conclusions

Four nearly adjacent watersheds receiving identical atmospheric NO₃ ⁻ inputs but differing significantly in long-term patterns of stream NO₃ ⁻ export showed variable δ¹⁵N, δ¹⁸O, and Δ¹⁷O dynamics of stream NO₃ ⁻ during 2010. Our results indicate that:

• (1) Atmospheric deposition can serve as the dominant source of NO ⁻₃ in some streams, particularly in watersheds characterized by low microbial nitrification rates and low stream NO ⁻₃ concentrations.

• (2) Both flux-based and concentration-based estimates should be considered when evaluating NO ⁻₃ source contributions to streams, as the relative importance of atmospheric NO ⁻₃ may differ depending on which approach is used.

• (3) Watershed hydrologic status influences the export of atmospheric NO ⁻₃ to streams, with greater proportions of atmospheric NO ⁻₃ in stormflow than baseflow.

• (4) Nitrate source apportionments based on δ¹⁸O–NO ⁻₃ can over-estimate contributions from atmospheric deposition relative to Δ¹⁷O–NO ⁻₃ -based estimates. Because it is not affected by mass-dependent fractionations that commonly occur during biological processes, Δ¹⁷O–NO ⁻₃ is a more robust and conservative tracer of atmospheric NO ⁻₃ than δ¹⁸O–NO ⁻₃ .