Topography plays a critical role in controlling rates of nitrogen (N) transformation and loss to streams through its effects on reaction and transport, yet few studies have coupled measurements of soil N cycling within a catchment to hydrologic N losses and sources of those losses. We examined the processes controlling temporal patterns of stream N export using hydrometric methods and dual isotopes of nitrate (NO3−) in a small headwater catchment on the coast of Northern California. Soil nitrate pools accumulated in the hollow during the dry summer due to sustained rates of net nitrification and elevated soil moisture, and then contributed to the first flush of NO3− in macropore soil-water and stream water in the winter. Macropore soil-waters had higher concentrations of all forms of N than matrix soil-waters, especially in the hollow. A plot of stream water δ15N versus δ18O values in NO3− indicated that NO3− was primarily derived from nitrification or microbial NO3−. Further analysis revealed a mixing of two microbial NO3− sources combined with seasonal progressive denitrification. Mass balance estimates suggested microbial NO3− was consumed by denitrification when conditions of high NO3−, dissolved organic matter, and soil-water contents converged. Our study is the first to show a mixing of two sources of microbial NO3− and seasonal progressive denitrification using dual isotopes. Our observations suggest that the physical conditions in the convergent hollow are important constraints on stream N chemistry, and that shifts in runoff mechanisms and flow paths control the source and mixing of NO3− from various watershed sources.
 Interactions among carbon (C), nitrogen (N), and hydrologic cycles exert strong controls on sources and export of nitrate (NO3−) to headwater streams [Lohse et al., 2009]. In particular, the balance between plant-microbial reactions and the hydrologic transport of reaction products determine the amount and form of N that enters streams [e.g., Ocampo et al., 2006a, 2006b]. Studies have shown that the hydrological coupling of subsurface flow between upland and riparian zones within a catchment determines catchment-scale NO3− removal [Vidon and Hill, 2004], nutrient dynamics [Stieglitz et al., 2003], and dissolved organic carbon (DOC) dynamics [McGlynn and McDonnell, 2003]. However, quantifying the interactions between flow paths and hydrologic and biogeochemical controls on N reaction and transport has been inherently challenging owing to issues of linking spatially and temporally variable sources of both water and N in a watershed.
 Ecosystems that have excess N can negatively affect water quality or add to greenhouse gas emissions [Vitousek et al., 1997]. In highly seasonal climates, like semiarid and Mediterranean ecosystems, N limitations or excesses may be both highly temporally and spatially variable, and water can both directly and indirectly influence N transformations [Austin et al., 2004; Davidson et al., 1991]. When soils are rewet following long dry periods, rapid increases in C and N cycling can occur, creating excessive N availability [Austin et al., 2004; Davidson et al., 1993; Fierer and Schimel, 2002] and enhancing the potential for solution loss to streams [Lohse et al., 2009]. These pulses have also been described as hot moments—periods characterized by disproportionately high rates of biogeochemical cycling [McClain et al., 2003]. The products of these periods then pass through a complex physical system and experience a variety of subsequent processes, such that the biogeochemistry of the solutions that eventually enter streams are the result of temporally and spatially distinct events.
 Dual isotopes of NO3− in combination with hydrometric techniques and other tracers have been successfully used to distinguish sources and processing of N in surface water and catchments owing to differences in the stable isotope ratios of the two main NO3− sources, atmospheric deposition and microbial nitrification, the microbial process of converting ammonium (NH4+) to NO3− (referred to microbial NO3− hereafter) [e.g., Barnes et al., 2008; Burns and Kendall, 2002; Kendall et al., 2007; Mayer et al., 2002; Pardo et al., 2004] (see Curtis et al.  for review). The δ18O values of NO3− in atmospheric deposition in units of per mil (‰) relative to Vienna Standard Mean Ocean Water (VSMOW) are typically high (>60‰) based on a more recent survey of precipitation analyzed using the denitrifier method [Elliot et al., 2007; Kendall et al., 2007]. The δ18O values of microbial NO3− are thought to be derived from water and soil-atmosphere oxygen (O2) with two oxygen atoms derived from water (H2O) and one from O2 [Anderson and Hooper, 1983; Hollocher, 1984; Kumar et al., 1983], with theoretical δ18O-NO3− values from nitrification ranging from −10 to +10‰, assuming δ18O-H2O of −25 to −5‰ and δ18O-O2 of +23.5‰ [Kroopnick and Craig, 1972], though values have been reported as high as +16‰ [Snider et al., 2010]. The theoretical range for δ18O values of NO3− assumes that (1) no fractionation occurs during incorporation of O from water or O2, (2) the proportion of O from H2O to O2 is the same in soil as is observed in lab cultures, (3) the δ18O signature of H2O that is used by microbial NO3− is the same as bulk water, and (4) δ18O of soil O2 used by microbes is equal to δ18O of the atmosphere [Kendall, 1998; Kendall et al., 2007].
 Numerous challenges have emerged with applying the dual isotope method to NO3− source identification (see Xue et al.  for review). In particular, the assumptions behind the sources of oxygen in microbial NO3− have been increasingly questioned. For example, recent studies indicate that oxygen exchange between water and intermediates in the nitrification pathway alters the oxygen isotope ratio of NO3− [Buchwald and Casciotti, 2010; Casciotti et al., 2010; Kool et al., 2011; Snider et al., 2010]. Other studies suggest the possibility that the δ18O value of O2 in the soil may be increased relative to atmospheric O2 due to fractionation associated with soil respiration [Mayer et al., 2001; Spoelstra et al., 2007]. Another challenge is that studies have shown that atmospheric NO3− [e.g., Pardo et al., 2004] or anthropogenic sources of reduced or organic forms of nitrogen (fertilizer, manure, septic waste) [Barnes and Raymond, 2010; Burns et al., 2009] can be readily taken up, processed, and oxidized to NO3− by microbes so that the δ18O-NO3− values fall within the microbial NO3− range even though the NO3− is not derived from natural sources such as soil. An additional complicating factor is that under anaerobic conditions, NO3− can be used as a terminal electron acceptor in the process of denitrification where NO3− is reduced to gaseous products of N and reduced forms of C such as dissolved organic matter (DOM) are oxidized to carbon dioxide (CO2). Denitrification results in isotopic enrichment of both δ15N-NO3− and δ18O-NO3− values, often observed in a ratio of 2:1 [Aravena and Robertson, 1998; Bottcher et al., 1990; Kendall, 1998; Kendall et al., 2007], although a 1:1 ratio has been proposed based on denitrification culture studies [Sigman et al., 2005] and evidence of possible biased δ18O-NO3− values with the earlier offline reaction techniques [Revesz and Bohlke, 2002]. Finally, in theory, sources of NO3− and mixing of those sources can be separated using mixing models and dual isotopes of NO3− [Kendall, 1998]; however, this has not been empirically demonstrated because of spatially and temporally variable sources of both water and NO3− in a watershed.
 In this study, we examined the patterns of soil solution and stream N losses in a small, coastal zero-order catchment in Marin County, California. Specifically, we were interested in understanding how the spatial and temporal patterns of hydrological flow along a topographic gradient relate to the sources of stream NO3− concentrations. Previous findings by Sanderman et al.  showed that shifts in soil-water sources in this catchment explained changes in biochemical characteristics and 14C age of the DOM delivered to the stream. We, therefore, expected that shifts in runoff mechanisms and putative flow paths over time would have significant impacts on the delivery of NO3− sources to streams. In addition, findings from Sanderman et al.  indicated that upland soils become DOM production limited, whereas the highly productive soils in the hollow of the zero-order catchment act as a near-infinite DOM source, thus providing a reduced energy source for microbially mediated N transformations such as denitrification. Based on these findings, we expected higher and sustained rates of N cycling in the convergent hollow compared to the divergent backslopes and toeslopes in this seasonally water-limited environment. In particular, we expected higher rates of denitrification in the hollow compared to other positions owing to higher DOC, NO3−, and soil-water contents.
2. Site Description
 This study was conducted in a well characterized, intensively instrumented, 3.9 ha first-order watershed located 2 km from the coast in the Tennessee Valley (TV) area of the Golden Gate National Recreation Area in Marin County, CA (37°51.5″N, 122°32.2″W) [Montgomery and Dietrich, 1989, 1995; Sanderman and Amundson, 2009, 2010; Sanderman et al., 2008, 2009; Yoo et al., 2009] (Figure 1). The area experiences a Mediterranean climate with mean annual air temperature of 15.3°C and mean annual precipitation of 1150 mm. Winter storms are typically long-duration events (12–36 h) of low intensity (2–5 mm/h) that originate from moist Pacific air masses. For the 2005 and 2006 hydrologic years (HY), beginning on 1 October 2004 and ending on 30 September 2006, precipitation measured at our site was 1260 and 1540 mm, respectively. A coastal vegetation community consisting of a mixture of annual European grasses, various forbs, woody scrubs, and to a lesser extent native perennial grasses dominates this landscape [Sanderman et al., 2009]. A previous study of atmospheric N deposition on the California coast (approximately 95 km north of this site) documented low atmospheric N inputs, an average of 1.62 kg N/ha/yr as dissolved inorganic N (DIN), during this same study period [Ewing et al., 2009]. No anthropogenic N inputs as fertilizer, septic or manure sources, other than burrowing gophers and other wildlife, exist in the catchment.
 Extensive hydrologic and geomorphic characterization of this site has shown three distinct, sequentially initiated, runoff generating regimes depending on rainfall dynamics, and antecedent soil moisture conditions [Montgomery and Dietrich, 1989, 1995; Sanderman et al., 2009]. During unsaturated conditions, water slowly infiltrates the soil matrix as throughflow and follows water pressure gradients through the subsurface to the channel head. After the subsurface soil becomes saturated, water moves as macropore flow along cracks and through gopher burrows to the channel head. Finally, water moves as saturation overland flow (SOF), a combination of direct precipitation onto saturated areas and return flow, defined as water that returns to the surface and flows along the ground surface to the channel head after the ability of the soil matrix and macropore network to transmit water has been exceeded [Montgomery and Dietrich, 1995]. Strong hydraulic conductivity contrasts exist at both the soil/saprolite and saprolite/bedrock contacts [Montgomery and Dietrich, 1995] and create a seasonal shallow ground water table above the bedrock and a transient perched water table above the saprolite in response to large winter storms. The largest peak flows occur when there is direct rainfall onto already saturated ground, SOF as defined by Dunne and Black  (Figure 1b). Saturation overland flow occurred two and seven times during the winter months of HY 2005 and 2006, respectively.
 Soils are classified as Typic Haplustolls. Soil profiles were previously described at seven positions along a topographic sequence from ridge to hollow in the study catchment; details are provided by Sanderman et al. . Upland soil profiles in the catchment have a prominent series of dark mineral (A) horizons, averaging 40 cm but thickening downslope, that appear to be well mixed by burrowing gophers and other animals [Yoo et al., 2005]. Within the hollow, colluvial soils are generally thicker, show little vertical variation in physical properties with depth, and contain significantly more organic C than on the hillside [Sanderman et al., 2009]. Depth to bedrock varies from ∼20 cm on the ridges to >2 m in the hollow.
3.1. Hydrometric, Soil Solution, and Stream Monitoring
 The catchment was instrumented for hydrometric monitoring and soil solution collection along a toposequence from ridge to toeslope and within the adjacent hollow prior to this study [Sanderman et al., 2009]. In brief, three tipping bucket rain gauges (Spectrum Instruments Inc., Plainfield, IL) were dispersed along the slope to record rainfall at hourly intervals. Rainwater was collected for analysis of atmospheric deposition (bulk, including both wet and dry) in high-density polyethylene (HDPE) bottles with glass wool-lined funnels placed on stakes at three positions above the vegetation canopy along the slope of the catchment (n = 21 rainwater sample events). Similarly, throughfall (TF) was collected at a similar frequency in high-density polyethylene (HDPE) bottles using glass wool-lined funnels dispersed under shrub canopies at three positions along the slope (n = 19 TF). An additional subset of rainwater and throughfall samples with insects and other detritus were discarded and not analyzed for chemistry (n = 5). At seven locations along the toposequence, soil matric potential in units of pressure head (cm) was monitored continuously using Spectrum Watermark sensors (Spectrum Technologies, Aurora, IL). Adjacent to the Spectrum sensors, an array of ceramic cup tensiometers at depths of 15, 30, 60, and 90 cm were installed, fitted with silicon bungs and monitored manually with a needle pressure transducer (SDEC, France).
 Zero-tension pan lysimeters (ZT) and tension lysimeters (PT) were used to quantify macropore and matrix soil-water chemistry, respectively. We used 150 cm2 ZT lysimeters installed at 20 cm depth in all hillslope locations, in addition to samplers installed at 20, 35, and 50 cm in the hollow (site 6) to sample macropore soil-water. We used Prenart Super Quartz tension (PT) lysimeters (Prenart Equipment ApS, Denmark) installed (1) at a depth of 10 cm at the ridge, (2) 20 cm at the shoulder, (3) at 20 and 50 cm on the backslope (hereafter sites 1–3 referred to the backslope locations collectively), and (4 and 5) 20, 50, and 100 cm depths at the two toeslope and (6 and 7) upper and lower hollow positions to sample matrix soil-water (Figure 1). Throughout HY 2005 and 2006, matrix and macropore soil-waters were collected at multiple depths along the hillslope and within the hollow following storm events (n = 303 ZT, n = 146 PT).
 Discharge (Q) was monitored using a small 120° V-notch weir installed 5 m below the channel head. Stage (water height in the weir, cm) was continuously recorded with a Heron Dipper-log pressure transducer (Heron Instruments, Inc., Ontario, Canada). Water samples were collected by hand in acid-washed HDPE bottles. Samples were collected throughout the course of the rainy season at all discharge levels for HY 2005 and 2006 but with increased intensity during storm events for HY 2006 (n = 80 stream, 15 SOF).
3.2. Solution Analytical Procedures
 All water samples were filtered to <0.7 mm using Whatman GF/F glass fiber filters (Whatman International Ltd., UK) within 48 h of collection and stored at 4°C in opaque acid-washed HDPE bottles until analysis. Rainwater, throughfall, soil-water, and stream water were analyzed for ammonium-N (NH4+, values reported as N) following the phenol seawater method [Alpkem, 1992b] on an Alpkem continuous-flow colorimetric autoanalyzer (Astoria-Pacific, Clackamas, Oregon). Method detection limit (MDL) was 1 µg/L for NH4-N in water. Samples were analyzed for NO3−-N and nitrite-N (NO2−-N) using the NO3− reduction method with cadmium (Cd) metal for NO3− + NO2 on the Alpkem autoanalyzer [Alpkem, 1992a]; NO2− was analyzed separately without the Cd column. Nitrite concentrations in all samples were <0.01 mg N/L and reported as NO3− concentrations (NO3− + NO2−) hereafter. Method detection limit (MDL) was 1 μg/L for NO3-N in water. Dissolved organic nitrogen (DON) as N was calculated as the difference between total dissolved nitrogen (TDN) and DIN (NH4+ + NO3− + NO2−). Total dissolved nitrogen was measured by NO3− colorimetric analyses following persulfate oxidation of all N forms to NO3− [D'Elia et al., 1977]. The MDL for TDN was 8 μg/L measured as NO3−-N. Major anions, chloride (Cl−) and sulfate (SO42−), were analyzed on a DIONEX ion chromatograph 3000 equipped with AS17 column (DIONEX, San Jose, CA). Major cations, calcium (Ca2+) and magnesium (Mg2+), were analyzed on a Perkin Elmer inductively coupled plasma-optical emission spectroscopy (Perkin Elmer, Waltham, MA). The analytical precision for all forms of N, anions and cations was 10% for samples less than 1 mg/L and less than 5% for samples less than 1 mg/L.
3.3. Source Waters
 To link rainwater, macropore and matrix soil-water chemistry to stream water chemistry, we employed diagnostic tools for mixing models of stream chemistry to determine end-members [Christophersen et al., 1993; Hooper, 2003; Hooper et al., 1990]. Four solutes were considered suitable for end-member mixing: Cl−, SO42−, Ca2+, and Mg2+. End-members were selected by determining those soil solutions whose concentrations bounded the stream water concentrations in two-dimensional (2-D) plot of one solute against another. Bi-plots of all possible combinations of solutes were visually inspected to determine which combination had end-members that bounded all stream chemistry observations. From this exercise, one combination of solutes, Mg2+ and Ca2+, was identified that bounded the stream water observations. This approach assumed that the stream water arises from a mixture of soil-water solutions, which to a first approximation are considered to be invariant in time and space [Hooper et al., 1990].
3.4. Dual Isotopes of Nitrate
 A select set of soil solution and stream water samples from HY 2006 were sent to UC Davis for dual isotope analysis of NO3− following the denitrifier method described by Sigman et al.  for δ15N in NO3− and by Casciotti et al.  for δ18O in NO3−. Samples were analyzed for δ15N and δ18O using a ThermoFinnigan GasBench + PreCon trace gas concentration system interfaced to a ThermoScientific Delta V Plus isotope-ratio mass spectrometer (Bremen, Germany). Values are reported in parts per thousand (‰) relative to atmospheric N2 and VSMOW, for δ15N and δ18O, respectively, using the equation:
where Rsample = ratio heavy to light isotope (e.g., 15N/14N) of a sample.
 Samples were corrected using international reference standards USGS32, USGS34, USGS35 supplied by NIST (National Institute of Standards and Technology, Gaithersburg, MD). Analytical precision was 0.4‰ for δ15N-NO3− and 0.8‰ for δ18O-NO3−.
 Stream water δ15N-NO3− and δ18O-NO3− values were plotted with theoretical distributions of rainwater, fertilizer, and soil N, with the theoretical atmospheric δ18O-NO3− values plotted in the range observed for the denitrifier method [Kendall et al., 2007]. Diagnostic tests were used to evaluate mixing and sources of stream δ15N-NO3− and δ18O-NO3− values [Kendall, 1998; Kendall et al., 2007]; stream and soil-water δ15N-NO3− values were plotted against the inverse of NO3− concentrations to evaluate mixing and then plotted against the natural log of NO3− concentrations to evaluate fractionation associated with denitrification [Kendall, 1998]. In addition, δ18O-NO3− values were plotted against δ18O-H2O values of different possible source waters to determine whether in situ nitrification was associated with ambient waters [Wankel et al., 2006; McMahon and Bohlke, 2006]. Stable isotope analyses (δD and δ18O) on a subset of stream, soil solution, and rainwater samples were performed on a liquid water stable isotope analyzer DLT-100 (Los Gatos Research, Mountain View, CA) with 0.2‰ and 0.5‰ analytical precision for δ18O and δD, respectively, using VSMOW as a reference standard. We used these values to predict δ18O-NO3− values based on the theoretical microbial nitrification model presented by Kendall  that assumes that in situ nitrification uses two oxygen atoms from water and one oxygen atom from O2 [Anderson and Hooper, 1983; Hollocher, 1984; Kumar et al., 1983] where:
 The δ18O-NO3− values were predicted based on three possible sources of δ18O-H2O: (1) rain water using a range of samples from the local meteoric water line (LMWL) generated for this study period, (2) soil-water, or (3) stream water, and we assumed a δ18O-O2 value of +23.5‰ [Kroopnick and Craig, 1972].
3.5. Soil Nitrogen Pools and Transformation Rates
 Soil inorganic N pools and processes were measured seasonally along a toposequence from backslope to toeslope and along the axis of the hollow to help to explain patterns of soil solution and stream losses. Specifically, the surface soils (0–10 cm depth) were sampled along the toposequence in the early wet winter season (January 2005), spring (April 2005), and late dry summer/early autumn (August/September 2005) to determine soil exchangeable mineral N pools and transformation rates of net mineralization and nitrification. Five soil cores were collected and composited from each of the seven topographic positions (two hollow, two toeslope, and three backslope positions, Figure 1), kept on ice in the field, and then processed immediately in the lab. Soils were sieved to <2 mm, and one 25 g soil subsample was dried at 105°C to determine gravimetric soil moisture. Another 15 g subsample was extracted with 75 mL of 2N potassium chloride (KCl) to determine exchangeable mineral N pools and then another soil subsample was incubated under aerobic conditions for 7 days in the dark to determine rates of net mineralization and nitrification following methods described by Lohse and Matson . Extracts were analyzed on the Alpkem for NH4+-N and NO3−-N using the salicylate TDN and NO3− reduction methods [Alpkem, 1992a, 1992c], respectively. One subsample from the summer season was incubated for 4 h under anaerobic conditions using an acetylene block method with amendments of NO3− and C as described by Groffman et al.  to determine denitrifying enzyme activity (DEA), a measure of the maximum rate of denitrification possible under ideal conditions. Archived soils from the seven soil profiles described above were analyzed for total N and δ15N on a Finnigan Delta S isotope ratio mass spectrometer with a front-end Carlo Erba autoanalyzer at UC Berkeley (Thermo Finnigan, Bremen, Germany).
 To determine differences in soil nitrogen cycling among topographic position and season, we employed repeated measures analysis of variance (RMANOVA) on soil properties and processes such as gravimetric soil moisture, ammonium, and nitrate pools. For these analyses, we grouped toeslope, hollow, and backslope positions (n = 2 and 3). Analysis of means on transformed ranks was performed to determine significant differences among position and season (p < 0.05). All statistical analyses carried out with JMP 9.0.2 statistical software (SAS, Cary, NC).
3.6. Mass Balance
 Nitrogen mass balance was estimated for HY 2006 for which we had higher temporal resolution data. We estimated bulk N deposition for NO3− and TDN by multiplying average NO3− and TDN concentrations ± standard errors to the observed rainfall (mm/h) and summing the total NO3− and TDN load for the water year. Stream export for NO3− and TDN was estimated using log concentration-log discharge relationship to predict NO3− and TDN concentrations with associated discharge. The maximum annual production of soil microbial NO3− in the hollow surface soils, defined here as the amount of NO3− produced from soil nitrification, was estimated from seasonal rates of net nitrification and the size of the hollow as defined by maximum extent of saturated area (0.34 ha) in the hollow shown in Montgomery and Dietrich , as well as the extent of the colluvium deposits (1.4 ha) [Montgomery and Dietrich, 1995]. Rates were assumed constant over each seasonal time period. Maximum potential removal of NO3− within the hollow was estimated from potential denitrification assays, maximum extent of saturated area (0.34 ha), depth (2.5–5 m) of the hollow [Montgomery and Dietrich, 1995], and a range of days of saturation within the hollow (30–365 days), based on the observation by Montgomery and Dietrich  that positive piezometric potential occurred throughout the year in the deepest portion of the hollow indicating that the hollow remained saturated at depth even during drought periods.
4.1. Rainfall and Throughfall Chemistry
 Rainfall and TF concentrations in the coastal catchment were quite low for all forms of N, anions, and cations relative to matrix and macropore soil-waters and stream chemistries (Figure 2). Specifically, NO3− was the dominant form of N in rainfall followed by DON and then NH4+, with concentrations averaging 0.17 ± 0.09 mg N/L for NO3− (mean ± standard deviation), 0.05 ± 0.05 mg N/L for NH4+, and 0.14 ± 0.17 mg N/L for DON. In contrast, DON was the dominant form of N in TF averaging 0.35 ± 0.34 mg N/L, whereas NO3− was lowest with a mean value of 0.034 ± 0.031 mg N/L followed by NH4+ with a mean of 0.15 ± 0.39 mg N/L. Concentrations in rainfall averaged 4.97 ± 2.33 mg/L for Cl−, 1.23 ± 0.36 mg/L for SO42−, 0.25 ± 0.22 mg/L for Ca2+, and 0.34 ± 0.21 mg/L for Mg2+, and increased in TF, averaging 8.79 ± 4.78 mg/L for Cl−, 2.85 ± 1.87 mg/L for SO42−, 0.77 ± 0.42 mg/L for Ca2+, and 0.52 ± 0.26 mg/L for Mg2+ (Figure 2).
4.2. Spatial Patterns of Soil Solution and Associated Stream Chemistry
4.2.1. Macropore Soil-Waters
 Macropore soil-water as measured by zero tension (ZT) lysimeters showed strikingly different patterns in solute chemistry compared to matrix soil-water as measured by Prenart tension (PT) lysimeters, particularly for different forms of N (Figure 2). Indeed, macropore soil-water concentrations for all forms of N were much higher than matrix soil-water solution concentrations across all positions. Ammonium concentrations were relatively low in macropore solutions compared to other forms of N, ranging from as high as 0.031–4.73 mg/L on the ridge (ZT 1) to 0.023–0.12 mg/L in the toeslope positions (ZT 4–5). Nitrate concentrations were the highest, and also the most variable in the hollow (range: 0.005–46.70 mg N/L) compared to other positions—particularly the toeslope positions (ZT 4–5), where concentrations ranged from 0.002 to 1.79 mg N/L (Figure 2b). Dissolved organic nitrogen was high as a proportion of total dissolved nitrogen (TDN) in the backslope positions (64 ± 4%) and in the hollow (71 ± 2%) (Figure 2c). In the hollow (ZT 6), average NO3− concentrations increased from 1.76 ± 4.94 mg N/L at 10 cm depth to 3.52 ± 9.70 mg N/L at 20 cm depth and then decreased to 0.66 ± 1.52 mg N/L at 35 cm and 1.18 ± 2.58 mg N/L at 50 cm depth (Figure 2b).
 Macropore soil-water concentrations of other major anions and cations—Cl−, SO42−, Ca2+, and Mg2+—increased relatively systematically from backslope positions (ZT 1–3), toeslope (ZT 4–5), hollow (ZT 6–7), to stream (Figures 2d–2g). In contrast to NO3− and DON that declined with depth in the hollow, Cl−, SO42−, Ca2+, and Mg2+ in macropore solution all increased slightly with depth (Figures 2d–2g). Average Cl− concentrations increased from 17.23 ± 15.61 at 10 cm depth to 26.13 ± 17.41 mg/L at 50 cm depth, for example, whereas concentrations increased slightly from 4.82 ± 3.47 to 7.44 ± 2.70 mg/L from 10 cm to 50 cm depth for SO42−, 5.10 ± 6.84 to 5.61 ± 2.06 mg/L for Ca2+, and 6.22 ± 8.28 to 7.14 ± 2.62 mg/L for Mg2+.
4.2.2. Matrix Soil-Waters
 In contrast to forms of N, which were lower in the matrix soil-waters than macropore solutions (range 0.002–0.73 mg N/L for NO3−) (Figures 2a–2c), concentrations of major anions and cations in matrix soil-waters typically increased with depth and were higher than the macropore soil-waters, especially in the deep matrix soil-water at the toeslope (100 cm at PT 4) (Figures 2d–2g). For example, at PT 4, the toeslope position, average Cl− concentrations in matrix soil-water increased from 18.52 ± 2.25 to 73.46 ± 18.60 mg/L from 10 cm depth to 50 cm depth, 4.91 ± 1.01 to 12.20 ± 1.62 mg/L for SO42−, 4.78 ± 1.13 to 17.88 ± 3.67 mg/L for Ca2+ (maximum 23.7 mg/L), and 5.93 ± 1.51 to 22.10 ± 4.39 mg/L for Mg2+ (maximum 28.86 mg/L).
4.2.3. Stream Water and Saturation Overland Flow
 Stream water chemistry showed overlap with SOF and the hollow shallow and deep macropore soil-waters for all solutes (Figure 2), although SOF had slightly higher Mg2+ to Ca2+ concentrations than stream chemistries, and concentrations of NO3−, DON, and TDN were lower in SOF than stream water (Figure 2). Average stream NO3− concentrations, for example, were 0.57 ± 0.31 mg N/L compared to 0.43 ± 0.45 mg N/L in SOF (Figure 2b), whereas average TDN concentrations were 0.95 ± 0.44 mg N/L in the stream compared to 0.78 ± 0.36 mg N/L in SOF (TDN not shown). In contrast, average stream concentrations for Mg2+ and Ca2+ were 7.06 ± 1.13 mg/L and 4.53 ± 0.75 mg/L, respectively, compared to 6.93 ± 1.67 mg/L for Mg2+ and 4.56 ± 1.00 mg/L for Ca2+ in SOF (Figures 2f and 2g). Stream NO3− and TN concentrations were correlated (r = 0.79), and DON was correlated to TN (r = 0.74) and Cl− (r = 0.49). Stream concentrations of Ca2+ and Mg2+ were highly correlated (r = 0.94) across the full range of discharges. Otherwise, other major anion and cations showed weaker correlations with each other (r = <0.25).
4.3. Sources of Stream Water and Chemistry
 Three end-members, precipitation as rainfall, deep toeslope soil-water, and SOF, were required to bound the stream water observations (Figure 3). A mixing diagram with all stream water points in the interior of the triangle described a mixture of these three solutions for the two solutes plotted, Ca2+ and Mg2+. Two end-member mixing of deep toeslope soil-matrix waters and rainwater explained 89% of the variation in stream chemistry and all soil solutions, both matrix and macropore (r2 = 0.89, p < 0.0001). Addition of SOF as an end-member helped to explain the residual variance in stream chemistry concentrations. Further examination of soil waters showed substantial overlap of surface and subsurface hollow soil waters with stream waters. Backslope and toeslope soil-waters were a mix of two components, rainwater and deep soil-matrix water in the toeslope (Figure 3b).
4.4. Temporal Patterns of Soil Solution, Saturation Overland Flow, and Stream Chemistry
 In HY 2005, winter storms commenced in October, and very high NO3− concentrations (range: 2.32–46.70 mg N/L as NO3−) as well as DON concentrations (not shown) were observed in surface macropore solution in the hollow (Figure 4), although stream flow did not commence until late December. Subsequent peaks in NO3− concentrations in macropore soil-waters typically corresponded to storm events when pressure heads were close to zero at all depths in the hollow indicating that soils were saturated (Figures 4b and 4c). Macropore soil-water concentrations followed similar temporal patterns in the HY 2006 with high solution losses as NO3− prior to the onset of seasonal stream flow. Pressure heads in the hollow approached zero more often in HY 2006 compared to HY 2005 and displayed more positive pore pressures in HY 2006 compared to HY 2005. Similar to HY 2005, peaks in NO3− concentrations in the macropore soil-water coincided with peaks in saturation. However, in HY 2006, deep soil-water NO3− concentrations were sometimes higher than surface soil-waters possibly due to higher positive pore pressures than in HY 2005. Ammonium and DON concentrations were generally low in the hollow macropore soil-waters throughout the season (<0.08 mg/L for NH4+ and <0.6 mg N/L for DON) with the exception of first flush responses (NH4+: range: 0.12–3.15 mg N/L, DON: 1.35–19.11 mg N/L).
 On 28 December 2004 in HY 2005, precipitation resulted in near-saturated soil moisture conditions in the hollow as indicated by pressure heads close to zero at all depths (Figure 4b) and resulted in the onset of seasonal stream flow in the channel and a flush of high NO3− and DON in soil-water and stream (Figures 4c and 4d). Ammonium and NO2− concentrations in stream flow were negligible (<0.01 mg N/L). Subsequent peaks in stream water NO3− concentrations during the winter and early spring corresponded to the saturation of the hollow and/or contributions from SOF (Figure 4). Hydrologic year 2006 followed similar patterns but stream flow commenced on 18 December 2005, and a large flood, the flood peak falling between a 25–50 year return interval, occurred on 31 December 2005 and delivered 210 mm of rain in 28 h on soil that had already received 330 mm rain in the prior 2 weeks.
 Nitrate and TDN concentrations in the stream were significantly correlated with discharge (NO3−: r2 = 0.38, TDN: r2 = 0.42, p < 0.0001) (Figure 5). Stream water NO3− and TDN concentrations increased with increasing discharge following the first flush of the hollow, and then showed dilution of NO3− with the early storms (e.g., 31 December 2005 to January, 2006), and increasing NO3− concentrations with increasing discharge in mid and late storm events (February to May). DON concentrations varied similarly and were also significantly correlated to discharge (r2 = 0.2, p < 0.0001) (data not shown) whereas NH4+ remained low and invariant across the range of discharge (0.01–0.04 mg N/L) (r2 = 0.02).
4.5. Analysis of Dual Isotopes of NO3− in Stream Water and Hollow Soil Solution
 Stream water δ15N-NO3− and δ18O-NO3− values ranged from +1 to +12‰ and −2 to +10‰, respectively. A plot of stream water δ15N-NO3− and δ18O-NO3− values with theoretical distributions of rainwater, fertilizer, and soil NH4+, rainwater NH4+, and septic and manure sources (Figure 6) showed that stream water δ15N-NO3− and δ18O-NO3− values plotted linearly in a ratio close to 1 (slope = 0.9). Further dual isotope analysis of soil-waters showed that δ15N-NO3− and δ18O-NO3− values ranged from +1.0 to 8.4‰ and −2.6 to +5‰, respectively. Initial soil-water δ15N-NO3− and δ18O-NO3− values from the hollow surface horizon (18–26 December 2005) overlapped with stream water values prior to the flood event on 31 December 2005 (Figure 6b). After this flood event, both surface and subsurface soil-water δ15N-NO3− and δ18O-NO3− values plotted linearly along a line with a slope of 0.50 ± 0.1. Deep soil-water from the hollow δ18O-NO3− values plotted higher than surface soil-water values from the hollow, +2.6 to +5.0‰ compared to −2.6 to −1.2‰, an offset of ∼5‰ and constrained the stream water δ15N-NO3− and δ18O-NO3− values (Figure 6b).
 Stream water δD-H2O and δ18O-H2O ranged from −28.3 to −31.9‰ and −4.2 to −5.5‰, respectively, and departed from a local meteoric water line (LMWL) (δD = 7.4 δ18O + 10.0) (Figure 7a). Rainwater δ18O-H2O values ranged from −6.5 to −14.8‰, whereas soil-water δ18O-H2O values ranged from −4.0 to −5.4‰ (mean 4.8‰) and overlapped with stream water values. The δ15N-NO3− in stream water was positively correlated with the inverse of NO3− concentrations (r = 0.66, p = 0.001) (Figure 7b) and was negatively correlated to the natural log of NO3 concentrations (r = 0.63, p = 0.0024) (Figure 7c). Stream δ18O-NO3− values were also somewhat negatively correlated to stream δ18O-H2O values (r = 0.44, p = 0.09) (Figure 7d) and tended to be higher at low discharge (Figure 7e).
 The predicted δ18O-NO3− values based on rainwater δ18O-H2O values of −5.8 to −14.8‰ as source water for in situ nitrification ranged from −2.0 to +4.0‰, the range of shallow soil-water δ18O-NO3− values (Figures 6b and 7f). The predicted microbial δ18O-NO3 values, based on soil-water δ18O-H2O values (−4.0 to −5.5) as the source water, were +4 to +5.5‰, the range for the mid and late subsurface soil solution δ18O-NO3− values (Figures 6b and 7f). Early δ18O-NO3 subsurface soil-water values overlapped with rainwater as the source of water for nitrification. The δ15N-NO3− values in soil-waters after the December flood event were significantly and negatively correlated to the natural log of NO3− concentrations (r2 > 0.9) (Figure 7g).
4.6. Soil N Pools and Processes Explaining Patterns of Solution and Stream NO3 Losses
 Soil N pools and transformation rates varied seasonally and with topographic position (Table 1). Gravimetric soil moisture was highest during the winter months and decreased through the spring to late summer. Soil moisture varied significantly with season (season, p = 0.0036) and topographic position × season (p = 0.068) (Table 2). Soil moisture was similar among topographic positions during the wet winter but significantly higher in the hollow compared to the backslope positions during the late, dry summer (Table 1).
Table 1. Soil Moisture and Nitrogen Pools and Processes With Topographic Position and Seasona
Analysis of means ± standard errors on log-transformed ranks was performed to assess significant differences among position (p < 0.05). Lower case letters indicate significant differences among position.
Gravimetric soil moisture (%)
24.54 ± 2.15
30.81 ± 1.54
29.61 ± 1.13
12.64 ± 4.24
22.21 ± 1.21
26.53 ± 3.23
6.65 ± 1.77b
8.84 ± 0.80ab
14.67 ± 1.31a
Soil exchangeable NH4 (kg N/ha)
2.54 ± 0.31
2.31 ± 0.15
2.81 ± 0.87
1.43 ± 0.08
1.95 ± 0.20
2.06 ± 0.06
5.45 ± 0.53
7.45 ± 0.27
5.34 ± 0.75
Soil exchangeable NO3 (kg N/ha)
0.05b ± 0.01
0.15 ± 0.08b
0.40 ± 0.04a
0.07 ± 0.03
0.56 ± 0.01
0.34 ± 0.14
2.54 ± 0.61
0.81 ± 0.38
10.42 ± 8.43
Net N mineralization (kg N/ha/d)
0.28 ± 0.06
0.27 ± 0.10
0.64 ± 0.36
−0.01 ± 0.02b
0.00 ± 0.04b
0.42 ± 0.02a
0.04 ± 0.08b
0.26 ± 0.11b
0.53 ± 0.07a
Net N nitrification (kg N/ha/d)
0.36 ± 0.03
0.44 ± 0.16
0.94 ± 0.24
0.02 ± 0.02b
0.25 ± 0.08b
0.75 ± 0.02a
0.00 ± 0.00c
0.09 ± 0.06b
1.15 ± 0.18a
Potential denitrification (kg N/ha/d)
0.58 ± 0.05b
0.48 ± 0.1b
0.91 ± 0.04a
Table 2. Results From Repeated Measures ANOVA Tests for Soil Moisture, N Pools, and Process Ratesa
Significant values are indicated in bold.
Gravimetric soil moisture (%)
Position × season
Soil exchangeable NH4 (kg N/ha)
Position × season
Soil exchangeable NO3 (kg N/ha)
Position × season
Net N mineralization (kg N/ha/d)
Position × season
Net N nitrification (kg N/ha/d)
Position × season
 Soil NH4+ pools were elevated in the spring in the hollow and in the toeslope positions compared to the backslope position and were substantially elevated in all positions in late summer (Table 1). Soil exchangeable NH4+ pools varied significantly with season (p = 0.008) and position × season (p = 0.013) (Table 2). Pools of exchangeable NO3− were elevated in the winter season in the hollow relative to toeslope and backslope. Pools of exchangeable NO3− were the highest in the late summer, particularly in the hollow position, averaging 10.42 kg N/ha in the hollow compared to 0.81 kg N/ha in the toeslope (Table 1).
 Soil N transformation rates were higher in the hollow than other topographic positions (Table 1). Net mineralization rates varied significantly with position (p = 0.042) and season (p = 0.047) (Table 2). In particular, net mineralization rates in the hollow samples were significantly elevated relative to toeslope and backslope positions in the spring and late dry summer. Net nitrification rates varied significantly with position (p = 0.0007) (Table 2). Rates of net nitrification were significantly elevated in the hollow relative to toeslope and backslope positions during the spring and in the late summer (Table 1). Consistent with these findings, potential rates of denitrification as measured by DEA were significantly higher in the hollow compared to toeslope and backslopes. Rates of potential denitrification in the hollow were similar in magnitude to rates of nitrification (Table 1).
 Soil total N (as percent) was higher and more enriched in δ15N values in the hollow positions compared to toeslope and backslope positions (Figure 8). Soil N generally decreased with soil depth, whereas δ15N values increased with depth as well as downslope topographic position. For example, hollow surface soil δ15N values were almost 2‰ higher than the ridge surface soil values, 5.3‰ compared to 3.8‰. In addition, δ15N values increased with depth in the hollow, from 5.3 to as high as 7.0‰. Soil C:N ratios tended to decrease with depth and were all less than 12 (Figure 8c).
4.7. Mass Balance
 Estimates of mass balance for HY 2006 showed that N inputs were in close balance to N outputs in stream flow in this Mediterranean catchment. Bulk N deposition was 5.5 ± 0.5 kg N/ha/yr as TDN with 2.6 ± 0.5 kg N/ha/yr as NO3− (average ± standard error), whereas total N outputs in stream flow were 5.3 ± 1.0 kg N/ha/yr as TDN, with 3.4 ± 1.0 kg N/ha/yr as NO3−. Internal cycling of N in the hollow dominated the budget, with soil nitrification producing between 33 and 130 kg N/ha/yr as NO3−. Under anaerobic conditions, rates of potential denitrification in the saturated portion of the hollow, 10–113 kg N/ha/yr, were similar to annual NO3− production via nitrification.
5.1. Shifts in Flow Paths From Deep to Surface Soil-Water and Overland Flow Alter NO3− Chemistry
 Findings from our study supported our hypothesis that shifts in flow paths from deep to surface soil-water and overland flow in the hollow during storm events would alter the delivery of NO3− and more conservative elements to the stream. Whereas rainwater and throughfall input concentrations were relatively low and invariant, matrix and macropore soil-waters differed substantially in nitrogen and base cation concentrations and varied with topographic positions, particularly in the deep toeslope matrix soil-waters (Figure 2). Indeed, two end-members, rainwater and deep toeslope matrix soil-water, explained much of the variation in soil solution and stream water chemistry (89%) suggesting simple mixing of rainwater inputs and deep soil-water (Figure 3a). The consistent linear trend in Ca2+ and Mg2+ indicated that these cations were in ion-exchange equilibrium in the deep soil matrix prior to mixing with rainwater, consistent with findings by Norton et al. . This consistent linear trend (Figure 3) and substantial overlap of surface and subsurface hollow soil-waters with stream water chemistries (Figure 3b) also suggested that most of the water was passing through a small volume of soil, i.e., the hollow, just before entering the stream and attaining its final chemistry (Figure 3b). As Norton et al.  argued, for stream Ca2+ and Mg2+ concentrations to retain a strong linear relationship across a wide range of discharges, representing different flow routing from throughflow and macropore flow to SOF as observed in this study site [Montgomery and Dietrich, 1995; Sanderman et al., 2009], the stream waters needed to be a simple mixing of two constant composition end-members, have relatively uniform soil base saturation properties in the watershed and depth, and/or have all the water pass through the same small volume.
 Saturation overland flow emerged as a smaller but needed third end-member of stream chemistry that could explain deviations from this linear pattern (Figure 3). Specifically, the ratio of Ca2+ to Mg2+ in SOF was lower than deep soil-water Ca2+ to Mg2+ ratios and helped to explain stream concentrations with lower Ca2+:Mg2+ ratios (Figure 3). In this catchment, overland flow was a combination of return flow and direct precipitation onto saturated areas. Given that saturated areas were limited in extent relative to the size of the catchment, return flow was likely the dominant source of overland flow. The chemistry of the collected overland flow suggested that this water had interacted with the soil matrix. Thus, the most probable source of SOF was shallow soil-water—both the inorganic chemistry data presented here and the organic chemistry data presented in Sanderman et al.  suggested this to be the case. Collectively, these data indicated that shifts from deep matrix soil-waters to infrequent surface and saturation overland flow events contributed to observed stream chemistries.
 Dual isotopes of nitrate provided further evidence of temporal changes in runoff mechanisms and differential contribution of shallow versus deep soil solution to the stream during storm events (Figures 6 and 7). In the early HY 2006 storms, prior to the flood on 31 December 2005, surface soil-water δ15N-NO3− and δ18O-NO3− values overlapped with stream water values (Figure 6b) suggesting flushing of stored microbial NO3− from the surface soil into the stream, NO3− that had accumulated in the hollow and other locations during the dry summer season (Table 1). After the December flood event, little vertical mixing appeared to occur within the hollow prior to soil-water entering the stream; rather, lateral flow appeared to dominate during subsequent storm events and mixing of soil waters occurred just prior entering the stream, likely at the seepage face of the channel head. This interpretation was supported by the δ18O-NO3− and δ15N-NO3− values of the surface and subsurface hollow soil-waters plotting linearly, both with a slope of 0.5 but offset by 5‰, and constraining the stream δ18O-NO3− and δ15N-NO3− values (Figure 6b). If there was vertical mixing within the hollow, then there would have been substantial overlap in surface, subsurface and stream values isotopic values as observed in earlier storm events prior to hollow saturation. Temporal changes in NO3− isotope values, therefore, appeared to reflect the sequence in which runoff generation mechanisms linked source areas to streams as described by Montgomery and Dietrich . This sequential activation of runoff mechanisms has been postulated to control stream chemistry in other ecosystems [e.g., Bishop et al., 2004; Sebestyen et al., 2008].
 The finding of a shift in NO3− sources from deep to surface flow paths was also consistent with the findings of Sanderman et al. , who showed that these shifts in soil-water sources explained changes in numerous DOM characteristics. However, whereas Sanderman et al.  showed strong log relationships between the discharge and the DOM attributes (age, concentration, δ13C values) indicating strong hydrologic controls on the DOM chemistry, we found weaker, though significant relationships between discharge and NO3− concentrations (Figure 5) suggesting that NO3− was not just under physical controls. Rather, an interaction of biogeochemical reactions and hydrologic transport of microbial reaction products determined NO3− stream chemistry. For example, in the early winter prior to the onset of seasonal stream flow, high concentrations of soil microbial NO3− that had accumulated in the soil during the spring and summer were displaced and flushed out in macropore solution (Table 1 and Figures 4 and 5). In contrast, during the December 2005 flood, concentrations of NO3− decreased with increasing discharge, apparently due to high contributions of SOF or depletion of NO3− due to high transport and reduced NO3− production (reaction) under saturation conditions (Figure 5).
 Numerous studies have examined flushing of nutrients and the hydrologic mechanisms explaining this phenomenon. In particular, flushing has been considered by van Verseveld et al.  and others to include the following possibilities: (1) a rising water table that intersects high nutrient concentrations in the upper soil layer [Buttle et al., 2001; Creed et al., 1996], (2) vertical transport of nutrients, by preferential or matrix flow through soil to the soil-bedrock interface and then laterally downslope [Creed et al., 1996; Hill et al., 1999], and (3) vertical transport of nutrients and then laterally within the soil profile [Gaskin et al., 1989]. Our early storm event data appeared to support mechanism 2. After the ability of the soil matrix and macropore network to transmit water was exceeded and runoff was dominated by saturated return flow (31 December 2005), our data appeared to support mechanism 3. These shifts in flushing mechanisms over a storm season help to explain variable responses in runoff chemistry [Bishop et al., 2004; Kirchner, 2003].
5.2. Source of NO3−-Progressive Denitrification and Mixing of Microbial NO3− Sources
 Findings from our study also supported the hypothesis that shifts in flow paths from deep to surface soil-water and overland flow in the hollow during storm events would alter the sources of NO3− to the stream. Numerous studies have used dual isotopes of NO3− to evaluate sources of NO3− in seminatural catchments (see Curtis et al.  for review). Of these studies, most of them have been conducted in humid forested or agricultural catchments of the northeastern US and Canada, with a few studies in southern California, Colorado and western US National Parks [Campbell et al., 2002; Michalski et al., 2004; Nanus et al., 2008]. A majority of the studies have shown that stream water NO3− is largely derived from microbial NO3− despite significant atmospheric NO3− or other anthropogenic N inputs [e.g., Pardo et al., 2004; Barnes et al., 2008; Kendall et al., 2007] and explained by microbes efficiently processing anthropogenic sources and resetting and/or overwhelming the atmospheric signal in stream flow, unless high frequency sampling during peak storm flow is used to capture the atmospheric NO3− signal [Sebestyen et al., 2008] or a combination of δ18O-NO3− and Δ17O-NO3− techniques are utilized [Michalski et al., 2004; Kendall et al., 2007]. Consistent with these findings, stream water δ18O-NO3− values in our catchment ranged from −2 to +8‰, consistent with NO3− derived from microbial processing (Figure 6a).
 We ruled out that atmospheric NO3− deposition was likely a significant source of NO3− in our catchment for the following reasons. One, we found that atmospheric deposition of NO3− was low on an annual basis, 2.6 ± 0.5 kg N/ha/yr, similar to Ewing et al. , and this input was low compared to high annual rates of net nitrification, 33–130 kg N/ha/yr suggesting that internal cycling of N likely swamped out any atmospheric signal that entered the ecosystem. Two, δ18O-NO3− values were low when discharge was high, particularly during the 31 December flood event (Figure 7) when an atmospheric NO3− signal in rainfall and/or SOF would likely have contributed the most to stream flow suggesting that atmospheric NO3− was not contributing to higher δ18O-NO3− values. Three, if high atmospheric δ18O-NO3− values were contributing to an increase in δ18O-NO3− in the soil-waters and stream δ18O-NO3− values, then we would have expected the surface hollow soil-waters δ18O-NO3− values to be higher than the subsurface δ18O-NO3−. Instead, the δ18O-NO3 values in the subsurface soil-waters were higher than the surface soil-water by ∼5‰. In the next sections, we posit that the patterns of stream δ15N-NO3− and δ18O-NO3− values were the products of fractionation associated with progressive denitrification of soil-waters and mixing of these two residual microbial NO3− sources with different δ18O values of water used in formation of microbial NO3.
5.2.1. Progressive Denitrification
 Results from dual isotope of NO3− analyses showed that stream water δ15N-NO3− and δ18O-NO3− values plotted linearly (slope of 0.9), indicating possible isotopic enrichment of the isotopes at a 1:1 ratio associated with denitrification (Figure 6). However, diagnostic tests of stream water δ15N-NO3− were inconclusive; stream water δ15N-NO3− values increased with the inverse of NO3− concentrations (Figure 7b), whereas they decreased with the natural logarithm of NO3− concentrations (Figure 7c), possibly indicating a combination of mixing and fractionation associated with denitrification.
 Further analysis revealed that the dual isotopes of NO3− for surface and subsurface soil-waters from the hollow both plotted linearly with slopes of 0.5 ± 0.1, though offset by ∼5‰, and constrained the stream NO3− isotope values (Figure 6b). These slopes agreed well with isotopic enrichment of δ15N-NO3− and δ18O-NO3− in a 2:1 ratio associated with denitrification. Indeed, diagnostic plots of these soil-waters showed a significant exponential increase in δ15N-NO3− with decreasing NO3− concentration (r2 > 0.91, p < 0.0001) (Figure 7g) providing supporting evidence of fractionation associated with denitrification [Mariotti et al., 1988; Kendall, 1998; Kendall et al., 2007].
 The isotopic fractionation plots revealed two straight lines, with different enrichment factors in the surface and subsurface soil-waters, −0.9‰ and −1.1‰, respectively, as well as different initial δ15N value for NO3−, 2.51‰ for the surface compared to 4.52‰ for the subsurface soil-waters (Figure 7g). These apparent enrichments factors (ε) for denitrification derived from the isotope fractionation plots were quite low (−0.9 to −1.1‰) compared to other studies that ranged from −40 to −1.5‰ [Mariotti et al., 1988; Kendall, 1998; Kendall et al., 2007; Sebilo et al., 2003; Lehmann et al., 2003]. These low ε values may indicate rapid denitrification rates as suggested by Mariotti et al. . The full range for δ15N values of NO3− in the soil-waters (8‰) and the apparent enrichment factors were lower than those observed in Burns et al. , who suggested relatively high fractionation associated with “riparian” denitrification (in the sense of Sebilo et al. ). Our low enrichment factors may indicate that denitrification was more indicative of “benthic” denitrification as described by Sebilo et al. , where the supply of microbial reactants of NO3− may have limited the rates of denitrification as runoff mechanisms and environmental conditions evolved over the hydrologic year. Indeed, temporally and spatially variable soil-water and soil NO3− concentrations and N process rates presented in this study and high DOM in the hollow as reported by Sanderman et al.  (Figure 2 and Table 1) may have led to spatially and temporally variable limitations and excesses in the supply of NO3− that likely controlled rates of denitrification. Further research is merited to follow up on these patterns at higher spatial and temporal resolution given our relatively limited number of soil-water samples analyzed for isotopes from the different horizons. However, other supporting evidence, denitrification enzyme assays, in particular, indicates extremely high potential for denitrification exists in the hollow soils, 0.91 kg N/ha/d, when environmental conditions (high NO3−, DOM, and low O2) for denitrification were met (Table 1). Additionally, increases in soil δ15N values in the hollow surface horizons compared to backslope soils (Figure 8) appeared to reflect the increasing proportion of inorganic N lost via denitrification with changes in topographic position [Amundson et al., 2003; Brenner et al., 2001]. The 2‰ difference between the initial δ15N value for NO3− at the different soil depths (Figure 7h) may in part be explained by the ∼2‰ difference between the soil δ15N values at the different depths, 5.3‰ in the surface horizon compared to 7.0‰ at 50 cm depth. Combined, these data together with mass balance indicating approximate balance of N inputs and outputs suggested that the process of denitrification was consuming most of the microbial nitrate produced when high concentrations of reactants (nitrate, DOC) and saturated conditions converged.
 The use of dual isotopes to identify denitrification has been challenging because of possible mixing of two sources showing the same trend line as denitrification [Xue et al., 2009]. Only a few studies have identified kinetic fractionation indicative of denitrification in soil-water and streams [Ohte et al., 2004; Panno et al., 2006; McMahon et al., 2006; Burns et al., 2009]. For example, Burns et al.  observed linear seasonal patterns of δ18O-NO3−:δ15N-NO3− but suggested that mixing of two or more distinct sources may have caused to these patterns. In combination with other tracers, Osaka et al.  found little evidence of denitrification in soil-waters, whereas they showed evidence of denitrification in the shallow groundwater in their catchment, similar to other shallow groundwater studies [Bottcher et al., 1990; Heffernan et al., 2012; McMahon and Bohlke, 2006; Mengis et al., 1999].
 Several possible factors may help to explain the strong signal of denitrification in our coastal Mediterranean catchment compared to others. One, atmospheric NO3− and other sources of NO3− as manure or septic have been significant and most likely confounded the isotopic signal of microbial NO3− values in other studies as suggested by Michalski et al.  and Nanus et al. . These sources were not present or at least not significant in our study catchment. Two, our catchment underwent pronounced changes in seasonal climate from a relatively hot, dry summer to a cool, wet catchment with variable saturation in space and time. These pronounced wetting and drying cycles resulted in “hot spots and hot moments” (in the sense of McClain et al. ) in N cycling within the catchment (Table 1) leading to high potential for periods of high production of NO3− via nitrification and periods of high consumption of NO3− via denitrification. Studies have shown that perched water tables and fluctuating or rising water tables result in highly redox active sites for coupled nitrification-denitrification because oxidized and reduced species are in close proximity to each other [Hedin et al., 1998; Pett-Ridge et al., 2006]. Strong hydraulic conductivity contrasts existed at both the soil/saprolite contacts in this study resulting in a seasonal shallow ground water table above the bedrock and a transient perched water table above the saprolite in response to large winter storms [Montgomery and Dietrich, 1995]. The hollow in our headwater catchment was consistent with riparian zone or catchment-stream interfaces, which have been considered a critical zone controlling rates of N loss to streams [Cirmo and McDonnell, 1997; Hedin et al., 1998; Ocampo et al., 2006a, 2006b; Triska et al., 1989, 1993].
5.2.2. Mixing of Two Residual Microbial NO3− Sources
 Findings from our study also suggested that stream water δ18O-NO3− values represented a mixture of two residual microbial NO3− sources with different δ18O values of water used in formation of microbial NO3− (Figure 7f). These differences could explain the higher δ18O values for NO3− in subsurface soil-waters compared to the surface waters (Figure 6b). Specifically, comparison of observed and predicted δ18O-NO3− based on the theoretical microbial nitrification model (equation (2)) and possible sources of δ18O-H2O suggested NO3− was derived from in situ nitrification (Figure 7f) and that the ∼5‰ difference in the δ18O-NO3 observed between the shallow surface solution and deep soil solution (Figure 6b) could be explained by rain water being the dominant source water for in situ nitrification in the surface soil-waters (after the December flood event), whereas ambient soil-water appeared to contribute to δ18O-NO3− subsurface soil-water values in combination with rainwater (Figures 7e and 7f). For example, the predicted δ18O-NO3− soil-water values were −2 to +4‰ given the observed rainwater δ18O-H2O values of −5.8 to −14.8‰ (Figure 7f) and explained the range of observed shallow soil-water δ18O-NO3− values, with the exception of one sample prior to the flood event that looked like soil-water δ18O-H2O values (Figures 6b and 7f). The predicted microbial δ18O-NO3− values from soil-water δ18O-H2O values (−4 to −5.5 ‰) were quite narrow and ranged from +4 to +5.5‰. These values overlapped with the mid and late subsurface soil-water δ18O-NO3− values (Figures 6b and 7f). Stream water δ18O-NO3−, therefore, represented a mixture of these two temporally variable source waters and progressive denitrification as described above.
 An alternative explanation for higher δ18O values in nitrate in the subsurface soil layers that we cannot exclude at this time is that the δ18O values reflect different sources of O2 and thus different δ18O values in NO3−. Sanderman and Amundson  showed aeration limitation of soil respiration in the hollow when the volumetric water content exceeded 0.4. These authors found that when soils were unsaturated, soil respiration rates were quite high, with resulting soil CO2 concentrations as high as 20,000 μmol CO2 mol−1 air in the hollow. Under high CO2 production and low diffusivity conditions, soil O2 may have become enriched in δ18O. The fact that only δ18O isotope values appear to be higher in the deeper soil layers are consistent with a possible 18O enrichment of the residual O2. However, δ18O-O2 values were not measured in this study and beyond the scope of this study. One of the few studies that analyzed δ18O-O2 along with δ18O-NO3 by Spoelstra et al.  did not observe a significant difference in δ18O-O2 values of atmospheric O2 compared to δ18O-O2 values at 55 cm depth. Future studies need to assess the role of respiration in affecting the stable isotope ratios of NO3− produced from nitrification to explain higher than expected values of δ18O-NO3 [Snider et al., 2010].
 We linked soil N cycling to stream chemistry using a combination of hydrometric and dual NO3− isotopes analyses. We found that shifts in runoff mechanisms and putative flow paths altered sources of NO3− from deep subsoil sources to surface soils as well as the processes dominating delivery of NO3− to streams. End-member mixing models, overlap of surface and subsurface soil-waters in the hollow with the stream chemistry, and dual isotopes of NO3− suggested that that mixing of these surface and subsurface hollow solutions was occurring just prior to entering the stream.
 Differences in N process rates in space and time in this very small watershed helped to explain the patterns of NO3− in soil-waters and stream water. Strong topographic controls on N cycling resulted in higher and sustained rates of nitrification in the hollow of the catchment compared to backslopes and toeslopes in this seasonally water limited environment. Findings from our study suggests that the convex hillslope positions have temporally variable rates of N cycling, or hot moments as described by McClain et al. , and turn on and off with soil moisture controls. In contrast, the hollow appears to be a sustained “hotspot” of biogeochemical activity. Sustained rates of N cycling in spring and late summer in the hollow due to available soil moisture likely explained the large pulse of nitrate observed in soil-waters and in stream waters with the onset of stream flow. Results from our study also suggested that the hollow had higher rates of denitrification when higher carbon and nitrate and water flow converged. Indeed, mass balance estimates suggested that most of the microbial nitrate produced from nitrification was consumed by denitrification when environmental conditions for denitrification were satisfied.
 To our knowledge, this is the first study to demonstrate mixing of two microbially derived nitrate sources for stream water, and progressive denitrification within the soil. Our study suggests that the small volume of the hollow is a filter that constrains the final N stream chemistry and that shifts in runoff mechanisms and flow paths change the source and mixing of nitrate from deep to surface soils. The data reveal that the N chemistry of stream waters, particularly in highly seasonal climates, is a complex integral of unique geochemical environments that ultimately shape watershed functions [McDonnell et al., 2007].
 This work was funded with a grant to R. Amundson by the Kearney Foundation of Soil Science and funding from University of Arizona Lohse Start up funds. Kathleen Lohse, now at Idaho State University, was supported by the National Science Foundation under award EPS-0814387 to complete analysis and writing. We also thank three reviewers and the comments of the editor that vastly improved the manuscript.