Hydrological process controls on nitrogen export during storm events in an agricultural watershed


  • Department of Biological Environment, Akita Prefectural University, Akita 010-0195, Japan.

R. JIANG, Laboratory of Soil Science, Graduate School of Agriculture, Hokkaido University, Kita-Ku Kita-9 Nishi-9, Sapporo 060-8589, Japan. Email: jiangrui@chem.agr.hokudai.ac.jp


The dynamic characteristics of nitrogen (N) and suspended solids (SS) were investigated in stream water during four storm events in 2003 in the Shibetsu watershed, eastern Hokkaido, Japan. Analysis showed that total nitrogen (TN), nitrate-N (inline image-N), dissolved organic nitrogen (DON), particulate nitrogen (PN) and SS concentrations all peaked sharply during the rising limb of the discharge hydrograph, but peaks in PN and SS were more significant than the peak in dissolved N. Particulate N and SS consistently displayed clockwise hysteresis with higher concentrations during rising flows, whereas inline image-N and DON showed different patterns among the storms depending on the antecedent soil moisture. An M (V) curve, defined as the nutrient mass distribution versus the volume of discharge, showed that a “first flush” of PN, inline image-N, DON and SS was observed; however, the distribution of nutrient loads in the discharge was different. Particulate N and SS had a shorter flushing characteristic time constant (t1/e, defined as the time interval required for a decline in nutrient concentrations in discharge water to e−1 [37%] of their initial concentrations), but contributed 80% of fluxes during the first 50% of the discharge, whereas the longer flush time (t1/e) of inline image-N and DON with slowly decreased concentrations led to half loads during the recession of the discharge. These data indicate that flush mechanisms might be distinguished between particulate nutrients and dissolved N. Analysis showed that the concentrations of PN and SS derived from soil erosion were related to surface run-off. In contrast, inline image-N originated from the near-surface soil layer associated with the rising shallow groundwater table and mainly flushed with subsurface run-off. The different flushing mechanisms implied that different watershed best management practices should be undertaken for effectively mitigating water quality degradation.


Quantifying the export of nitrogen (N) from catchments has become a significant issue for land managers over the past 20 years. Although many studies have demonstrated that the characteristics of nutrient export are complicated because of a variety of factors, such as geographical, hydrological, climatic, biochemical and anthropological factors, and although each of these factors is not a fundamental process, they share some similarities. First, N export had a significant variable signal at a temporal scale. In particular, rainfall events and/or snowmelt seasons contribute to high concentrations and fluxes of N (Baron and Campbell 1997; Brooks and Williams 1999; Hatano et al. 2005; Inamdar et al. 2004; McNamara et al. 2008; Mitchell et al. 1996; Zhang et al. 2007). Second, flushing of N during a snowmelt season or during storm events has been observed (Brown et al. 1999; Burns 2005; Burns et al. 1998; Creed and Band 1998a,b; Creed et al. 1996; Inamdar et al. 2004; McHale et al. 2002; Zhang et al. 2007). Last, flow paths have a close relationship with N export (Jia et al. 2007; Jiang et al. 2008; Zhang et al. 2007).

There has been considerable interest in identifying the sources, flow paths and transport mechanisms responsible for the export of N, particularly inline image-N, within watersheds at seasonal and rainfall event scales. These studies have improved our understanding of some of the processes, but many are not yet clearly understood. The mechanisms of inline image-N export are usually explained by contradictory processes. Creed and Band (1998a,b) found a clockwise, discrete hysteresis pattern of inline image-N concentration from glaciated catchments in the Canadian Shield and attributed this pattern to the flushing of inline image-N from near-surface soil layers as a result of the rising groundwater table, but they did not find any proof of the source of inline image-N. In contrast, Brown et al. (1999), Inamdar et al. (2004) and McHale et al. (2002) believed that inline image-N export occurred via deep flow paths and that the rise in inline image-N concentrations was associated with the rapid displacement of till water by infiltrating precipitation. Hill et al. (1999) indicated that the biogeochemistry of the organic horizon could regulate patterns of inline image-N loss in subsurface run-off movement by preferential flow pathways in forest soils. Ocampo et al. (2006) explained that the dynamic of the shallow ephemeral perched aquifer drove a shift from hydrological controls on inline image-N discharge during the “early flushing” stage to an apparent biogeochemical control on inline image-N discharge during the “steady decline” stage of the flushing response.

Less attention has been paid to dissolved organic N (DON) compared with inline image-N. Recent studies suggest that contributions of DON can be significant and can constitute a major portion of the total N solute export (Campbell et al. 2000; Willett et al. 2004). Although stream nitrate and DON may originate from similar shallow subsurface and surface flow paths during storm events, differences in the flushing response among inline image-N and DON have been found (Cooper et al. 2007; Sebestyen et al. 2008).

Despite numerous investigations of N dynamics, studies on the relationship between the distribution of N fluxes and stream discharge in agricultural or forested watersheds during storm events are scarce. Quantification of nutrient fluxes in sewage systems and urban watersheds indicate that these watersheds present strong “first flush” for most storms and constituents (Barco et al. 2008; Bertrand-Krajewski et al. 1998; Lee et al. 2002). In general, the term “first flush” has been used to indicate that the mass emission rate is higher during the initial portions of run-off than during the last portion (Kondolf and Wilcock 1996; Lee et al. 2002). This characteristic of “first flush” has been used to define different pathways for particulate nutrients and dissolved nutrients (Jiang et al. 2009) and has been used for watershed best management practices (BMPs), such as enhancement of sediment and nutrient removal efficiency by treating the first stage of run-off using sedimentation devices or filters (i.e. ditches, tanks and ponds) (Barco et al. 2008).

Several studies have examined N export in Shibetsu watershed (Hayakawa et al. 2006, 2009; Woli et al. 2004). Hayakawa et al. (2009) reported that net N input (NNI) amounted to 55 kg N ha−1 year−1 in the Shibetsu watershed and that N export from the watershed outlet accounted for 27% of the NNI. A number of studies have indicated that land use has a significant positive correlation with inline image-N export and that agricultural N was a dominant source (Hayakawa et al. 2006; Woli et al. 2004). However, these studies do not focus on the mechanisms controlling N export to streams at a high temporal resolution during storm events. Therefore, the present study focuses on the dynamic concentrations and flux distribution of N in discharge during storms. The objectives of the present study were to: (1) assess temporal patterns in N concentrations during storm events, (2) evaluate the distribution of N fluxes in discharge within a storm, (3) clarify whether flushing of N export exists among inline image-N, DON and PN and any variation, (4) examine potential sources of N export from the Shibetsu watershed.

Materials and methods

Description of the watershed

The Shibetsu watershed is located in eastern Hokkaido (outlet as shown in Fig. 1; 43.634′N, 145.085′E), Japan. The watershed area is 679 km2 and the upper parts dominated by forest are covered by Volcanogenous Regosols; the downstream areas are mainly used for agricultural purposes and are covered by Cumulic Andosols, Gray Lowland Soils and Peat Soils. The region has a mean slope of 4.28°, with a maximum value of 34°. The slopes are more gentle and concave in the downstream areas compared with the upstream areas. This region has a hemi-boreal climate, characterized by warm summers and cold winters. Precipitation averages 1147 mm year−1 and the annual mean temperature is 5°C, with the lowest mean monthly temperature in February (−8.3°C) and the highest mean monthly temperature in August (18.0°C) (1978–2002 average; Japan Meteorological Agency 2007, http://www.jma.go.jp). The watershed consists of agriculture (51.4%), forest (45.6%), urban area (1.4%) and waste land and road (1.6%). The Nakashibetsu town, covering most of the Shibetsu watershed, has a large grassland area, and dairy farming is the main occupation. The dominant forest vegetation is Japanese larch, Larix kaempferi. Grassland (covered by Phleum pratense) occupies more than 95% of the agricultural land area and the remaining area is cultivated with maize (Zea mays L.), sugar beet (Saccharum officinarum), potato (Solanum tuberosum) and Japanese radishes (Raphanus sativus Linn.). The human population was estimated to be 24,000 persons (35 people km−2), concentrated in the middle part of the watershed.

Figure 1.

 Location of the study watershed and water sampling site.

Watershed monitoring, sampling and analysis

Base flow water samples were grabbed from the stream outlet in 1 L polypropylene bottles once per month from March to November 2003. An automated water sampler (ISCO 3700, Isco, Lincoln, NE, USA) was installed at the outlet of the Shibetsu watershed and water samples were collected during four storm events: 20 June (E1), 11 July (E2), 30 September (E3) and 23 October (E4) 2003. The autosampler was triggered when the rainfall was >4 mm per 30 min, with intervals of 15 min to 1 h during the rising stage of discharge, and 2–6 h during the receding section. Water samples were transported to the laboratory quickly after a storm and then stored at 4°C until analysis. Rainfall was measured by a tipping-bucket rain gauge (0.2 mm) placed in an open area near the automated sampler. A daily stream water stage, measured every 10 min, was obtained from the website of the Ministry of Land, Infrastructure and Transport, Japan (http://www1.river.go.jp). Discharge was calculated using calibrated formulas based on monitoring by local government. Groundwater wells in near-stream forests were constructed of 5 cm (internal diameter) polyvinylchloride pipes and groundwater levels were recorded using pressure transducer and capacitance water level probes at 10 min intervals. A shallow groundwater well was cored to a depth of 2.94 m under the ground surface where a coarse sandy sediment layer was intersected. A deep groundwater well, with a depth of 12.52 m, was cored to the aquifer of the Mashu pumice layer. Soil samples were collected once in August 2008 from 27 sites representing most of the watershed area. Each sample was taken in triplicate with an auger from depths of 0–20, 20–50 and 50–100 cm. The three samples from each depth were homogenized before analysis.

Suspended solids (SS) were measured from 800 mL subsamples that were filtered through pre-weighed glass microfiber filters (47 mm; Whatman GF/F, Whatman International Ltd., Maidstone, England). The filters were dried at 90°C for 24 h and weighed again. Total N (TN) was determined using alkaline persulfate digestion and HCl-acidified ultraviolet (UV) detection. After filtering through 0.2 μm membrane filters, water samples were used for analyzing total dissolved N (TDN), inline image-N, inline image-N, inline image-N and Si. Concentrations of TN and TDN were determined using alkaline persulfate digestion and HCl-acidified UV detection. inline image-N and inline image-N were determined by ion chromatography (QIC Analyzer; Dionex, Sunnyvale, CA, USA); inline image-N was determined by colorimetry using the indophenol blue method; and Si was determined colorimetrically by the molybdenum blue method. Particulate N (PN) and DON were calculated by subtracting the concentration of TDN from TN, and inorganic N (inline image-N, inline image-N and inline image-N) from TDN, respectively. Soil samples were air-dried and ground to pass through a 2 mm sieve for inline image-N analysis. inline image-N in soil was extracted using a 1:5 soil : 2 mol L–1 KCl solution and the concentrations were determined by colorimetry.

Data analysis

Kruskal–Wallis tests and multiple comparisons using Steel–Dwass tests were used to examine the variation in water chemistry.

Antecedent precipitation index

Several researchers have used relative antecedent precipitation index (API) values to compare soil moisture among pre-storm conditions (Christopher et al. 2008; Inamdar and Mitchell 2006; McDonnell et al. 1991). The API index was calculated for the four storm events in our study, and is defined as:


where x = 7 and 21 days before a rain event and Pi (mm) is the total precipitation on the ith day before the event. We used API7 for calculating the surface soil moisture and API21 for the deep groundwater situation because the surface Volcanogenous soil usually has high infiltration rates and water can move quickly into the shallow groundwater aquifer. However, the water needs more time to seep into the deeper soil layer through the coarse sandy sediment layer.

Run-off coefficient

The run-off coefficient is defined as the total volume of discharge divided by the total volume of precipitation (amount of precipitation multiplied by watershed area), which is another indicator of wetness in the watershed.

Time constant

A time series of the discharge and the concentration of N in the discharge highlight the export behavior of N from a watershed. We assumed that there was an exponential decline in N concentrations during the receding limb of the discharge. According to Creed and Band (1998a), this decline can be described by:


where Nt is the concentration of N in the discharge waters at time = t (mg L−1), N0 is the concentration of N in the discharge waters at t = 0, the time that the peak concentration of N is observed (mg L−1), t is time in h and k is the constant proportionality factor (h−1). The time constant (t1/e) is defined as the time interval that must elapse in order for the concentrations of N in the discharge waters to decline to e−1 (37%) of their initial concentration.

M (V) curve and mass first flush ratio

An M (V) curve, which is described as the normalized load (Mi) versus the normalized water discharge volume (Vi), was used to analyze the load distribution in discharge (Bertrand-Krajewski et al. 1998):


where N is the total number of measurements, j is the index from 1 to N, and Ci and Qi are the instantaneous concentration and water discharge. If the M (V) curve is above the bisector, the “first flush” phenomenon occurs (Geiger 1987; Saget et al. 1995) and the magnitude of the first flush can be quantified for each storm and for each water-quality parameter using the mass first flush (MFF) ratio (Barco et al. 2008). The MFF ratio was calculated as:


where n is the index in the storm corresponding to the percentage of the water discharge, ranging from 0 to 100%. By definition, the MFF ratio equals zero at the beginning of a storm and always equals 1.0 at the end of the storm. The MFF ratio is a useful tool for quantifying first flush and can be statistically characterized or used in regressions or other investigations to understand the magnitude of a first flush and storm or catchment characteristics. For example, an MFF20 equal to 2.5 means that 50% of the mass is contained in the first 20% of the discharge water.

In addition, the concentrations of inline image-N and inline image-N in the bulk rainfall samples during different storm events were very low (i.e. inline image-N was <0.15 mg L−1 and inline image-N was <0.2 mg L−1). Therefore, atmospheric input of N with rainfall was not considered to be an important source of stream water N during the hydrological events.


Hydrological storm events

The hydrological characteristics of the four storm events varied considerably (Table 1). The total rainfall ranged from 50 to 91 mm. Although a single peak was observed for all events (Fig. 2), the peak discharges varied from 51.4 to 167.7 m3 s−1 (Table 1). The fast hydrological response of the watershed to rainfall was reflected in the length of time to the peak (Tp, time from the start of the rain to the peak of stream water discharge). The longest Tp was found in E2, which had the lowest antecedent precipitation index (API7) value (Table 1). In contrast, E3, with the highest antecedent soil moisture (19.3 for API7 and 28.4 for API21), had the shortest Tp (Table 1). The run-off coefficient increased with an increase in the value of API21, except for E4; but it also increased with an increase in groundwater level, except for E3, during the four storms from early summer to autumn (Table 1; Fig. 2), suggesting that antecedent soil moisture together with seasonal factors could have controlled watershed wetness.

Table 1.   Characteristics of each hydrological storm event in the study area
Storm eventDatePrecipitationRainfall duration (h)APIXDischargePercentage of annual dischargeRun-off coefficient
Total (mm)Max (mm h−1)API7API21Total (×106 m3)Peak (m3 s−1)Tp (h)
  1. APIX, antecedent precipitation index determined for 7 and 14 preceding days; Tp, time to peak (the time from the outset of the rain to the peak of stream water discharge). El, E2, E3, and E4 represent the storm events on 20 June, 11 July, 30 September and 23 October 2003, respectively.

Figure 2.

 Precipitation, stream discharge and groundwater level (only one well is shown, the others are similar) during four storm events (E1–E4) in 2003.

The groundwater level in the near-stream forests was analyzed because it might be critical for describing the evolution of solute signatures during the storm events. The focus in the present study was on the temporal response of groundwater level in near-stream forests versus stream discharge. The shallow and deep groundwater levels along with discharge and precipitation are presented in Fig. 2. The groundwater level during the storm events indicated surface saturation and deep seepage in the near-stream forests. Importantly, these water levels showed that: (1) the maximum shallow groundwater level occurred at the precipitation peak and just before the stream discharge peak, implying that the shallow groundwater quickly responded to the rainfall and contributed to the stream discharge, (2) the deep groundwater rose slowly at the rising limb of discharge, but kept increasing after recession of the discharge, (3) surface saturation was maintained even after precipitation ceased and during recession of the discharge.

Comparison of nitrogen concentrations between storm events and base flow

The nutrients and SS concentrations in the base flow differed among constituents (Table 2). The most prevalent form of N was inline image-N (54% of TN, 59% of DTN and 95% of DIN). Dissolved organic nitrogen accounted for 36% of the TN and PN accounted for only 9% of the TN. The concentrations during storm events were highly variable among N species. The dominant form during the storm events was not only inline image-N, but also PN. inline image-N accounted for 8–70% of the TN with an average of 37, 27, 50 and 45% for E1, E2, E3 and E4, respectively; whereas PN accounted for 0–86% of the TN depending primarily on the peak discharge, with averages of 46, 57, 20 and 21% for each storm event. The PN: inline image-N ratio (Fig. 3) showed that PN was the dominant form on the rising limb of the discharge, and an increase in rainfall increased the length of the dominant time (i.e. E2); whereas inline image-N was dominant on the receding limb. Dissolved organic nitrogen ranked second, ranging from 15 to 34% during the storm period. inline image-N concentrations were very low as base flow and were below the detection limit in many samples. According to Kruskal–Wallis H and Steel–Dwass tests (Table 2), discharge, SS, PN and inline image-N all showed significant differences among storm events and base flow. However, inline image-N and DON differed only among storm events; whereas TN did not show any difference. In brief, the concentrations of all forms of N (except for inline image-N and inline image-N) were significantly higher than those observed in forested watersheds, regardless of the base flow or storm events.

Table 2.   Discharge, suspended solids and nitrogen concentrations during the storm and baseflow events
Storm eventsDi scharge (m3 s−1)SS (mg L−1)inline image -N (mg L−1)TN (mg L−1)DON (mg L−1)PN (mg L−1)NH4+-N (mg L−1)inline imageN/ TN(%)DON/ TN(%)PN/ TN(%)
  1. Different superscript letters (a– c) following the median values indicate significant differences (Steel–Dwass test, P < 0.05). DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids; TN, total nitrogen.

Figure 3.

 The ratio of particulate nitrogen (PN) : inline image-N during four storm events (E1–E4).

Dynamic concentrations of nitrogen during storm events

The concentrations of TN, PN and SS changed notably and had similar patterns in all storm events (i.e. they peaked sharply during the rising limb of discharge and then dropped rapidly) (Fig. 4). The TN and PN concentrations are shown in Table 2 and Fig. 4. A significant positive correlation was found between TN and PN (R2 = 0.98, P < 0.001). In addition, PN and SS were significantly positively correlated (R2 = 0.89, P < 0.001). The dissolved N concentrations did not show significant changes, although peak concentrations also occurred before the discharge peaked. inline image-N concentrations varied slightly with ranges of 0.5–0.8, 0.3–0.7, 0.5–1.0 and 0.4–0.8 mg L−1, whereas the ranges in DON concentrations were 0.1–0.6, 0.1–0.5, 0.2–0.8 and 0.3–0.7 mg L−1 during the E1, E2, E3 and E4 storms, respectively. The maximum value of PN in E2 was much larger than the values recorded in the other storm events, probably because of the antecedent dry soil conditions (low API7 value; Table 1).

Figure 4.

 Temporal changes in nitrogen (N) and suspended solid (SS) concentrations, discharge, rainfall and shallow groundwater level during four storm events (E1–E4).

The peaks in dissolved N followed the peaks in the shallow groundwater level or after the water table was elevated to a certain level (Fig. 4). This corresponding response was most obvious in E3, when concentrations of inline image-N and DON both showed two peaks corresponding to the shallow groundwater level hydrograph; whereas PN did not show a second peak similar to dissolved N. Therefore, we speculated that dissolved N was sensitive to changes in the shallow groundwater.

Discharge-concentration patterns

Although N and SS both peaked before the discharge peaks, the relationships between the concentrations and discharge were different. Particulate N and SS displayed consistently clockwise hysteresis in all storm events, whereas dissolved N showed no consistent pattern among the storms. Counter-clockwise concentration-discharge relationships for inline image-N and DON were found in E1, no hysteresis in E3, and clockwise hysteresis existed in E2 and E4 (Fig. 5).

Figure 5.

 Patterns of nitrogen (N) and suspended solids (SS) during four storm events (E1–E4). Arrows indicate the time course.

Discharge-weighted mean concentrations of the nutrients and rainfall conditions during each rising section have been analyzed to investigate the run-off mechanisms. Table 3 shows that nutrient concentrations were significantly positively correlated with cumulative rainfall, whereas there was no uniform correlation with rainfall intensity in the storm events.

Table 3.   Pearson correlation coefficients (r) between discharge-weighted mean concentrations of suspended solids and nitrogen and rainfall condition during the rising limb of discharge in each storm event
Storm eventr value for the N concentration and rainfall intensityr value for the N concentration and accumulative rainfall
SSinline imageNDONPNSSinline image-NDONPN
  1. *P < 0.05; **P < 0.01. DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids.


As the N and SS concentrations all peaked quickly during the rising section of discharge and declined during the receding section, we investigated the dynamic characteristics of the concentrations during the receding limb of the discharge hydrograph to determine the flow path contributions in the catchment and the potential N export affected by a storm event. The t1/e values of PN and SS were shorter than the dissolved N values, and t1/e for inline image-N was the longest (Table 4). The reason for these differences could be that PN and SS are likely to be flushed away by overland flow and dissolved nutrients could have been removed by both surface and subsurface flow (Zhang et al. 2007). Moreover, we noticed that the t1/e values for all nutrients were longer in E1 than in the other storms and the longer time of export might have led to increased nitrate loads.

Table 4.   Summary statistics for regressions describing the exponential decline (k) in nitrogen concentrations during the storm events
Storm eventk (h−1)r2Time constant t1/e (h)
  1. DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids.

 El−0.039 ± 0.0070.72 ± 0.4825.87 ± 3.93
 E2−0.071 ± 0.0070.89 ± 0.4814.01 ± 1.26
 E3−0.203 ± 0.0220.84 ± 0.494.92 ± 0.49
 E4−0.463 ± 0.1220.67 ± 0.942.16 ± 0.45
 Average−0.194 ± 0.0390.78 ± 0.5911.74 ± 1.53
inline imageN
 El−0.010 ± 0.0010.90 ± 0.0697.35 ± 9.00
 E2−0.038 ± 0.0050.81 ± 0.0926.10 ± 3.13
 E3−0.036 ± 0.0100.53 ± 0.1427.48 ± 6.11
 E4−0.063 ± 0.0160.66 ± 0.1515.86 ± 3.20
 Average−0.037 ± 0.0080.72 ± 0.1141.70 ± 5.36
 El−0.019 ± 0.0020.84 ± 0.1552.64 ± 6.08
 E2−0.083 ± 0.0120.78 ± 0.2012.00 ± 1.52
 E3−0.081 ± 0.0120.82 ± 0.1612.39 ± 1.55
 E4−0.156 ± 0.0190.93 ± 0.106.43 ± 0.72
 Average−0.085 ± 0.0110.84 ± 0.1520.87 ± 2.47
 El−0.035 ± 0.0050.79 ± 0.3328.29 ± 3.67
 E2−0.245 ± 0.0180.94 ± 0.294.09 ± 0.27
 Average−0.139 ± 0.0110.86 ± 0.3116.19 ± 1.97

M (V) curves of the nitrogen load distribution versus discharge during storm events

An M (V) curve helps to understand the nutrient mass distribution versus the water discharge volume relationship. Figure 6 shows the normalized loads of N and SS as a function of normalized discharge for the four storm events. All curves were above the bisector, which showed the “first flush” of constituents. The PN and SS curves were above the curves of inline image-N and DON, indicating a stronger first flush for particulates, probably because PN and SS, which are associated with soil erosion, are more prone to movement during the early stage of a storm.

Figure 6.

 M (V) curve: normalized mass first flush relative to normalized discharge. DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids; TN, total nitrogen.

First flush ratios (MFF30 and MFF50) were used to quantify the magnitude of the first flush and to analyze load distribution. Table 5 shows that the MFFn of PN and SS were higher than those of inline image-N and DON. On average, inline image-N, DON, TN, PN and SS transported 35, 39, 49, 66 and 71% of the total loads in the first 30% of the water discharge during the four storms, respectively; and 56, 63, 70, 80 and 88% of loads for the first 50% of the water discharge. These data indicated that the first flush of dissolved N was weaker than PN and SS and the contributions of the first flush to the export loads were different between particulate nutrients and dissolved N, which might be ascribed to different flush mechanisms for the different forms of N in a watershed.

Table 5.   Mass first flush (MFF30 and MFF50) values for quantifying the magnitude of the first flush
Storm eventMFF30MFF50
inline imageNDONTNPNSSinline imageNDONTNPNSS
  1. DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids; TN, total nitrogen.


Contribution of storm events to annual nitrogen loads

We estimated the loads of N and SS during the storm events and found that TN accounted for 27.7% and inline image-N, DON, PN and SS accounted for 16.2, 24.7, 45.4 and 37.2%, respectively, of the total annual loads during the four storms (Table 6). The largest storm, E2 with the lowest value of API7, contributed the largest loads, particularly PN and SS.

Table 6.   Nitrogen loads during the storm events and the percentage of the annual load
 Load (kg)Percentage of the annual load
SSTNinline imageNDONPNSSTNinline imageNDONPN
  1. DON, dissolved organic nitrogen; PN, particulate nitrogen; SS, suspended solids; TN, total nitrogen.



Impact of hydrological characteristics

Antecedent soil moisture is one of the most important factors for watershed hydrological response to rainfall. A previous study by Rusjan et al. (2008) showed that a higher API value could cause a faster hydrological response (shorter Tp). Comparing the API7 and API21 values in our study, the value of API7 was more consistent with the previous study, suggesting that the hydrological response depended more on the antecedent soil moisture of the shallow groundwater aquifer than the deep groundwater level. However, our study showed that the Tp was longer than other headwater streams probably because of the larger size of our watershed (Rusjan et al. 2008; Zhang et al. 2007, 2008). The shallow groundwater level showed that saturation in near-stream areas was dictated by antecedent moisture condition (i.e. ponding in E3 with the highest API value and antecedent groundwater level; Fig. 2; Table 1) and that shallow groundwater might be recharged by subsurface flow and return flow from contributing hill slopes and those flows are most likely to be responsible for the continued saturation of the near-stream areas even after precipitation cessation and hydrograph recession (i.e. E1 and E2; Fig. 2). Thus, subsurface flow could dominate the receding limb of the hydrograph and might be a big contributor to inline image-N export.

The run-off coefficient is another indicator of wetness in a watershed, which reflects not only antecedent soil moisture, but also other factors such as evapotranspiration, growth of vegetation and plant cover. Our increasing run-off coefficient values from summer to autumn appear to imply that plant growth and evapotranspiration processes also play an important role in the watershed hydrological cycle, except for the increasing groundwater level. The highest API and antecedent groundwater level both were observed in E3, but the largest run-off coefficient was in E4. This might be related to the decrease in water use by plants and evaporation in autumn.

Rainfall intensity was found to have a negative correlation with inline image-N, but a positive correlation with PN (Zhang et al. 2008). However, there was no correlation between nutrients and rainfall intensity in our study, probably owing to a lag in the hydrological response to rainfall (longer Tp; Table 1). A correlation between nutrient concentrations and cumulative rainfall showed that the amount of rainfall was a controlling factor for nutrient export in the Shibetsu watershed. To analyze the correlation between nutrient concentrations and rainfall intensity, further monitoring needs to be conducted in the headwater stream.

Major sources of PN, inline image -N and DON

Temporal changes in stream chemistry reflect the sequence in which hydrological flow paths link source areas to streams. As catchment wetness increases during storm events, the nutrients in the stream reflect the sources that are available and the amount of water that flows through the source areas of those nutrients. Although the concentrations of inline image-N, DON and PN all were larger on the rising limb of the discharge hydrograph, the differences including the export time, the export patterns and the export load distribution in the discharge imply that they could have resulted from different sources with different export mechanisms.

Previous studies have reported a significant positive correlation between nutrient concentrations and discharge, but these relationships varied depending on study sites as well as rain events (Ahearn et al. 2004; McNamara et al. 2008). Our study showed that different patterns not only existed among storm events, but also among different nutrients at the same study site, which implied variation in a mechanism of nutrient export. Particulate N always had a clockwise pattern and significant “first flush” indicated that it was mainly sourced from the surface flow as a result of soil erosion, and a significant correlation with SS could provide evidence for this. In contrast, the patterns of dissolved N varied with different shallow groundwater level or antecedent soil moisture and the “first flush” was very weak. Thus, transport of dissolved N might be better related to subsurface run-off or groundwater, which is always affected by shallow aquifer or antecedent soil moisture. Therefore, we speculated that PN could have been derived from the surface soil, whereas dissolved N appeared to be mainly transported by subsurface flow or originated from the groundwater.

Inamdar et al. (2004) believed that inline image-N export occurred via deep flow paths and a rise in inline image-N concentrations was associated with the rapid displacement of till water by infiltrating precipitation, using base cations (Ca2+ and Mg2+) as indicators. McHale et al. (2002) and Iqbal (2002) also reached a similar conclusion. We used Si as an indicator, which is only derived from rock weathering in a deep layer, and the result is shown in Fig. 7. An opposite trend was found between Si and inline image-N concentrations, which was different from the parallel trends previously observed (Inamdar et al. 2004), implying that inline image-N was not from deep groundwater; and the mechanism must differ from the explanation of Inamdar et al. (2004).

Figure 7.

 Concentrations of Si and inline imageN during the four storm events (E1–E4).

Creed and Band (1998a) attributed the early peaks in the inline image-N concentration to the “flushing” hypothesis, in which nutrients are leached from near-surface layers by a rising water table followed by quick lateral transport of these leached nutrients to the stream via near-surface subsurface stormflow on the hillslope and/or surface, saturation excess run-off in the riparian zone. Hydrologically, a key feature of this perceptual model is full-column saturation of the soil profile or water table rises high enough into the soil profile to encounter near-surface soils with high transmissivity and thus the potential for significant lateral flow to occur. This “transmissivity feedback” has now become a common hypothesis to explain lateral flushing of labile nutrients (Bishop et al. 2004). In our study, the shallow groundwater level rose to above a depth of 20 cm under the soil surface during the storm events, and above the ground surface in the wet antecedent condition in E3 (Fig. 2). This indicates that the hydrology requirement for inline image-N flushing was met in our study. In addition, a key requirement of the “flushing” hypothesis biogeochemically also exists, that is, the ready availability of excess nutrients in the near-surface soil horizons (Qualls and Haines 1991; Weiler and McDonnell 2006). Previous studies have found a highly significant positive correlation between inline image-N concentrations in stream water and the proportion of upland area in Shibetsu watershed under a baseflow condition (r = 0.89, Hayakawa et al. 2006; r = 0.84, Woli et al. 2004), which indicated that the upland area was the most important factor for determining inline image-N concentrations. Woli et al. (2004) also stated that the regression coefficient had a significant positive correlation with the cropland surplus N, chemical fertilizer N and manure fertilizer N. Meanwhile, our soil chemical analysis revealed that the inline image-N concentration at a depth of 0–20 cm in the soil layer was significantly higher than that at 20–50 cm and 50–100 cm depths (Fig. 8). The above studies all indicated that higher N application rates resulted in greater field surplus N accumulated in the top 0–20 cm soil layer and had the potential to leach into the stream in the Shibetsu watershed. In addition, local famers usually apply fertilizer onto grasslands in spring and summer, and apply manure in spring or autumn (the annual application rate of chemical fertilizer N is 80 kg ha−1 and an equal amount of manure fertilizer N is added; Agricultural Department of Hokkaido Government 2002); there is a high risk of inline image-N leaching during the rainy seasons if fertilizer management is not appropriate (Jia et al. 2007). Thus, we believe that the source area of inline image-N flushing was variable in the near-surface soil layer in upland areas because of the rising shallow groundwater level. The shallow groundwater level consistently reached a maximum (closest to the soil surface) before the peak in inline image-N concentration and discharge (Fig. 4), supporting the flushing of inline image-N.

Figure 8.

inline image-N concentrations in the soil profile (ANOVA, P < 0.05, n = 27).

On the basis of the variable source area (VSA) concept, for a given watershed, N flushing may have been regulated not by the total VSA, but rather by the rate of change in the expanding source area (dVSA/dt). Although the rate would have been regulated by topography (Creed and Band 1998a), we believe that the rate may also be regulated by the antecedent shallow groundwater level, which affects the water connectivity of the watershed in a rain event and may lead to a different dVSA/dt. For example, E1 needs more time to connect the hydraulic connectivity because of the shallow groundwater level and dry antecedent soil moisture, which makes inline image-N flux move more slowly from the variable source areas at the first stage (smaller values of dVSA/dt). This may be one reason for the counterclockwise direction of the hysteresis loops of E1 as a “prolonged flushing mechanism” (Rusjan et al. 2008; Weiler and McDonnell 2006).

A similar pattern of concentrations between DON and inline image-N in all storm events was observed in our study (Fig. 9). Several studies have found DON peaks on the rising limb similar to our finding (Buffam et al. 2001; Hill et al. 1999; Inamdar and Mitchell 2007; McHale et al. 2000; Vanderbilt et al. 2003). However, Vanderbilt et al. (2003) attributed this pattern to the flushing of decomposing leaf litter; Inamdar and Mitchell (2007) and Hill et al. (1999) reported that stream DON was derived from throughfall. In contrast, Hagedorn et al. (2000, 2001) found DON peaks on the recession limb and attributed this to mobilization of DON during its passage through the forest canopy and organic-rich topsoil. In the case of the Shibestu watershed, we found that DON peaked after a maximum value of the shallow groundwater level and the trends observed for DON and the shallow groundwater level were very similar, particularly in E3 (Fig. 9). Therefore, we speculate that DON may correlate with groundwater rising, and that it may lead to flushing, just like inline image-N. However, we have no evidence to confirm the source of DON, owing to a lack of analyses of throughfall, litter and soil water. However, we noticed that the concentrations of DON in E3 and E4 (in autumn) were significantly higher than the concentrations in E1 and E2 (Table 2). Hagedorn et al. (2001) attributed high DON exports to elevated decomposer activity and the availability of fresh leaf litter in autumn. Hayakawa et al. (2006) reported a positive relationship between upland area (grassland area) and DON concentrations in the Shibetsu watershed. Decomposing of grass litter and manure application in autumn could be the reason for the relationship. Thus, our finding may confirm the flushing of decomposed leaf litter on the forest and decomposed manure and leaf litter on the grassland.

Figure 9.

 Concentrations of inline image-N and dissolved organic nitrogen (DON) and the shallow groundwater level during the four storm events (E1–E4).

Implication of the “first flush” for watershed management

Two observations about the “first flush” effect in our study may have important implications for the management of water resources: (1) high concentrations and fluxes of SS and PN exported rapidly for a short time at the beginning of the discharge, (2) slow recession of nitrate concentration and long flushing time in the falling limb lead to half of the inline image-N flux export. The higher initial SS and PN concentrations early in the run-off that are associated with a first flush have important interaction with the removal efficiency of the BMP. It has been noted that many BMPs, such as ponds and wetlands, operate at great efficiency in removing particles including PN during storm events if the pond or wetlands can effectively reduce the nutrients that are transported at the rising limb of discharge during the first flush (Luo 2008). Gentle sloping landform has the potential to last long time of water retention; and agricultural landscapes have typically been found to leach greater amounts of inline image-N (Hayakawa et al. 2006; Woli et al. 2004). The geological characteristics associated with agricultural operations in Shibetsu watershed might lead to a long flushing time of inline image-N, which is likely to contribute to a much higher inline image-N concentration and flux to the water body. Thus, appropriate fertilizer management in summer and autumn in the Shibetsu watershed, when rainfall is high, is apparently crucial for preventing excess N losses to streams during storm events. In addition, Hayakawa et al. (2006) studied the impact of wetlands on inline image-N concentration by comparing the Shibetsu watershed with another wetland watershed and found that wetlands play an important role in the attenuation of inline image-N. The spatial distribution of wetland or riparian areas in the watershed and their connectedness with the stream network will be a key for the delivery of inline image-N and thus regulate the flushing response. However, this application must be carefully undertaken because the regulation of wetlands or riparian forests with respect to inline image-N flushing is complicated (Inamdar and Mitchell 2006).


Although concentrations of N and SS all peaked at the rising limb of discharge, PN and SS consistently displayed clockwise hysteresis between concentration and discharge and had a short flush time constant (t1/e), whereas dissolved N had a long t1/e and different patterns among storms associated with antecedent soil moisture. These differences indicate that PN originated from the surface soil and was transported by surface flow. In contrast, dissolved N was related to subsurface flow. inline image-N was leached from near-surface soil layers by the rising water table, followed by quick lateral transport of the leached inline image-N to the stream via subsurface flow on the cropland. Although the concentration patterns of DON and inline image-N were similar, DON could have been closely related to groundwater rising and decomposed leaf litter and manure. Further study is required to take throughfall, litter and soil water into consideration.

Although the first flush of N and SS were observed, PN and SS were more significant and contributed 80% of fluxes within the first 50% of the discharge, whereas dissolved N with slowly decreased concentrations led half loads to export during the recession of the discharge. The different flushing mechanisms between particulate nutrients and dissolved N imply that different BMPs should be conducted for effective mitigation of N export on a watershed scale.


We thank Dr Toshiya Saigusa, Mr Osamu Sakai, Mr Hiroyuki Koda and Mr Toshinobu Koba, Hokkaido Konsen Agricultural Experiment Station, for their help and suggestions on this study. This study was partly supported by Japanese Grants-in-Aid for Science Research from the Ministry of Education, Culture, Sports, Science and Technology (No.14209002 and No.19201008) and The Strategic International Cooperative Program “Comparative Study of Nitrogen Cycling and Its Impact on Water Quality in Agricultural Watersheds in Japan and China” by the Japan Science and Technology Agency. The research grant was provided by the “Hokkaido Regional Development Bureau for the Restoration Project in the Shibetsu River” and the Global COE program “Establishment of Center for Integrated Field Environmental Science”, MEXT, Japan.