JEQ Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 9 August 2006
Published in J Environ Qual 35:1742-1755 (2006)
DOI: 10.2134/jeq2005.0373
© 2006 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Volk, J. A.
Right arrow Articles by Ullman, W. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Volk, J. A.
Right arrow Articles by Ullman, W. J.
Agricola
Right arrow Articles by Volk, J. A.
Right arrow Articles by Ullman, W. J.
Related Collections
Right arrow Water Quality
Right arrow Watershed and Landscape Processes
Right arrow Ecosystem Management
Right arrow Nutrient Cycling
Right arrow Nitrogen

TECHNICAL REPORTS

Landscape and Watershed Processes

Nitrogen Loads through Baseflow, Stormflow, and Underflow to Rehoboth Bay, Delaware

J. A. Volka, K. B. Savidgeb, J. R. Scudlarkb, A. S. Andresc and W. J. Ullmanb,*

a Watershed Assessment Section, Delaware Department of Natural Resources and Environmental Control, 820 Silver Lake Boulevard, Suite 220, Dover, DE 19904
b College of Marine Studies, University of Delaware, 700 Pilottown Road, Lewes, DE 19958-1298
c Delaware Geological Survey, University of Delaware, Newark, DE 19716-7501

* Corresponding author (ullman{at}udel.edu)

Received for publication September 26, 2005.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
A detailed study of water and nitrogen (N) discharge from a small, representative subwatershed of Rehoboth Bay, Delaware, was conducted to determine total N loads to the bay. The concentrations of ammonium (NH4+), nitrate + nitrite (NO3– + NO2–), and dissolved and particulate organic N were determined in baseflow and storm waters discharging from Bundicks Branch from October 1998 to April 2002. A novel hydrographic separation model that accounts for significant decreases in baseflow during storm events was developed to estimate N loads during unsampled storms. Nitrogen loads based on gauged flows alone (7100–19 100 kg/yr) significantly underestimated those based on land use–land cover (LULC) and estimated N export factors from different classes of LULC (32 000–40 600 kg/yr). However, when ungauged underflow and associated N loads were included in the total loads (25500–33800 kg/yr), there was much better agreement with LULC export models. This suggests that in permeable coastal plain sediments, underflow contributes significantly to N fluxes to estuarine receiving waters, particularly in drier years. Based on the similarity in LULC, N loads from the Bundicks Branch subwatershed were used to estimate upland loads to the entire Rehoboth Bay Watershed (259 000–316000 kg/yr). These N loads from the watershed were much greater than those from direct atmospheric deposition (49000–64 500 kg/yr) and from a local wastewater treatment plant (9700–13700 kg/yr). While the watershed was the principal source of N at all times during the year, the relative contributions from the watershed, wastewater, and direct atmospheric deposition varied predictably with season.

Abbreviations: LULC, land use–land cover • PON, particulate organic nitrogen • RBWTP, Rehoboth Beach Wastewater Treatment Plant • TDN, total dissolved nitrogen • TN, total nitrogen


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
MANY ESTUARIES along the Mid-Atlantic and Gulf coasts of the United States suffer from eutrophication caused by elevated nutrient loadings from domestic, municipal, industrial, and agricultural practices in their watersheds (Quinn et al., 1989; USEPA, 1998, 2002; Bricker et al., 1999). Direct surface water and ground water discharge from upland watersheds, atmospheric deposition, wastewater discharge, and in some cases, exchange with the coastal ocean contribute nutrients to these estuaries. In the estuary, nitrogen (N) and phosphorus promote primary production, leading to high levels of plant biomass, changes in plant community structure, and potentially noxious and toxic algal blooms (Vollenweider, 1992; Jorgensen and Richardson, 1996; National Research Council, 2000; de Jonge et al., 2002). Eutrophication resulting from nutrient over-enrichment can increase estuarine turbidity, alter habitats necessary for indigenous flora and fauna, reduce biodiversity, and lead to large diurnal excursions in oxygen concentration, which in extreme cases, can lead to fish and shellfish kills (Vollenweider, 1992; Delaware Department of Natural Resources and Environmental Control, 1995; de Jonge et al., 2002).

The small coastal bays of the mid-Atlantic are particularly susceptible to the problems of eutrophication. These bays, many of which are shallow and poorly flushed by tidal processes, are characterized by high human population pressure, much of which is focused in seasonal resort communities, and have extensive rural-residential and agricultural land use in their watersheds (Quinn et al., 1989; USEPA, 1998, 2002). These land use–land cover (LULC) characteristics result in extensive export of nutrients to estuarine receiving waters.

The mid-Atlantic Coastal Plain is characterized by low topographic relief and highly permeable soils and aquifers. As a result, a substantial fraction of the discharge and associated nutrient loads from the watershed travels through subsurface pathways to receiving streams or directly to estuarine receiving waters (Dillow and Greene, 1999). While the discharge to streams can be gauged for the determination of nutrient loads, the discharge that bypasses gauging stations is often ignored in nutrient balances.

The objective of this research was to determine the seasonal and interannual variability of total N loads from the watershed through all available pathways including ungauged ground water discharge to a shallow estuarine lagoon in the mid-Atlantic Coastal Plain. This information is needed by regulators to better manage N export from this watershed. To achieve this objective, water and N budgets and the flow processes that control these budgets were studied in detail at a small, representative subwatershed.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Study Location
Rehoboth Bay (Fig. 1 ) is the northern-most member of the chain of interconnected coastal bays along the Atlantic coastline of the Delmarva Peninsula. The bay is small (74 km2), drains a watershed of 150 km2 with no major tributaries, and has no direct connection to the adjacent coastal ocean. Rehoboth Bay is linked to the ocean through Indian River Bay to the south, and to Delaware Bay to the north through the Lewes-Rehoboth Canal. The coastal-plain watershed is dominated by agricultural and forested land uses (Table 1) but has substantial urban and suburban centers that have central sewage disposal in the northeast (Rehoboth Beach) and to the southwest (Long Neck). The population is predominantly seasonal, reflecting the nearby beach resorts, although the number of year-round residents is increasing along with increases in residential and commercial land uses throughout the watershed (Ames and Dean, 1999; Delaware Office of State Planning Coordination, 1999). Our studies were focused on Bundicks Branch (Fig. 1), a small watershed that drains to Rehoboth Bay and has LULC characteristics similar to that of the larger watershed (Table 1).


Figure 1
View larger version (60K):
[in this window]
[in a new window]
  		Fig. 1. The Inland Bays watershed showing the locations of Rehoboth Bay, the Bundicks Branch and Millsboro Pond subwatersheds and gauging stations, the Rehoboth Beach Wastewater Treatment Plant (RBWTP), the urban and sewered areas of the watershed, and the Georgetown and Cape Henlopen rain collection sites.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Land use characteristics of the Rehoboth Bay Watershed and the Bundicks Branch subwatershed (1997 data; Delaware Office of State Planning Coordination, 1999).

 
Hydrographic Data
The U.S. Geological Survey-Water Resources Division (USGS) determined stage heights and stream discharge at 15-min intervals for Bundicks Branch (at Robinsonville, Delaware; USGS DE01484654; Fig. 1) during the period of August 1998 to March 2000. A rating curve was developed and maintained by USGS following standard stream gauging protocols (Rantz and others, 1982a, 1982b) during this period. After March 2000, stage heights were recorded at 15-min intervals using an Isco 6700 automated water sampler equipped with an Isco 720 pressure sensor (Teledyne Isco, Lincoln, NE) and the rating curve was modified and maintained independently of the USGS. During the period when USGS determined discharge, step corrections were applied to the rating curve when new discharge measurements indicated that changes in stream cross-sections had shifted the rating curve from a previous state. From March 2000 to April 2002, rating curves were prorated smoothly with time between flow measurements as intermediate recalibrations were performed with insufficient frequency to determine step corrections. The uncertainty in these measurements is likely greater than during the earlier period.

Baseflow Sampling
Baseflow water samples were collected every 2 to 4 wk at Bundicks Branch from October 1998 to April 2002. Grab samples were collected in 1-L plastic bottles and a portion of each sample was immediately filtered in the field through glass-fiber filters (GF/F; Whatman, Brentford, UK) to separate dissolved from particulate N phases. The filtered and unfiltered samples were kept on ice and in the dark until returned to the laboratory (within 6 h). Filtered samples were frozen until analyzed for dissolved N concentrations. Particles in the remaining unfiltered samples were collected on GF/F filters for the determination of particulate organic N (PON) and other parameters. All of the water quality data collected at this site is available electronically from the Delaware Geological Survey (Andres et al., 2002).

To determine how the frequency of baseflow sampling affected the determination of nutrient exports from the Bundicks Branch catchment to Rehoboth Bay, samples were collected at a higher frequency during three time periods. On one occasion, samples were manually collected for nutrient analysis every day for a month. On two occasions during baseflow periods, samples were collected every hour for 1 d using the Isco 6700 sampler.

Stormwater Sampling
Water samples were collected automatically using the Isco 6700 water sampler during 17 storms from July 1999 to April 2002 (Table 2). Sampling was initiated by an increase in stage height of 0.02 ft (0.61 cm) above the previously established baseflow height. After initiation, the sampler collected samples at fixed 2- or 3-h intervals for up to 46 h or until the discharge returned to within 10% of pre-storm values. Samples were kept in the dark until retrieved from the field for laboratory processing. During the warmer months, the Isco 6700 sample compartment was loaded with ice or ice packs to keep samples cool after collection. In the laboratory, water samples were filtered and processed for the determination of dissolved N species and PON as described above for the baseflow samples.


View this table:
[in this window]
[in a new window]
  		Table 2. The measured and estimated flow-weighted average total nitrogen (TN) concentration in stormwater (CSF) used to estimate storm loads from the Bundicks Branch subwatershed for specified periods.

 
Analytical Methods
Dissolved N concentrations were determined by automated segmented-flow colorimetry using a Flow Solutions IV Analyzer (O/I Analytical, College Station, TX). Dissolved ammonium (NH4+) was determined using the phenol hypochlorite method (Glibert and Loder, 1977; Grasshoff and Johansen, 1972); nitrate + nitrite (NO2– + NO3–) was determined by the sulphanilamide/N(1-napthyl) ethylene diamine method after cadmium reduction of NO3– to NO2– (hereafter this parameter is referred to as nitrate, NO3–) (Glibert and Loder, 1977); and total dissolved nitrogen (TDN) was determined as NO3– after oxidation in an autoclave by potassium persulfate (D'Elia et al., 1977; Glibert and Loder, 1977; Solorzano and Sharp, 1980). Dissolved inorganic N (DIN) was calculated from the sum of NO3– and NH4+ concentrations. Dissolved organic N (DON) was determined as the difference between TDN and (DIN).

Particulate organic nitrogen (PON) was determined by combustion using a Carlo Erba Model EA1108 CHNS Analyzer (CE Elantech, Lakewood, NJ), calibrated using ethylenediamine-tetraacetic acid and phenylalanine. Total N loads from the watershed were calculated based on total N concentrations (TN = TDN + PON).

Conceptual Model
The TN load to an estuary is the sum of dispersed loads from the watershed, point discharges to the estuary, dispersed loads from the atmosphere, and in some circumstances, contributions from adjacent water bodies. In the case of Rehoboth Bay, routine monitoring by the National Atmospheric Deposition Program at the nearby Cape Henlopen Site (Fig. 1) provided the basis for determining direct atmospheric N deposition to the bay surface during the study period (Scudlark et al., 2005; National Atmospheric Deposition Program, 2004). (Atmospheric deposition to the upland watershed was included in the N export from the watershed as determined below.) The Rehoboth Beach Wastewater Treatment Plant (RBWTP; Fig. 1) is the only point source of N to Rehoboth Bay. Average daily discharge and TN concentrations from the RBWTP, as reported monthly to the Delaware Department of Natural Resources and Environmental Control, provided the basis for determining N loads from this source. No present estimate exists for the net import of N from adjacent water bodies (Indian River Bay and Delaware Bay), but these are thought, based on low concentrations and net outflow of water, to be negligible compared to other inputs and are neglected in the present analysis. This paper is primarily focused on the determination of the discharge and N loads from the upland watershed to Rehoboth Bay, for the comparison with loads from RBWTP and direct atmospheric deposition.

Nitrogen may be transported from a watershed to its receiving waterbody by streams, as either baseflow or stormflow, by ground water, or by atmospheric processes. Each of these pathways exhibits its own chemical characteristics and spatial and temporal variability and all of these components must be properly summed to yield a total load to the receiving body (Freeze and Cherry, 1979; Genereux and Hooper, 1998). In many estuarine watersheds, the export of N to the estuary may be approximated by the sum of baseflow and stormflow loads, each of which may be calculated as the product of discharge and concentration over a specified time interval:

Formula 1[1]
where Li are loads, Ci are concentrations, Qi are water discharges, and the subscripts i = G, BF, and SF refer to gauged, baseflow, and stormflow, respectively. At a stream gauging station, the sum of baseflow and stormflow discharges, QG, is determined from stage height and a rating curve:

Formula 2[2]

Nitrogen concentrations, CG, are determined by the analysis of collected samples. While it is possible to determine CBF from sampling during no-rainfall periods, it is not possible to determine CSF directly. It is also not practical to sample and determine N concentrations at the high frequency needed to properly estimate stormloads during all storm periods. Therefore, storm loads, LSF, must be calculated from the difference between gauge loads, LG, and baseflow loads, LBF, for sampled storms to serve as the basis for estimating N export for unsampled storms:

Formula 3[3]

This approach, however, requires that gauged flow, QG, be separated into its baseflow, QBF, and stormflow, QSF, components during and shortly after storms. For this work, a novel hydrographic separation algorithm appropriate to the coastal plain setting was developed (see below).

For reasons also discussed below, a third loading term is necessary in the porous and permeable coastal plain sediments of Bundicks Branch where a significant fraction of the total discharge, with its N burden, is found to travel below, adjacent to, and/or parallel to the stream-gauging station where QBF and QSF are determined. This additional component of discharge is termed "underflow" (Larkin and Sharp, 1992; Jackson, 1997; Sharp, 1999). The overall transport of N from the watershed to Rehoboth Bay is therefore given by:

Formula 4[4]
where LUF, CUF, and QUF refer to the load and concentration of N, and discharge of water associated with the underflow component. In this study, QUF was estimated from contrasting hydrological balances at Bundicks Branch and a nearby larger watershed (Millsboro Pond, Fig. 1), which has unit discharges (discharge normalized to watershed area) typical of the Delmarva Peninsula (Johnston, 1976). CUF was estimated from ground water data collected around the Rehoboth Bay Watershed (see below). Equation [4] was used to estimate the total export of N from Bundicks Branch. The total export normalized to total watershed area (unit export) was then used to infer the discharge and loading to the estuary from the unsewered areas of the Rehoboth Bay Watershed. Nitrogen export from sewered areas was determined from the measured ground water concentrations in these areas and unit discharge at Millsboro Pond. Exports from the entire watershed were then compared to the N loads from wastewater and atmospheric sources.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Gauged Discharge
Substantial variation in average monthly discharge was observed at Bundicks Branch during this study (Fig. 2 ). During the period of October 1998 to February 2000 the monthly rainfall at Georgetown and Cape Henlopen reached the long-term average (National Atmospheric Deposition Program, 2004; University of Delaware Research and Education Center, 2004) only twice. From March 2000 to August 2001, long-term monthly averages were exceeded eight times; from September 2001 to April 2002 the averages were exceeded only once. The gauged discharges clearly reflect the rainfall patterns with the period of March 2000 to August 2001 showing more than 2.5 times the gauged discharge of the periods before and after (Fig. 2).


Figure 2
View larger version (32K):
[in this window]
[in a new window]
  		Fig. 2. Rainfall (at both Georgetown and Cape Henlopen) and gauged discharge at Bundicks Branch during the study period. Monthly rainfall did not vary greatly across the watershed. The baseflow contributions to the gauged discharge, QBF, determined by the traditional and alternative hydrographic separation models are shown. There are substantial differences between these models only during the wettest months.

 
Hydrograph Separation
There are a number of algorithms for separating the total gauged discharge, QG, of a stream during and after storms into two components, baseflow (or pre-event, QBF) and storm (or event, QSF) flow (Gray, 1970; Freeze and Cherry, 1979; Maidment, 1993; Fetter, 1994). Most of these traditional algorithms assume that QBF remains essentially constant or increases slightly during storm periods and assign the remaining discharge to stormflow, QSF, an assumption that is consistent with the behavior of larger streams and rivers. CBF is also assumed to remain constant. In preliminary application, these algorithms did a poor job of modeling the average hydrographic and hydrochemical character of Bundicks Branch and of determining the average N concentration of storm water, CSF. The traditional algorithms yielded many impossible negative estimates of storm concentrations and loads calculated from Eq. [1] and [2] during storm periods assuming constant QBF and CBF:

Formula 5[5]

This suggested that QBF and/or CBF decreased dramatically during storms. Our intensive daily sampling clearly indicated that CBF varied little with time over short time scales (Jennings, 2003), consistent with the deeper transport pathways associated with baseflow waters, and therefore QBF during storms must have been significantly overestimated by the traditional algorithms. The high frequency sampling also indicated that multivariate estimation of N concentrations (e.g., Cohn et al., 1989) during unsampled periods was unnecessary.

The importance of underflow in this coastal plain setting is not surprising. A number of authors have noted that the streams may switch from discharge- ("gaining") to recharge- ("losing") dominated streams in response to significant increases in stage height and discharge. Bonell et al. (1998) noted this behavior in tropical "high-rainfall, response-dominated" catchments in Australia and used an isotopic tracer to demonstrate that baseflow can be totally suppressed at the peak of storm discharge. Wegner (1997) noted a complete reversal of the hydraulic gradient in the riparian zone of an Alaskan stream during snow melt, leading to significant bank and aquifer recharge in this system. Perhaps most relevant to this study is the work of Johnston (1976), who reported reversals of the riparian hydraulic gradient and reductions of up to 50% of integrated baseflow to coastal plain streams during storms on the Delmarva Peninsula. According to Johnston (1976), this gradient reversal generally dissipated within 24 h. Bundicks Branch is a small and "flashy" stream that shows significant changes in stage height and discharge in response to storm events and it is therefore reasonable to assume that it has hydrographic behavior similar to that found in the above studies.

No fully conservative tracers (stable isotopes, Cl–) were determined at Bundicks Branch and chemical sampling was limited to a subset of all storms. We therefore developed a relationship based on stage height, a parameter measured at all times during the study, to separate the stormflow, QSF, and baseflow, QBF, components of discharge during storms:

Formula 6[6]
where HG is the measured stage height, Ho is the stage height just before the storm, and Hmax is the peak stage of the storm. This algorithm has the effect of reducing QBF to zero for a short period at the peak of the stormflow, and yielded only a few apparently negative concentrations of CSF when applied to the sampled storms. It was therefore used consistently throughout this study. Figure 3 shows a representative analysis of discharge during a storm event on 18 Mar. 2002 using both the traditional separation method and the method described by Eq. [6]. The traditional method represents a minimum estimate of QSF and the alternative method represents a minimum estimate of QBF (Fig. 2). Based on the method given by Eq. [6], QSF {approx} 63% (QBF {approx} 37%) of QG for an average storm.


Figure 3
View larger version (29K):
[in this window]
[in a new window]
  		Fig. 3. Comparison of the traditional hydrographic separation model with the alternative model used in this study. The traditional method yields a maximum value of gauged baseflow (QBF), a minimum gauged stormflow (QSF), and therefore apparently negative values of stormflow nitrogen concentration (CSF) and storm loads (LSF) during some periods during storms. The alternative procedure yields a minimum QBF, maximum value of QSF, and only rarely negative values of CSF and LSF.

 
Nitrogen Concentration and Speciation in Baseflow
The principal form of N in baseflow discharge at Bundicks Branch was NO3– (278 ± 42 µM), with dissolved organic N having secondary importance (26 ± 21 µM). Ammonium and PON were significantly lower in concentration at all times of the year (1.6 ± 1.2 µM and 1.8 ± 1.1 µM, respectively). The average concentration of TN was 307 ± 48 µM.

There is limited evidence of a repeatable seasonal pattern in N concentrations and speciation. Only PON concentrations showed a recurring minimum in the fall. However, it does appear that dissolved N concentrations responded to discharge levels. Generally, higher concentrations of NO3– and TDN were found during the periods of lower flow than during periods of higher flow (Fig. 4 ) reflecting the higher NO3– concentrations in the surficial aquifer that becomes the dominant source of stream water during low-rainfall and low-flow periods (Denver, 1989; Andres, 1991; Hamilton et al., 1993; Denver et al., 2004).


Figure 4
View larger version (18K):
[in this window]
[in a new window]
  		Fig. 4. Dissolved N concentrations (NO3– and total dissolved nitrogen, TDN) at Bundicks Branch decreased with increasing baseflow discharge (QBF) during the study period. The concentrations at low flow appear to be a good estimate of median ground water concentrations in the Bundicks Branch subwatershed.

 
Baseflow Nitrogen Loads
Daily baseflow loads of each N species at Bundicks Branch were calculated for the period of October 1998 to April 2002 as the product of average daily QBF (determined by the hydrographic separation model described by Eq. [6] during storm periods) and the corresponding measured or the linearly interpolated concentrations of N (see below). From these results, monthly baseflow loads at the Bundicks Branch gauging site were calculated by summation (Fig. 5 ).


Figure 5
View larger version (48K):
[in this window]
[in a new window]
  		Fig. 5. Relative magnitude and seasonal distribution of total nitrogen (TN) loads (LT = LBF + LSF + LUF, where BF is baseflow, SF is stormflow, and UF is underflow) from the Bundicks Branch watershed to the Rehoboth Bay ecosystem during the study period. Baseflow N loads [= LBF(TDN) + LBF(PON), where TDN is total dissolved nitrogen and PON is particulate organic nitrogen] were highest during high flow periods as TN in baseflow was less variable during the experimental period. Total dissolved N predominated at all times, but PON loads had a seasonal pattern and were largest during the wettest periods.

 
Baseflow N loads varied from approximately 500 kg/mo during the low flow periods to up to 2400 kg/mo during the higher flow periods (Fig. 2 and 5). The primary dependence of baseflow loading on discharge is consistent with the observation that N concentrations varied significantly less than discharge (Fig. 4).

Validity and Precision of Baseflow Loads
Several assumptions and estimates were necessary to compute daily baseflow loadings and ultimately monthly (Fig. 5) and annual N loads. First, it was necessary to assume that one discrete baseflow sample adequately represents the average daily concentrations in the stream on the day of collection. Second, it was necessary to assume that linearly interpolated N concentrations between biweekly to monthly baseflow samples adequately describe daily concentrations on days for which no samples were collected.

To verify these assumptions and to estimate the uncertainties of calculated baseflow loadings, nutrient concentrations were investigated with high frequency sampling (hourly to test the first assumption and daily to test the second). Results of both experimental procedures were similar (Jennings, 2003). Total dissolved N concentrations did not vary by more than 5% within a day or between succeeding baseflow sampling dates and, therefore, uncertainties in N loads are largely controlled by the uncertainties in the discharge measurements (±10%; James et al., 2001). However, large fluctuations in PON (and other particulate parameters), apparently unrelated to precipitation or other hydrologic factors, were observed on both time scales. As a result, uncertainties assigned to PON loads could exceed 40% (Jennings, 2003). At Bundicks Branch, however, PON represents only a small fraction of TN loads during baseflow (Fig. 5). Therefore, the large uncertainty associated with PON loads had little impact on calculated TN loads from baseflow.

Nitrogen Concentrations and Speciation in Storm Discharge
As with the baseflow samples, NO3– was the predominant form of N in stormwater discharge, although other species dominated at different periods of the storm hydrograph and in storms of different magnitudes (Fig. 6 and 7) . In a typical small storm, the concentrations of N species varied little during the storm and were similar to the concentrations observed at baseflow (Fig. 6). There was slight evidence of increases in PON at the beginning of such storms due presumably to the resuspension of sediments from the streambed by increased flow velocities ("first flush") and some hint of an increase in dissolved organic N concentrations at the beginning and end of the storm hydrograph.


Figure 6
View larger version (28K):
[in this window]
[in a new window]
  		Fig. 6. Discharge components, N concentrations and speciation, and N loads during the storm of 10 Apr. 2002, the smallest of the sampled storms. During this storm N concentrations and speciation varied little with discharge and therefore the gauged loads were proportional to discharge.

 

Figure 7
View larger version (40K):
[in this window]
[in a new window]
  		Fig. 7. Discharge components, N concentrations and speciation, and N loads during the storm of 10–12 Aug. 2001. The storm had two peaks and showed clear evidence of a "first-flush" effect during the initial rise in the hydrograph, with particulate organic nitrogen (PON) becoming a dominant component of the N load.

 
During larger storms, there were significant changes in N speciation and TN concentrations (Fig. 7). In a typical large storm, high concentrations of PON at the beginning of the storm dominated the discharge and loading and dissolved organic N concentrations rose and remained high during the entire storm period. By the end of the storm, concentrations of N species returned approximately to the levels found in the baseflow. During the storm shown in Fig. 7, there were actually two peaks of discharge that reflect the timing of the rainfall. The high levels of PON achieved during the first rising hydrograph are not attained during the second, suggesting that N speciation during storms may be a function of storm frequency as well as intensity, a factor not investigated in this study.

Storm Loads of Nitrogen
For each sampled storm event, measured discharges, QG, were integrated over the 2- or 3-h sampling intervals, centered on the actual sample time, and these were used with the measured N concentrations, CG, to determine gauged N loads, LG, for that interval (Fig. 6 and 7). After hydrographic separation, LBF was determined as the product of the estimated QBF and the interpolated baseflow concentration, CBF, from sampling before and after each storm. Storm loads from each interval (LSF) were then determined from Eq. [3]. As seen in Fig. 6 and 7, LBF during storms was always proportional to QBF and the remaining load was assigned to the stormflow (LSF). Finally, calculated loads for each sample interval were summed over the storm duration to yield event loads for gauged flow, baseflow, and stormflow.

The event loads of N attributable to stormflow for the 17 monitored storms at Bundicks Branch are shown in Fig. 8a . There was a great deal of variation between storm loads; however, this variability was mostly related to the variability in rainfall intensity and duration and therefore total discharge associated with each storm. To remove these effects, the weighted average N concentration in storm flow (Formula 6SF) was calculated for each storm (Fig. 8b). Because of the effects of longer-term meteorological fluctuations, no clear seasonal trends in Formula 6SF are observed.


Figure 8
View larger version (20K):
[in this window]
[in a new window]
  		Fig. 8. Storm loads determined by the alternative hydrographic separation method for 17 sampled storms at Bundicks Branch. Although there was substantial variation in the storm loads (a), there was substantially less variation in the average storm concentration (Formula 6SF, b), which can therefore be used to estimate storm loads for unsampled storms from discharge alone.

 
Gauged Nitrogen Loads
The N export characteristics of Bundicks Branch were calculated over various time scales using the above methods for determining baseflow and stormflow export. The interpolation of baseflow N concentrations allowed for the computation of daily, monthly, quarterly, and annual baseflow loads for a period greater than 3 yr. The sampled stormflows indicated that storm discharges contributed significantly to TN loads as well. However, only a small fraction of all of the storms were sampled for the determination of N concentrations and loads and therefore, to correctly account for storm loads over the longer time scales of this study, the loads attributable to unsampled storms had to be estimated.

For this study, Formula 6SF and LSF were estimated for unsampled storms from the results of load determinations from sampled storms. Formula 6SF varied little between adjacent storms (Fig. 8b) and it was therefore concluded that during the monitored periods (12 July 1999 to 10 Apr. 2000 and 19 May 2001 to 10 Apr. 2002), the mean Formula 6SF of adjacent sampled events was used to estimate the storm loads of unsampled storms occurring between those events (Table 2). This approach, however, was not justified for the period before or the longer periods between storm sampling at Bundicks Branch (1 Oct. 1998 to 11 July 1999 and 11 Apr. 2000 to 18 May 2001). To calculate N loads during storms during these unmonitored periods, the average Formula 6SF of all 17 monitored events (2.6 ± 1.1 g/m3) at Bundicks Branch (Table 2) was used. The gauged N loads for Bundicks Branch for the three calendar years of sampling are given in Table 3.


View this table:
[in this window]
[in a new window]
  		Table 3. Annual gauged and total nitrogen (TN) loads from Bundicks Branch to Rehoboth Bay and a comparison with estimates of N loads based on land use–land cover (LULC; Table 1) and export factors provided by the given reference. Results for water years are given in Jennings (2003).

 
The gauged N loads determined in this study are significantly smaller than previously estimated loads based on land use export factors (Table 3; Ritter, 1986; Horsley & Witten, Inc., 1998). This discrepancy suggests that a component of the N load to Rehoboth Bay from the Bundicks Branch subwatershed was not properly included in the gauged N loads and/or that the published export factors were too high.

Underflow and Associated Nitrogen Loads
In regions such as the Rehoboth Bay Watershed, where there is little topographic relief, shallow stream incision into the aquifer, and a highly permeable aquifer, ground water may bypass the local surface-water flow system as underflow. Instead of discharging to the nearest topographic lows (streams) where it can be measured at a gauging station, the underflow enters deeper regional flow systems that discharge at the mouths of major valleys (Freeze and Cherry, 1979; Larkin and Sharp, 1992). In coastal settings such as the Bundicks Branch subwatershed, underflow and its associated N load may discharge directly to nearby lagoons, estuaries, or directly to the ocean (Andres, 1987, 1992, 1995; Dillow and Greene, 1999; Schwartz, 2003; Krantz et al., 2005). Underflow and associated loads, therefore, must be estimated to correctly determine the total nutrient load from upland watersheds to downstream receiving waters.

The potential underflow at Bundicks Branch was estimated from the comparison of hydrological balances at Bundicks Branch and at Millsboro Pond (at Millsboro, DE, USGS DE0148525), the largest subwatershed of the Inland Bays that is located approximately 12 km southwest of Bundicks Branch (Fig. 1). In such watersheds the unit hydrologic balance may be described by (Fetter, 1994):

Formula 7[7]
where P* is precipitation, QG* is gauged surface water discharge, QUF* is underflow, and ET* is evapotranspiration, all normalized to watershed area, over a specified interval of time.

In this study, QUF* was determined on a monthly basis from the contrasting average discharge at Bundicks Branch and Millsboro Pond. Millsboro Pond has an average annual unit discharge (0.35 m/yr) similar to that reported as typical for the Delmarva Peninsula by Johnston (1976) and therefore QUF* can be presumed to be small. Given the close proximity and similar LULC in each watershed, there should be similar values of monthly precipitation and evapotranspiration. Therefore the differences in monthly unit gauged flows at these two stations provide a minimum estimate of QUF* at Bundicks Branch (Fig. 9 ). This minimum estimate was used as the best estimate of QUF* at Bundicks Branch, although a higher value might have been appropriate if QUF* at Millsboro Pond was non-negligible.


Figure 9
View larger version (40K):
[in this window]
[in a new window]
  		Fig. 9. Monthly unit underflow (QUF*) at Bundicks Branch based on the difference between unit gauged flows (QG*) at Millsboro Pond and Bundicks Branch. This is a minimum estimate of QUF*, based on the assumption that there is no underflow at Millsboro Pond. Underflow was a significant fraction of total discharge at Bundicks Branch in all but the wettest periods and dominated in the driest periods.

 
There are a number of compilations of ground water quality data for the Rehoboth Bay Watershed that provided some guidance to estimating CUF (Andres, 1991; Sparco, 1995). In the Rehoboth Bay Watershed, NO3– is the dominant form of N in local ground waters (Denver, 1989) and concentrations are variable and not normally distributed (Table 4). The range of concentrations for the Bundicks Branch subwatershed and the larger downstream Love Creek catchment are statistically indistinguishable from that of the entire Rehoboth Bay Watershed (Table 4). For the purpose of this analysis, the median value (360 µM) for the entire watershed was used as an estimate of CUF. It should be noted that this estimate of CUF is similar to the concentrations of NO3– and TN determined at Bundicks Branch at low discharge when the streamflow was dominated by baseflow input of ground water that might otherwise discharge through underflow (Fig. 4). This provides further indication of the close link between fluxes of ground water, surface water, and chemical constituents. LUF was calculated on a monthly basis using the monthly estimate of QUF and the median concentration of nitrate in the watershed. LUF was clearly a significant fraction of the LT at Bundicks Branch (Fig. 5). During periods of high flow, LUF represented over half of LT; during periods of low flow, over 75% of the TN export from the watershed was through underflow.


View this table:
[in this window]
[in a new window]
  		Table 4. Concentrations of NO3– in ground waters of the Rehoboth Bay Watershed (100 µM N = 1.4 mg N/L).

 
Total Nitrogen Loads from Bundicks Branch
The estimated N export from Bundicks Branch indicates that TN loads vary over monthly, seasonal, and annual time scales. Within a single year at Bundicks Branch, monthly TN loads ranged from 1000 to 5000 kg/mo (Fig. 5). The TN export (LT = LG + LUF), however, had a reproducible seasonal trend that was related to seasonal water discharge patterns that was, in turn, related to seasonal precipitation and evapotranspiration patterns. The highest N loadings occurred during the spring freshet, while considerably lower loadings were observed during the drier summer months. Total annual N loads (including underflow contributions) varied by only 21% during the 3-yr sampling period (Table 3).

The observed seasonal variability of N loads has a potentially significant impact on the ecology of the water bodies that receive these loads and annualized LULC models hide this seasonality. There are a number of factors, in addition to the seasonal runoff patterns, that may influence the seasonal distribution of N loadings from the watershed to streams and ultimately to estuarine waters. These include the patterns of atmospheric deposition to the watershed (Scudlark et al., 2005), N fertilization and harvesting in the watershed, and the patterns of temporary storage of N in soils, biomass, and aquifers. All of these processes, together with external inputs and internal cycling, contributed to the observed discharge patterns at Bundicks Branch and may be important at other similar coastal plain sites.

Reproducible seasonal patterns in LT were observed during both wet and dry periods and in spite of the lack of observed seasonal patterns in LBF or LSF (Fig. 5). This indicates that, when all pathways of water transport are properly accounted for, TN export from watersheds and loads to receiving waters were less dependent on hydrological forcing than on LULC-related N inputs and management practices within the watershed. This is likely to be most true in the case of watersheds with high permeability soils and aquifers, such as those found in many coastal plain settings in the mid-Atlantic region. Models that estimate N export from such watersheds based on surface water measurements alone are likely to severely underestimate total loads to receiving waters and incorrectly determine the seasonal patterns of loading. Such models, therefore, may provide incomplete or erroneous information to managers concerned with eutrophication problems.

Total annual N loads from Bundicks Branch determined in this study compare favorably with estimated loads from LULC export models (Table 3). Ritter (1986) computed N loadings to the Inland Bays based on land use export coefficients determined in the watershed for three broad land use categories. When the export coefficients determined by Ritter (1986) were applied to the LULC for Bundicks Branch (Table 1), the estimated annual load from the Bundicks Branch subwatershed to the receiving waters (32 000 kg/yr) was almost identical to the annual loads measured in this study. Horsley & Witten, Inc. (1998) determined annual N loads to the Inland Bays using an approach similar to that of Ritter (1986) but with a larger number of land use categories and export coefficients from a number of sources (both local and regional). Applying the export coefficients from the Horsley & Witten, Inc. (1998) model to the Bundicks Branch subwatershed yielded a somewhat higher N load (40 600 kg/yr) to Rehoboth Bay, 27% above the loads determined by Ritter and 30% above those found in this study (Table 3).

Nitrogen loads from the Bundicks Branch watershed for the 1999 water year made in a recent HSPF (Hydrological Simulation Program-Fortran; Bicknell et al., 1996) modeling effort (6476 kg/yr; Gutiérrez-Magness and Raffensperger, 2003) corroborate our estimates of surface water export of N for the same period (7028 kg/yr). This agreement, however, is not truly independent as the HSPF model was calibrated, in part, by the same data presented here. Although this model matches the gauged N export at the Bundicks Branch site, it does not account for the quantity of underflow and N export associated with the underflow. As a result, the HSPF model significantly underestimated total unit N exports from the subwatershed and therefore cannot be reliably extrapolated to the determination of TN loads to Rehoboth Bay.

Total Nitrogen Loads to Rehoboth Bay
The total direct loads of N to Rehoboth Bay are derived from four sources. First, there is the contribution from unsewered areas of the watershed, which have similar LULC to Bundicks Branch and therefore can be estimated from the observed discharge from this subwatershed. We used the unit N export determined for Bundicks Branch, uncorrected for slight LULC differences between this subwatershed and the remaining unsewered areas of the Rehoboth Bay (Table 1), as an estimate of this contribution. This estimate does not include the possibility of N attenuation in the aquifer, which is thought to be small based on high dissolved oxygen levels that would inhibit denitrification (Denver et al., 2004), or attenuation during discharge through estuarine sediments. Second, there is the contribution from the RBWTP, which was determined from average monthly discharge and concentration reported to the Delaware Department of Natural Resources and Environmental Control. Treated wastewater effluent from the sewered areas to the southwest of Rehoboth Bay—the Long Neck region—are discharged by spray irrigation and this contribution is therefore taken into account in the ground water export calculation. Third, there is the contribution from streams and ground water in the sewered areas of the watershed. Neither of the sewered areas have major streams that discharge to Rehoboth Bay. Therefore, the unit discharge estimated from the Millsboro Pond subwatershed serves as an estimate of total ground and surface water discharge. In the case of the Rehoboth sewer district, the unit discharge was reduced to account for the discharge that passes through the RBWTP. Median ground water concentrations for the sewered areas (Table 4) were used as the estimate of the concentration of all discharge from these areas. Lastly, there is an atmospheric contribution due to wet and dry deposition directly to the bay surface that was estimated from data collected at the nearby Cape Henlopen rain site (Scudlark et al., 2005; National Atmospheric Deposition Program, 2004).

From the available data, each of these contributions was estimated on a monthly (Fig. 10 ) and annual (Table 5) basis. Over the three years of our study, upland sources (including atmospheric deposition on the watershed) contributed approximately 82% of the N load to Rehoboth Bay, while direct atmospheric deposition and the RBWTP delivered 15% and 3%, respectively (Table 5). Due to seasonal patterns in atmospheric deposition, watershed loading, and discharge from the RBWTP (due to the seasonal resort population), however, the relative importance of each source varied through the year (Fig. 10).


Figure 10
View larger version (38K):
[in this window]
[in a new window]
  		Fig. 10. Comparison of monthly N loads from the watershed to Rehoboth Bay with loads due to atmospheric deposition and municipal wastewater (Rehoboth Bay Wastewater Treatment Plant, RBWTP). Nitrogen loads from the watershed predominated at all times of the year, but the relative importance of N sources varied predictably through the year.

 

View this table:
[in this window]
[in a new window]
  		Table 5. Comparison of loads to Rehoboth Bay from direct atmospheric deposition, the Rehoboth Bay Wastewater Treatment Plant (RBWTP), and the watershed (including gauged, ungauged, and total) for the water years of 1999 to 2001 and for the summer quarters of 1999 to 2001.

 
During the summer (June–August), primary productivity in Rehoboth Bay is largely N-limited (Ullman et al., 1993). Direct atmospheric deposition of N also becomes a relatively larger source (26%), due to the increases in NO3– and NHx (ammonium + ammonia) concentrations in the atmosphere (Scudlark et al., 2005) and decreased watershed discharge and loads due to high evapotranspiration and plant uptake. Overall, the annual values and the range of seasonal values are similar to those produced by other studies that have used different data sets and different assumptions (Ritter, 1986; Horsley & Witten, Inc., 1998; Scudlark et al., 2005; Delaware Department of Natural Resources and Environmental Control, 2000). These results suggest that N-loading models based on LULC may be successfully used in coastal plain settings as long as ungauged underflow is properly included in the TN loads.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
More than 80% of the TN budget of Rehoboth Bay, Delaware, is delivered through ground water and surface water pathways. Direct atmospheric deposition accounts for 15% of the N input. During the summer months, however, the atmospheric fraction becomes more significant (26%) due to reduced runoff (gauged and ungauged flow) from the watershed and higher atmospheric concentrations.

To properly estimate total loads to estuarine receiving waters from coastal plain watersheds with little relief and high soil and aquifer permeability, the discharge of baseflow, stormflow, and underflow must be measured or estimated together with concentrations of N in all of these contributions. Detailed studies of the water balance in a small coastal plain watershed indicate that underflow represents a significant fraction of the total annual discharge (50–75%) and N load (43–75%) that is transported from the watershed to estuarine receiving waters. When this fraction is included in the total export from this watershed, the annual loads are significantly higher and less variable (32 000 ± 3000 kg/yr) than those estimated from gauged flow alone (15 000 ± 7000 kg/yr).

Nitrogen loading models based on export factors from contrasting categories of LULC (32 000–40600 kg/yr) reproduce the measured and estimated loads of N fairly well only when the ungauged (underflow) contribution to the total load is included in the N budget. This suggests that better estimates of N loading would result from the routine inclusion of underflow contributions to nutrient cycling in porous coastal plain settings.


   ACKNOWLEDGMENTS
 
This research was supported principally by USEPA Star Grant R826945 (W.J. Ullman, project director), by the Center for the Inland Bays (J.R. Scudlark, principal investigator), and by the NOAA Air Resources Laboratory (NADP-AIRMON Program; J.R. Scudlark, principal investigator). The Delaware Department of Natural Resources and Environmental Control, the U.S. Geological Survey, and the Delaware Geological Survey provided additional support. J.A. Volk (formerly J.A. Jennings) was supported by a Delaware Water Research Center Graduate Fellowship.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow