Environmental Management

, Volume 47, Issue 3, pp 398–409

Nutrient Response Modeling in Falls of the Neuse Reservoir

Authors

    • Modeling and TMDL Unit, Department of Environment and Natural ResourcesNorth Carolina Division of Water Quality-Planning
    • Department of Marine, Earth, and Atmospheric SciencesNorth Carolina State University
  • Jie Li
    • College of Ocean and MeteorologyGuangdong Ocean University
    • Department of Marine, Earth, and Atmospheric SciencesNorth Carolina State University
Article

DOI: 10.1007/s00267-011-9617-4

Cite this article as:
Lin, J. & Li, J. Environmental Management (2011) 47: 398. doi:10.1007/s00267-011-9617-4
  • 158 Views

Abstract

In order to study system responses of Falls of the Neuse Reservoir (Falls Lake) to varied nutrient loadings, a coupled three-dimensional hydrodynamic and eutrophication model was applied. The model was calibrated using 2005 and 2006 intensive survey data, and validated using 2007 survey data. Compared with historical hydrological records, 2005 and 2007 were considered as dry years and 2006 was recognized as a normal year. Relatively higher nutrient fluxes from the sediment were specified for dry year model simulations. The differences were probably due to longer residence time and hence higher nutrient retention rate during dry years in Falls Lake. During the normal year of 2006, approximately 70% of total nitrogen (TN) and 80% of total phosphorus (TP) were delivered from the tributaries; about 20% (TN and TP) were from the sediment bottom. During the dry years of 2005 and 2007, the amount of TN released from sediment was equivalent to that introduced from the tributaries, indicating the critical role of nutrient recycling within the system in dry years. The model results also suggest that both nitrogen and phosphorus are limiting phytoplankton growth in Falls Lake. In the upper part of the lake where high turbidity was observed, nitrogen limitation appeared to dominate. Scenario model runs also suggest that great nutrient loading reductions are needed for Falls Lake to meet the water quality standard.

Keywords

NutrientsEutrophicationWater quality modelingFalls LakeBenthic flux

Introduction

Water quality issues in lakes, ponds and reservoirs have long been concerns worldwide. Based on an assessment report by the European Environment Agency (EEA) published in 2009 (EEA 2009), both nitrate and phosphorus concentrations in European rivers and lakes generally decreased during the last 14 years, reflecting the effect of measures to reduce agricultural inputs of nitrate and the general improvement in wastewater treatment and reduced phosphate content of detergents over this period. The United States (U.S.) has approximately 40 million acres of lakes, ponds, and reservoirs. Johnson (1989) reported that for the decade following the passage of the Clean Water Act in 1972, the majority of the lake area experienced a decline in water quality in the U.S. Based on the most recent (2002–2008) reports by the states to the U.S. Environmental Protection Agency (EPA), about 14% of the assessed lakes, ponds, and reservoirs were impaired by nutrients.

Nutrients are the third most reported cause for lake impairment in the U.S., following mercury and polychlorinated biphenyls (PCB). Nutrient enrichment in lakes and ponds is often associated with eutrophication symptoms such as increased frequency of algal blooms, expanded oxygen depletion of bottom waters, decreased water clarity, and abnormal pH values, which further impact aquatic life (Cloern 2001; Boesch 2002). In addition, excess nutrient pollution in water-supply lakes and reservoirs is expected to promote increased harmful cyanobacterial abundance and have a direct adverse impact on the value of the potable water resources (Burkholder 2002, 2009; Touchette and others 2007).

Despite the well-accepted relationship between nutrient enrichment and other symptoms of eutrophication, a universal threshold of nutrient load to different systems is not readily defined, partly because of the variation of hydrological and geophysical characteristics between systems leads to different nutrient capacities (Cloern 2001; Scheffer and van Nes 2007). Historical nutrient loadings and compositions of nutrient types often matter as well (Johnes 1999). Without in-depth examination, the response and sensitivity of an individual system to varied nutrient loadings and its controlling mechanisms are not known. Such information is often required to assist with environmental policy development and selection of management strategies.

In order to study the responses of chlorophyll a concentrations (chl a) to varied nutrient loadings in Falls of the Neuse Reservoir and to assist with nutrient management strategy development in the watershed, a coupled three-dimensional hydrodynamic and eutrophication model was applied to the system. The model results were then analyzed to examine the controlling factors of variations of chl a concentrations within the system.

Methods

Study Area

Falls of the Neuse Reservoir (Falls Lake) is the primary water supply for the City of Raleigh, and surrounding towns in Wake County, North Carolina (NC) of the U.S. Falls Lake is located in the upper portion of the Neuse River Basin, near the fast-developing Triangle Area in North Carolina. The dam is located just north of the City of Raleigh, and the lake extends northwest about 35.4 km upstream to the confluence of the Eno, Flat, and Little Rivers near the City of Durham. The reservoir was constructed and filled by 1983; it is currently operated by the United States Army Corps of Engineers (USACE).

The main stem of Falls Lake extends from relatively deeper channels (around 15 m) close to the dam upward to shallower arms (less than 1 m) downstream of the confluence of the Eno, Little, and Flat Rivers. The average depth of the reservoir is about 3.2 m. During summer, the reservoir is usually strongly stratified with hypoxic bottom waters in the deeper region. The total surface area of the lake at normal pool is about 50.2 km2 (12,410 acres), and the total volume is about 0.162 km3 (131,395 acre-feet).

Falls Lake drains a watershed area of approximately 1994 km2 (770 square miles). Fifty-eight percent of the area is covered by forest, 18% by agriculture and 11% by development (DWQ 2009a). A calibrated watershed model applied to the upper five sub-watersheds (Flat River, Eno River, Little River, Ellerbe Creek, and Knap of Reeds Creek) estimated that agriculture and point sources are the major contributors of total nitrogen (TN) and total phosphorus (TP) to the lake (DWQ 2009a)

An intensive survey conducted by NC Division of Water Quality (DWQ) during 2005–2007 shows that chl a concentrations in Falls Lake have frequently exceeded the State’s water quality standard of 40 μg/l (DWQ 2009b). A nutrient management strategy is being developed for Falls watershed. In addition, Falls Lake is on the State’s impaired waters list for chl a standard violations.

Field Observation

An intensive survey was conducted by DWQ from March, 2005 to September, 2007 to support model calibration and validation. Data were collected approximately every two weeks at thirteen stations within Falls Lake (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig1_HTML.gif
Fig. 1

Study area and model grid. The depth of the model cell ranges from about 1 to 10 m (e.g., NEU010—1.3 m; NEU13B—2.7 m; NEU018E—5.2 m; NEU020D—8.8 m)

Physical parameters including water temperature, pH, conductivity, and dissolved oxygen were measured at multiple depths in the water column at intervals equal to or less than 1 m. For the algae and nutrient data, composite photic-zone water samples were collected at each station. A photic-zone composite was obtained by lowering an integrated depth sampling device (Labline water sampler) to twice the Secchi depth and filling while slowly raising to the surface (DWQ 2006a). The water samples were then analyzed for chl a, ammonia/ammonium (NHx), nitrite/nitrate (NOx), total Kjeldahl nitrogen (TKN), phosphate, total phosphorus (TP), and total organic carbon (TOC) concentrations in the lab. More detailed information about the collection procedures and analytical methods used are available (DWQ 2006a).

EFDC Model

The three-dimensional, coupled hydrodynamic and water quality model Environmental Fluid Dynamics Code (EFDC), was selected to simulate water quality variations in Falls Lake. EFDC was selected in order to have a reasonable representation of the complicated bathymetry of Falls Lake, and at the same time, to include three different algal groups (i.e., diatoms, blue-green algae, and other) observed most frequently in Falls Lake.

Environmental Fluid Dynamics Code has been identified as an acceptable tool for the development of Total Maximum Daily Loads by the U.S. Environmental Protection Agency (US EPA 1997). It has been successfully applied in many types of water courses in previous studies, including estuaries, lakes, and coastal seas (e.g., Kuo and others 1996; Shen and others 1999; Ji and others 2001; Lin and Kuo 2003; Shen and Haas 2004; Park and others 2005; Lin and others 2007, 2008; Xu and others 2008).

EFDC has four sub-models: hydrodynamic, sediment transport, eutrophication (water quality model), and other pollutant transport model. The hydrodynamic model was the foundational sub-model, which simulated water surface elevation, current, and temperature. These parameter inputs were provided to the other sub-models, and at the same time, biogeochemical processes regarding the concerned variables (e.g., sediments and nutrients) were calculated in the corresponding sub-models. A brief description of the hydrodynamic, sediment transport, and water quality sub-models follows.

The hydrodynamic sub-model was developed by Hamrick; a detailed description of the formulations and solution schemes used can be found at Hamrick (1992, 1996). The model solved the Navier–Stokes equations for a water body with a free surface. In the vertical direction, sigma coordinates, with the hydrostatic assumption, were used in the model. For this study, four vertical layers were used. Horizontally, curvilinear orthogonal grids were used. Cell lengths varied from around 60 m to about 1,000 m. A total of 519 cells were used in the model (Fig. 1).

Turbidity was relatively high in the upper part of Falls Lake. High concentrations of suspended sediment often block the light for phytoplankton growth and hence are closely linked with phytoplankton dynamics in the lake; therefore, the sediment transport model was activated in this study. The sediment transport model was coupled with the hydrodynamic model with the same time resolution. Multiple classes of sediments were simulated in the model (Kim and others 1998; Lin and Kuo 2003). In this study, three sediment classes were selected to represent washload, clay, and silt/fine sand.

The water quality sub-model of EFDC (Park and others 1995; Tetra Tech 2007) consists of a water column water quality model and a sediment diagenesis model (constant nutrient flux were assumed in this study) linked internally. The water column water quality model simulated the spatial and temporal distributions of 22 state variables in the water column (Table 1). For each state variable, a mass conservation equation was solved. The kinetic processes were formulated after CE-QUAL-ICM (Cerco and Cole 1993, 1994; Cerco 1995), with differences listed in Park and others (1995, 1998, 2005). A detailed description of kinetic processes and their mathematical formulations used in the eutrophication sub-model can be found in Park and others (1995), Tetra Tech (2007), and Ji (2008).
Table 1

EFDC model water quality state variables

(1) Cyanobacteria: Bc

(2) Diatom algae: Bd

(3) Green algae: Bg

(4) Stationary algae: Bm

(5) Refractory particulate organic carbon: RPOC

(6) Labile particulate organic carbon: LPOC

(7) Dissolved organic carbon: DOC

(8) Refractory particulate organic phosphorus: RPOP

(9) Labile particulate organic phosphorus: LPOP

(10) Dissolved organic phosphorus: DOP

(11) Total phosphate: PO4t

(12) Refractory particulate organic nitrogen: RPON

(13) Labile particulate organic nitrogen: LPON

(14) Dissolved organic nitrogen: DON

(15) Ammonia nitrogen: NH4

(16) Nitrate nitrogen: NO3

(17) Particulate biogenic silica: SU

(18) Dissolved available silica: SA

(19) Chemical oxygen demand: COD

(20) Dissolved oxygen: DO

(21) Total active metal: TAM

(22) Fecal coliform bacteria: FCB

Surface Boundary

Surface boundary was specified for both the hydrodynamic and the eutrophication sub-models of EFDC.

Air temperature, wind, cloud cover, relative humidity, solar radiation, and rainfall data were needed for calculation of the heat flux, water mass flux, and surface drag at the interface between air and water. These data were recorded from the RDU airport station and obtained from the NC State Climate Office. In cases of missing data from the RDU station, data from the Reedy Creek station was used.

In the eutrophication model, air depositions of NHx and NOx were included as nutrient sources into the surface layer of the water column. Two forms of air deposition were considered here: dry deposition and wet deposition. Dry depositions were the depositional nutrient fluxes during dry (non-raining) days. Wet depositions were the depositional nutrient fluxes accompanied with rainfall. Dry and wet deposition nutrient fluxes suggested by a data survey study from the Highway Stormwater Program of the NC Department of Transportation were adopted in this model study (NC DOT 2008).

For wet depositional flux, weekly wet chemistry data from the National Atmospheric Deposition Program (NADP) station NC41 (Finley Farms, Wake County, North Carolina) were used in the model.

For dry chemistry inputs, Clean Air Status and Trends Network (CASTNET) station PED108 (Prince Edward, Prince Edward County, Virginia) was found to be the best available source of weekly average data for NHx and NOx. The weekly fluxes calculated by Multi Layer Model and downloaded from http://www.epa.gov/castnet/data.html were used. The Prince Edward station is located approximately 65 miles north of Falls Lake watershed in Prince Edward County, Virginia.

River Boundary

Eighteen river inputs/outputs were specified in the Falls Lake modeling domain to represent 17 tributaries and one outward discharge at the dam. For gaged and monitored tributaries, observational data were used to derive the river inputs. The river inflow data was downloaded from the USGS web site at: http://waterdata.usgs.gov/nwis. The water quality observational data can be downloaded from the EPA STORET Program at http://www.epa.gov/storet/dbtop.html.

For un-gaged and un-monitored tributaries, concentration data from a close-by tributary were used and flow data from a close-by tributary multiplied by a distribution ratio were specified. The distribution ratio was calibrated based on the ratio between the un-gaged drainage basin area and the close-by tributary basin area.

Benthic Boundary

Benthic nutrient fluxes were measured by DWQ in April 2006 at two sites: NEU013B and NEU018E. Since certain temporal and spatial variations of the nutrient flux rates were reported, and neither the spatial nor the temporal resolutions of the data were enough to develop a dynamic representation of sediment nutrient fluxes for model use, constant values were specified in the model. The measured nutrient flux rates were used as an indication of the ranges of the parameter values and their order of magnitudes. The exact nutrient flux values were selected through model calibration processes and are listed in Table 2.
Table 2

Constant benthic nutrient fluxes specified in the model (unit: g/m2/day)

 

NHx

PO4

NOx

SOD

Field data

NEU013B

0.05

ND

ND

−1.39

NEU018E

0.01

ND

ND

−0.78

Model

2005

0.02

0.0023

0.00

−1.20

2006

0.01

0.001

0.00

−1.20

2007

0.02

0.0023

0.00

−1.20

ND no significant amount was found

Model Parameters

A number of parameters need to be specified in the model. Parameters were selected based on field observations whenever available. In cases where direct field data were not available, the selection of model parameters were based on literature review and model calibration processes. Table 3 lists some critical model parameters used in this study, together with values used in other studies.
Table 3

Parameters used in the model

 

Units

Fort Cobb Lakea

Chesapeake Bayb

Neuse Riverc

Pamlico Soundd

Cape Fear Rivere

Falls Lake

Maximum algal growth rate at 20°C

1/day

1.5–1.8

2.25–2.5

2.0

2.5

2.3

2.6

Nitrogen half saturation for algal growth

g/m3

0.05

 

N/A

0.01

0.05

0.05

Phosphorus half saturation for algal growth

g/m3

0.002

 

N/A

0.001

0.005

0.002

Background light extinction coefficient

1/m

0.475

N/A

N/A

0.475

0.55

0.475

Light extinction for TSS

1/m per (g/m3)

0.015

N/A

N/A

0.015

0.015

0.015

Light extinction for total suspended chlorophyll

1/m per (mg/m3)

0.041

0.017

N/A

0.017

0.017

Riley (1956)

Optimum temperature for algal growth

°C

23.0–32.0 (Bc)

20.0–27.5

N/A

20.0–26.0

20.0–26.0

23.0–32.0 (Bc)

17.0–23.0 (Bd)

17.0–23.0 (Bd)

20.0–25.0 (Bg)

20.0–25.0 (Bg)

Algal basal metabolism rate at 20°C

1/day

0.01–0.04

0.003–0.04

0.05 (as death rate)

0.01

0.01

0.01–0.04

Algal predation rate

1/day

0.01–0.1

0.01–0.215

0.1

0.12

0.01–0.1

Algal settling velocity

m/day

0.01–0.25

0.0–0.35

0.015–1.5

0.15

0.15

0.01–0.1

PO4 partition coefficient

l/mg

0.04

N/A

N/A

0.002

0.004

0.04

Carbon/chlorophyll

mg/μg

0.04–0.065

0.06

0.05

0.06

0.06

0.06–0.065

Nitrogen/carbon

mg/mg

0.167–0.176

0.167

0.14

0.14

0.16

0.167–0.176

PCprm1 (constant for phosphorus/carbon)

mg/mg

N/A

42

50

40

30

30

PCprm2 (constant for phosphorus/carbon)

mg/mg

N/A

85

N/A

0

40

0

PCprm3 (constant for phosphorus/carbon)

 

N/A

200

N/A

200

200

0

Minimum organic phosphorus hydrolysis rate

1/day

0.01–0.1

0.005–0.1

0.1 (rate at 20°C)

0.005–0.1

0.005–0.1

0.01–0.1

Minimum organic nitrogen hydrolysis rate

1/day

0.005–0.075

0.005–0.075

0.1 (rate at 20°C)

0.005–0.075

0.005–0.075

0.005–0.075

Maximum nitrification rate

1/day

0.2

0.07

0.05

0.07

0.09

0.06

Benthic PO4 flux

g P/m2/day

0.001

N/A

N/A

0–0.001

0.002

0.001–0.0023

Benthic NH4 flux

g N/m2/day

0.05

N/A

N/A

0–0.04

0.02

0.01–0.017

Benthic NO3 flux

g N/m2/day

0.002

N/A

N/A

0–0.004

0.002

0.000

aODEQ (1996); b Cerco and Cole (1994); c Lung and Paerl (1988); d Lin and others (2007); e Lin and others (2008)

Results

Field Observations

Long-term (1984–2007) averaged rainfall near Falls Lake appeared to be evenly distributed among different months; the USACE estimated long-term lake inflow showed a seasonal pattern of higher inflows during winter-spring, and lower during summer-autumn. Variations from the long-term average were quite different among 2005, 2006 and 2007 (Fig. 2). Comparatively, 2005 and 2007 were dry years with monthly inflows typically less than the long-term means. By contrast, much higher variations (both positively and negatively departing from the long-term average) were observed during 2006, which resulted in a year with total rainfall and inflows close to the long-term average.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig2_HTML.gif
Fig. 2

Monthly rainfall and inflow during 2005, 2006, and 2007, compared with long-term (1984–2007) data at Falls Lake

An along-channel gradient was observed in both chl a and nutrient (TN and TP) concentrations in Falls Lake (Fig. 3). Higher concentrations were observed in the upper part of the lake where both chl a and nutrient concentrations peaked at station ELL10, located at Ellerbe Creek Cove. The maximum chl a concentration observed within Falls Lake during the intensive survey period was 230 μg/l (observed on August 9, 2007 at station ELL10), slightly less than the historical (1984–2001) maximum concentration of 280 μg/l, observed on July 16, 1986 at station NEU10 (ELL10 was not monitored at that time).
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig3_HTML.gif
Fig. 3

The annually averaged Chlorophyll a (Chl a), TN and TP distributions along the channel of Falls Lake. Error bars indicate standard deviation

The spatial chl a gradient was likely due to the confluence of the five major tributaries of Falls Lake (the Eno, Little, and Flat Rivers and Ellerbe and Knap of Reeds Creek) at the upper part of Falls Lake, contribution more than 60% of flow and more than 70% of the tributary nutrient loading of the entire lake. As flow entered the lake and decelerated, phytoplankton started to grow, and nutrient levels tended to decrease as water reached the lower part of the lake.

Compared with the long-term average, the annually-averaged chl a concentrations appeared to be lower during 2005 and higher during 2007, especially at the upper part of the lake. The 2006 annually averaged chl a concentrations appeared to be between 2005 and 2007 values and relatively closer to the historical values at most stations. The particularly low annually averaged chl a concentrations during 2005 were due to unavailability of chl a data from April 11 to August 23, 2005 (the expected peak phytoplankton growing season), because of analysis problems identified at the DWQ lab (DWQ 2006b)

If we define the phytoplankton growing season in Falls Lake as starting from the first incidence (during a year) of an observed chl a concentration greater than the State’s standard of 40 μg/l, and ending at the last such incidence, then the growing season spanned through early January (57 μg/l was reported on January 4, 2007 at ELL10) to mid-November (51 μg/l was reported on November 15, 2005 at both NEU010 and NEU013), almost all year long.

In an effort to examine the controlling mechanisms of phytoplankton dynamics in Falls Lake, correlations were calculated between each water quality parameter observed during the intensive survey period. Results are listed in Table 4. The observed chl a concentrations appeared to be significantly correlated with both nitrogen and phosphorus, suggesting both forms of nutrient play important roles in phytoplankton dynamics in Falls Lake. Chl a was also positively correlated with TOC and water temperature. TOC was positively correlated with many of the water quality parameters examined, probably because TOC had multiple sources including local production (e.g., from phytoplankton) and tributary delivery. Although high concentrations of chl a occurred throughout most of the year, the highest concentrations were usually observed during summer when water temperature was high.
Table 4

Correlation between observed water quality variables

 

Chl a

TN

TP

TOC

DO

Temp.

Chl a

1

0.45

0.46

0.34

0.02

0.22

TN

 

1

0.89

0.34

−0.03

−0.001

TP

  

1

0.37

−0.06

0.04

TOC

   

1

0.09

0.1

DO

    

1

−0.63

Temp.

     

1

Underline indicates correlations that are statistically significant at P < 0.05

Model Calibration and Validation Results

Due to differences in hydrographs and rainfall patterns between 2005 and 2006, we decided to calibrate the lake model separately for 2005 and 2006, and use 2007 as the model validation period where all model parameters were taken from the 2005 model calibration processes. While calibrating 2005 and 2006 model runs separately, the goal was to change as few model parameters as possible. The parameters that varied (between 2005 and 2006 model simulations) were river input representations, organic matter settling velocities, and benthic nutrient fluxes.

In general, the model performed reasonably well regarding water level, chl a, TN, TP (Fig. 4), temperature, TOC and DO simulations (DWQ 2009b). Since a primary goal for this model study was to simulate the responses of the chl a standard exceedance rate to varied nutrient loadings, the model-predicted chl a exceedance rate at different levels was closely examined at station NEU013B as well as the total exceedance rate over all monitoring stations (Fig. 5). Overall, within the range of the observed chl a concentrations, the model-predicted chl a exceedence rate (fraction of time model-predicted chl a concentration exceeded certain values) followed the trend suggested by the observations. Both model results and observations suggest that at station NEU013B, chl a concentrations were evenly distributed between approximately 20 and 80 μg/l, while more than 70% of the observed and modeled chl a concentrations existed between 20 and 80 μg/l at all the monitoring stations. Chl a concentrations for all monitoring stations in Falls Lake were above 80 μg/l approximately 5% of the time.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig4_HTML.gif
Fig. 4

Comparisons between model-simulated and observed lake surface elevation, Chl a, TN and TP concentrations in 2005, 2006 and 2007

https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig5_HTML.gif
Fig. 5

A comparison of the model-simulated and observed fraction of times chlorophyll a concentrations were above certain values at NEU013B (upper panel) and over all monitoring stations (lower panel) in Falls Lake

Compared with 2006 model simulations, the 2005 and 2007 model simulations appeared to have slightly underestimated TN, TP and Chl a concentrations. The underestimation is probably because the constant values of the benthic nutrient flux specified in the model introduced more uncertainties during the dry years of 2005 and 2007. More data are needed to have a better representation of time-varying benthic nutrient fluxes in Falls Lake.

Responses to Nutrient Reduction

In order to examine the response of chl a exceedance rate to various levels of nutrient input in Falls Lake, multiple model runs were made for a combination of TN and TP reductions at 25, 50, 75, and 100% of their original loadings. Nutrient reductions were applied to the five upper major tributaries since they typically contributed greater than 70% of all tributary nutrient loading and the upper part of the lake had the highest chl a standard exceedance rate (DWQ 2009b). In order to save computational time, scenario runs were made on a yearly basis. 2006 was selected as the baseline year since it was recognized as the most close-to-normal year according to rainfall and lake level observations.

The model cell containing NEU013B was selected as the baseline area for the calculation of the corresponding chl a standard exceedance rate under different nutrient loading conditions. NEU013B was located just downstream of the conjunction of the five upper major tributaries, where high chl a concentrations were observed. Chlorophyll a concentrations at NEU013B were influenced by nutrient input from all of the five upper major tributaries. Model-simulated chl a concentrations at the surface and subsurface layer were averaged to be consistent with composite field samples collected from water columns above twice the Secchi disk depth.

The model results (Fig. 6) suggest that in order to lower the chl a exceedance rate from above 50% to below 10% (indicated as the thick line in Fig. 6) at the baseline area of NEU013B, large TN and/or TP reductions will be needed. Initially, the model-predicted chl a standard exceedance rate responded more readily to nitrogen than to phosphorus but became more sensitive to phosphorus loading differences when 30% or higher TP reduction was applied. This finding is consistent with the trend suggested by model-simulated limitation factors for phytoplankton growth.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig6_HTML.gif
Fig. 6

The model-predicted chlorophyll a standard exceedance rate at NEU013B in Falls Lake under different nutrient loading reduction scenarios

Limitation Factors for Phytoplankton Growth

Figure 7 shows the model-simulated limitation functions as Nutrient, Light, and Temperature at the monitoring stations in Falls Lake during 2006. The lower the value of the limitation factor is, the more stringent the limitation.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig7_HTML.gif
Fig. 7

Model-simulated limitation functions at the surface (upper panel) and bottom (lower panel) layers of the monitoring stations in Falls Lake during 2006. The lower the value of the limitation function is, the more stringent the limitation. The temperature limitation function shown here is for the 3rd algae group (“other algae,” i.e., algae species besides blue-green algae and diatoms)

The model results suggested that nutrients severely limited phytoplankton growth, especially at the surface. This finding also confirmed that water-quality conditions in Falls Lake were nutrient sensitive. At the upper portion of Falls Lake during 2006, nitrogen limited phytoplankton growth the majority of the time (Fig. 8). By contrast, at the lower portion of the lake, algal growth was limited by nitrogen and phosphorus at a similar frequency. The dominance of nitrogen limitation at the upper portion of Falls Lake probably explained the insensitivity of the model-predicted chl a exceedence rate to reduced phosphorus loading at the current conditions.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-011-9617-4/MediaObjects/267_2011_9617_Fig8_HTML.gif
Fig. 8

Model-simulated fraction of time phytoplankton is N or P limited at the surface layer of the monitoring stations in Falls Lake during 2006

The observed TN:TP mass ratio during the 2005–2007 intensive survey period varied between 2 (at NEU019E, r2 = 0.04) to 8 (at NEU020D, r2 = 0.45) at different locations in Falls Lake. The total TN:TP mass ratio within the entire lake was 5.5 (r2 = 0.79), indicating nitrogen was primarily limiting phytoplankton growth. This was consistent with the findings from the model.

Temperature dependency of phytoplankton growth rate followed a seasonal pattern (not shown). On a yearly basis, temperature limitation on phytoplankton growth was not as stringent as nutrient limitation. In the bottom waters, phytoplankton growth was limited by light especially in the lower part of the lake with deep channels. At such places, stratification may have blocked nutrients from being mixed into surface waters.

Discussion

Sources of Nutrient Loading

In order to understand the response of chl a levels to tributary nutrient loading, an analysis of model-represented nutrient sources and their relative magnitude was conducted. In addition to nutrient loading introduced from tributaries, other nutrient sources exist in Falls Lake, including benthic nutrient fluxes and atmospheric deposition (Table 5).
Table 5

Nutrient loading to Falls Lake

 

2005

2006

2007

TN (103 kg N/year) (% total)

TP (103 kg P/year) (% total)

TN (103 kg N/year) (% total)

TP (103 kg P/year) (% total)

TN (103 kg N/year) (% total)

TP (103 kg P/year) (% total)

Tributary

313.4

31.2

553.0

70.5

236.0

28.1

45.1

44

71.7

80.3

38.3

41.4

Atmospheric

36.2

N/A

45.6

N/A

34.6

N/A

5.2

 

5.9

 

5.6

 

Benthic

345.6

39.7

172.8

17.3

345.6

39.7

49.7

56

22.4

19.7

56.1

58.6

Among the nitrogen sources, atmospheric deposition appeared to have a minor contribution to the total nutrient loading in Falls Lake, accounting for 5–6%. Atmospheric deposition here refers to the summation of the dry and the wet deposition that were deposited directly onto lake surfaces. Air deposition to land surfaces within Falls watershed and later introduced into lake as runoff was accounted for in the tributary loading. Unfortunately, air deposition data for phosphorus is not available and hence not accounted in the model.

Tributary loading contributed around 70–80% of total nutrient loading during 2006, but less than 50% during dry years of 2005 and 2007. During dry years, nutrient flux from the lake bottom appeared to be relatively high, equivalent to the annual average total inputs from air and tributaries. This trend applies to both nitrogen and phosphorus. Although no significant benthic flux was measured for phosphorus during April 2006, bottom hypoxia, which often exist in deeper regions of Falls Lake during summer, may have promoted phosphorus release from bottom sediment.

The average residence time within Falls Lake was around 4 months during 2006, but around 6–7 months during 2005 and 2007. Exogenous nutrients tended to be retained in systems with slow transport and long residence times (Cloern 2001). Nutrient retention and recycling between sediment and water column appeared to play an important role in the pelagic nutrient cycle in Falls Lake.

It is also noteworthy that uncertainties are associated with the model estimation of the benthic nutrient fluxes. The benthic nutrient fluxes were specified in the model at a constant rate. The values were selected to be within the range of the observations in Falls Lake and fine-tuned through model calibration. On one hand, the relative magnitudes and inter-annual trends of the specified benthic nutrient fluxes were expected to be reasonable, on the other hand, uncertainties were still associated with the spatial and temporal variations of the nutrient fluxes and further monitoring will be needed for a better model representation.

Uncertainties Related with Nutrient Reduction Scenarios

In order to predict the allowable tributary loading for Falls Lake to meet the water quality standard, other sources of nutrient loading were assumed to be constant. This assumption arises from our lack of prediction capability on future conditions such as air deposition, weather condition, land use changes, and benthic condition.

As shown by Table 5, benthic nutrient fluxes tended to be higher during dry years with lower tributary loading and longer residence time. If the frequency of such drought incidences were to increase in the future as a result of climate change, that will shift the “normal year” as drier than what our assessment was based on (i.e., the 2006 condition). The possibly longer lake residence time will likely lead to more stringent nutrient reduction requirement or longer restoration time (Schindler 2006). In contrast, as tributary nutrient loading reductions are carried out, organic matter introduced into the lake will likely be reduced. As less organic matter settles to the lake bottom, nutrient fluxes released from the sediment into the water column will probably be reduced as well. If reduced benthic nutrient flux did occur, it would probably alleviate the nutrient enrichment problem in the lake, resulting in the water quality standard being met at a less stringent nutrient reduction scenario.

Nitrogen Versus Phosphorus

Model results suggested that both nitrogen and phosphorus were limiting primary production in Falls Lake, with nitrogen limitation appearing to dominate in the upper portion of the lake. The observed TN:TP mass ratio of 5.5 (12.2 by atoms) was lower than the Redfield ratio (16 by atoms), and the critical TN:TP value (44.2 by atoms) suggested by Elser and others (2009) to have a N-limited phytoplankton growth in lakes.

By contrast, lake primary production was generally thought to be phosphorus limited (Schindler 1977); the control of point sources of phosphorus reduced algal blooms in many lakes (Holtan 1981; Schindler 2006). Although some lakes were also reported to be naturally nitrogen-limited (Goldman and others 1993; Steinhart and others 2002), the increasing inputs of nitrogen from atmospheric deposition (especially in high deposition areas) have led many lakes to be phosphorus limited (Elser and others 2009). Increased nitrogen loading may reduce lake phytoplankton biodiversity by possibly favoring those species competing best for the limiting phosphorus (Elser and others 2009). In addition, eutrophication of estuaries and coastal seas where phytoplankton growth was primarily limited by nitrogen also called for reductions in nitrogen inputs from their drainage basins (Turner and Rabalais 2003).

Further, the ratio of nitrogen to phosphorus in anthropogenic nutrient inputs was usually low; as a result of accumulated anthropogenic input, highly eutrophic lakes tended to have lower TN:TP ratios. Nitrogen limitation is significantly more frequent in lakes of low ambient TN:TP (mass ratio ≤14) (Downing and McCauley 1992).

To control nitrogen inputs seemed a natural choice for highly eutrophic lakes, such as Falls Lake, where phytoplankton growth was dominated by nitrogen limitation. However, lake experiments (Schindler and others 2008) showed that reducing only nitrogen inputs favored algal blooms of the nitrogen-fixing cyanobacteria. Nitrogen fixation was enough to allow biomass to be produced in proportion to available phosphorus and the lake remained eutrophic. Reducing phosphorus loadings still appeared to be a must to reduce eutrophication.

In the case of Falls Lake, considering the research results listed above and the model indications, a great percent reduction in phosphorus together with a moderate percent reduction in nitrogen was recommended as a reasonable approach in reducing lake-wide eutrophication.

Acknowledgments

We appreciate the helpful advice provided by Falls Lake Technical Advisory Committee during the development period of the nutrient response model. The technical review of the model application by Tetra Tech is also appreciated. Finally, we are very thankful to Andy Painter and Kathy Stecker of NC Division of Water Quality for proof-reading this manuscript.

Copyright information

© Springer Science+Business Media, LLC 2011