Modeling the eutrophication of two mature planted stormwater ponds for runoff control

Ecological Engineering 61P (2013) 601–613

Contents lists available at ScienceDirect

Ecological Engineering journal homepage: www.elsevier.com/locate/ecoleng

Modeling the eutrophication of two mature planted stormwater ponds for runoff control Tove Wium-Andersen a , Asbjørn H. Nielsen a , Thorkild Hvitved-Jacobsen a , Hans Brix b , Carlos A. Arias a , Jes Vollertsen a,∗ a b

Aalborg University, Department of Civil Engineering, Sohngaardsholmsvej 57, 9000 Aalborg, Denmark Aarhus University, Department of Bioscience, Plant Biology, Ole Worms Allé 1, 8000 Aarhus C, Denmark

a r t i c l e

i n f o

Article history: Received 17 March 2013 Received in revised form 3 July 2013 Accepted 4 July 2013 Available online 14 August 2013 Keywords: Eutrophication model Stormwater Wet detention pond pH Dissolved oxygen

a b s t r a c t A model, targeting eutrophication of stormwater detention ponds was developed and applied to simulate pH, dissolved oxygen and the development of algae and plant biomass in two mature planted wetponds for run off control. The model evaluated algal and plant biomass growth into three groups namely; phytoplankton, benthic algae and macrophytes. The study evaluated large data sets from an intensive monitoring campaign of two Danish mature stormwater ponds which were used to calibrate and validate the model. One general calibration covering 6 consecutive months together with two additional short-term calibrations during summer (31 days) and winter (56 days) were carried out applying the data series from one of the ponds. The calibrations showed a good agreement between measured and modeled data for the time spans evaluated. The calibration showed that growth rates for the three groups of primary producers were approximately the same for the two calibrations. The validation for both trends of DO and pH were well simulated, even though pond Aarhus showed a better agreement of the absolute values. The eutrophication model could be successfully calibrated to two stormwater ponds and has the potential of providing a tool for improving the design of stormwater ponds by taking into account the behavior of the plant ecosystem. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Stormwater runoff from roads and urban surfaces often contribute to the deterioration of the aquatic environment. Runoff contains a wide range of contaminants in varying concentrations (NURP, 1983). Among these contaminants are heavy metals, suspended solids, nutrients, and organic micro-pollutants like PAH’s and biocides. If discharged without prior treatment, the pollutants can have detrimental impacts on the flora and fauna of surface waters. Therefore treatment is often required to protect the receiving aquatic environment. Due to the nature of stormwater runoff, treatment facilities must be robust and capable of handling an intermittent runoff pattern as well as a wide range of pollutant concentrations. One treatment technology which has proven robust and efficient in managing both flows and pollutants from stormwater runoff is the wet detention pond. A wet detention pond is an artificial lake with a permanent pool of water designed to detain stormwater runoff. While detained, the quality of the stormwater is improved by sedimentation, adsorption, chemical precipitation,

∗ Corresponding author. Tel.: +45 9940 8504; fax: +45 9940 8552. E-mail address: [email protected] (J. Vollertsen). 0925-8574/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ecoleng.2013.07.032

and biodegradation – hereby mitigating the negative impacts on the receiving waters. The specific load of nutrients to a wet detention pond can be significant and, like in a natural lake, can cause eutrophication of the pond (Mitsch and Reeder, 1991; Gelbrecht et al., 2005). In natural lakes the resulting algal bloom will lead to high water turbidity, high pH and periods of low dissolved oxygen (DO) concentration due to decomposition of dead algae (Torno et al., 1985). Most stormwater systems are located in urban areas and one of the purposes is to improve the environmental quality of the area. A eutrophic stormwater is not wanted as it leads to poor esthetic conditions, reduced ecological quality as well as decreased treatment performance. With respect to wet detention ponds, this process has not been studied in details but could potentially lead to low DO concentrations in the pond. Upon discharge, such water has detrimental impacts on the receiving water fauna and flora (Magaud et al., 1997). Low DO concentrations in the pond could furthermore lead to anaerobic conditions at the pond bottom, resulting in malodors and lowered redox conditions in the sediments. The latter can cause release of phosphorus and heavy metals previously immobilized in the pond sediments (Wong and Yang, 1997). The algal biomass can also absorb pollutants, which thereby are removed from the free water phase (Mehta and Gaur, 2005).

602

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Nomenclature a Apb BA bD,BA bD,MP bD,PP DICrunoff hi I I0 KDIC KI KL,O2 KL,CO2 KN KO2 KP kR,BA kR,MP kR,PP MP P pHinhib PP RSOD ScO2 /DIC SCS SDIC Si SN SO2 SCO2 SOS SP T U10 Vp XBA XMP XPP YC:O2 YN:C YN:O2 YO2 :C YP:C YP:O2 z ˛ ˇ

surface area per unit volume (m−1 ) bottom area of the pond (m2 ) benthic algae including epiphytes (gC m−3 ) decay rate of BA (s−1 ) decay rate of MP (s−1 ) decay rate of PP (s−1 ) dissolved inorganic carbon in runoff predicted value in the base case at the given time step i light intensity in the water column (J m−2 s−1 ) light intensity at the pond surface (J m−2 s−1 ) half-saturation constant for SDIC (gC m−3 ) half-saturation constant for I (J m−2 s−1 ) air–water mass transfer coefficient of O2 (m s−1 ) air–water mass transfer coefficient of CO2 (m s−1 ) half-saturation constant for SN (gN m−3 ) half-saturation constant for SO2 (gO2 m−3 ) half-saturation constants for SP (gP m−3 ) respiration rate of BA (s−1 ) respiration rate of MP (s−1 ) respiration rate of PP (s−1 ) macrophytes (gC m−3 ) base case parameter value bell-shaped function varying between 0 and 1, with optimum at pH equal to 7.5 phytoplankton (gC m−3 ) sediment oxygen demand rate constant (gO2 m−2 s−1 ) Schmidt number (/D) for SO2 and SDIC , respectively CO2 saturation concentration (gC m−3 ) dissolved inorganic carbon (gC m−3 ) normalized sensitivity index at time step i dissolved nitrogen (gP m−3 ) dissolved oxygen (DO) (gO2 m−3 ) concentration of dissolved CO2 (gC m−3 ) dissolved oxygen (DO) saturation concentration (gO2 m−3 ) dissolved phosphorus (gP m−3 ) water temperature (◦ C) wind speed 10 m above ground (10-min average speed) (m s−1 ) volume of the pond (m3 ) biomass of primary producers BA (gC m−3 ) biomass of primary producers MP (gC m−3 ) biomass of primary producers PP (gC m−3 ) C produced per O2 consumed by sediment oxygen uptake (gC gO2 −1 ) N produced per C in biomass produced by photosynthesis or by respiration (gN gC−1 ) N produced per O2 consumed by sediment oxygen uptake (gP gO2 −1 ) O2 produced per C in biomass produced by photosynthesis or by respiration (gO2 gC−1 ) P produced per C in biomass produced by photosynthesis or by respiration (gP gC−1 ) P produced per O2 consumed by sediment oxygen uptake (gP gO2 −1 ) water depth (m) attenuation coefficient for fresh water (m−1 ) attenuation coefficient for phytoplankton (m2 g−1 )



 grd  SOD  wg max,BA max,MP max,PP  KL KL a w | hi |

| P|

ratio between N or P taken up from the sediment to N or P taken up from the water phase (zero for PP and BA; between 0 and 1 for MP) temperature constant for growth, respiration and decay of PP, BA and MP temperature constant for the sediment oxygen demand temperature constant for the water/liquid mass transfer maximum specific growth rate of BA (d−1 ) maximum specific growth rate of MP (d−1 ) maximum specific growth rate of PP (d−1 ) constant in empirical gas–water mass transfer equation exponent in empirical gas–water mass transfer equation air density (g m−3 ) water density (g m−3 ) Switch function (0 or 1) to prevent extinction of primary producers during winter absolute difference in predicted value between the base case (calibration) and sensitivity case (±30%) at the given time step i absolute change in parameter value

To predict the DO concentration in a stormwater pond, it is necessary to simulate the growth and decay of the flora, together with a number of related processes. In other words, it is necessary to set up an eutrophication model of the system. Several eutrophication models have been developed for fresh waters and marine waters; for example the CAEDYM model, which is developed for lakes and reservoirs (Hipsey et al., 2006). However, a stormwater pond differs from a natural lake in several ways (Hvitved-Jacobsen et al., 2010): • Hydraulic load pattern and residence time: In contrast to natural lakes, stormwater ponds only receive input during runoff events. That is, the inflow pattern to a wet pond is characterized by periods of no inflow, interrupted by short periods of high inflow rates. Furthermore, in many ponds the outlet is restricted, resulting in large fluctuations in water volume and water level. Due to the high variability of the inflow pattern, the residence time of the water also varies significantly. • Nutrient load and load pattern: The nutrient load varies with the variation in inflow rate, causing larger temporal load variation for stormwater ponds compared to natural lakes. • Contaminants: The contaminants in stormwater runoff can be toxic to the aquatic flora and fauna, while it is typically not the case for the inflow to natural lakes. For these reasons, a targeted model is needed for simulation of eutrophication in stormwater ponds. Additionally, a rather simple model is preferred to minimize the data requirement for calibrating and validating the model. The objective of this study is to develop an eutrophication model targeted at stormwater ponds. The model shall predict pH, DO and biomass of primary producers under dynamic conditions, to yield better understanding of the biological, chemical and physical performance of such ponds. The model is calibrated and validated on data from two wet detention ponds. The ponds were equipped for continuous monitoring of flow, water levels, turbidity, pH, DO,

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

603

Table 1 Pond and catchment characteristics. Unit Type of catchment

Total catchment area Impervious catchment area Permanent wet volume Retention volume Permanent water depth, max Length to width ratio Average precipitation Annual design runoff

ha ha m3 m3 m – mm y−1 m3

Odense

Aarhus

Light industry

Residential (blocks of flats) 57.4 25.8 6900 1400 1.25 3:1 661 131,900

27.4 11.4 1990 1992 1.45 4.5:1 657 55,500

and temperature. Pollutant loads and discharges were measured on lumped samples. Fig. 1. Planting plan of the species selected for the pond at Aarhus. Modified from Danish Miljoeportal.

2. Methodology 2.1. Study site The studied ponds were constructed as part of an earlier research project in the cities of Aarhus and Odense, Denmark. The two ponds are in the following named pond Aarhus and pond Odense. They receive stormwater runoff from catchments of different land use and have different physical characteristics, Table 1. The ponds consist of one compartment without sediment forebay. The ponds were designed for a retention time of 72 h at a return period of 4 year−1 , i.e. 4 times per year the design residence is 72 h or less, and follow the recommendations of, e.g. Hvitved-Jacobsen et al. (1994), Pettersson et al. (1999), and Vollertsen et al. (2007). The two systems were built and planted in 2007 and have been operational ever since. Besides the hydraulic and structural design the concept comprised the inclusion of plant species with the potential of improving the performances while minimizing a possible visual impact generated by the structure. In the system located in Aarhus the planted pond only two species of plants were selected, namely Phragmites australis and Schoenoeplectus lacustris, in order to blend-in with the adjacent natural Lake Braband (Fig. 1). The second site located in Odense was built in an industrial quarter which permitted the use of a broader variety of plants (11 different species) to improve performance and simultaneously enhance the environmental quality of the site (Fig. 2) (for more details: Isteniˇc et al., 2012).

The measuring campaign revealed that both ponds were subject to illicit discharges. The pond in Aarhus received sporadic discharges of septic wastewaters, which was observed by wastewater odor at the pond inlet and confirmed by detection of E. coli in the inflow. The discharges occurred irregularly, sometimes with intervals of days, sometimes with intervals of months. The source hereof could not be identified. The pond in Odense received sporadic discharges of chemical waste with high concentrations of copper, zinc and lead. The discharges occurred approximately every half year, resulting in temporary copper concentrations above 1000 ␮g L−1 in the pond. The source hereof was not identified. 2.2. Data series The ponds were monitored for inlet and outlet flows as well as pollutant loads. Water levels and water quality parameters were monitored by online sensors in the ponds. Pollutants were measured by automatic sampling. Meteorological data (radiation, air temperature, wind direction and wind speed) were obtained from a nearby weather stations operated by the Danish Meteorological Institute, DMI. Online sensors were placed in the middle of the ponds. They monitored the pH (WTW SensoLyt 700 IQ), temperature (via pH meter), turbidity (WTW VisoTurb 700), DO level (WTW FDO 700 IQ), and water depth (Klay Hydrobar). Measurements were logged

Fig. 2. Planting plan of the species selected for the pond at Odense. Modified from Danish miljoeportal.

604

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

every minute. The sensors were routinely cleaned, the drift of signal logged, and the sensor recalibrated when needed. The data series were subsequently adjusted for sensor drifts and recalibrations. In both ponds the inlet flow was monitored using two fullflowing magnetic flow meters, one to measure the smaller flows (Krone optiflux, DN 150 mm) and one to measure the larger flows (Krone optiflux, DN 500 mm). The flow meters were protected against silting by a grit chamber. The flow meters were coupled in series and with a bypass of the smaller flow meter when flows exceeded its hydraulic capacity. There was furthermore installed a rectangular weir to bypass flow when the hydraulic capacity of the larger flow meter was exceeded. The flow over the weir was estimated from the water level over the weir. An auto-sampler (MAXX TP4-C, 24 sample bottles) was placed at the inlet of each pond after the grid chamber. Additionally, a similar auto-sampler was placed at the middle of the pond in Odense and at the end of the pond in Aarhus. At the inlets the sampling was flow-proportional, while it was time-proportional in the ponds. The auto-samplers were emptied every 14 days or when full. The auto-samplers were equipped with plastic containers and placed underground, i.e. dark and cold to minimized changes of the samples during storage. The subsamples from the bottles of the auto-samplers were pooled and analyzed for total concentrations of copper, lead, zinc, cadmium, nickel, chromium, mercury, total PAH (after USEPA), suspended solids, nitrogen as well as total and dissolved phosphorus. All analyses were performed by an accredited laboratory, following international or European standards (ISO 17294m, manual MK2260-GC/MS, DS/EN 872, DS/EN I 6878aut, manual SM17 udg. 4500 and DS/EN I 11905 auto). Furthermore, grab samples were taken from the ponds for determination of chlorophyll (analyzed after DS/EN 2201). The water samples from Odense covered the time span from mid April 2008 to late September 2009. The samples from Aarhus covered the time span from mid June 2008 to late September 2009. In April 2, 2009, iron salts were added to the sediment in the pond in Aarhus in order to enhance pollutant removal. Therefore, monitoring data obtained after this date was not used for modeling purposes. A period in January and February 2009 was also excluded as both ponds became ice-cover. This surveillance of the ponds has resulted in a large dataset, giving a unique and robust platform for calibrating and validating a model predicting pH and oxygen levels in the stormwater ponds. 2.3. Model outline An eutrophication model targeted at stormwater ponds (Fig. 3) was developed based on eutrophication models for freshwater

systems as described by Hamilton and Schladow (1997) and Jørgensen and Bendoricchio (2001). The model is relatively simple and calculates growth of primary producers subdivided into phytoplankton (PP), benthic algae including epiphytes (BA) and macrophytes (MP) as well as concentrations of DO, CO2 , N and P. Calculated changes in dissolved inorganic carbon (DIC) constituted, together with the pH and the DIC of the incoming stormwater, the basis for calculation of pH. These modeled pH and O2 concentrations were compared to measured data for the purpose of calibrating the model. A pond was modeled as one completely mixed compartment. The mixing of the water in both ponds has been studied by WiumAndersen et al. (2012) by analyzing the time passing from inflow events to response in online measurements (turbidity, pH, temperature and oxygen). The response times were found to be between 0.5 and 2 h and it was concluded that, during a runoff event, the ponds became completely mixed within hours. For a similar pond and using an inert tracer, Madsen et al. (2007) found that between events the studied pond became completely mixed within a day. Compared to the time scale of the biological processes, and considering that the ponds are shallow, it therefore seemed appropriate to use only one completely mixed compartment for simulation. To simulate the light penetration into the pond, the water column was within each time step divided in layers of 0.1 m, simulating plant and algae growth in each layer. After completion of all process simulations of one time step, the water column was again completely mixed prior to simulation of the next time step. 2.3.1. Conditions and process at the system boundary In the model, water only entered or left the ponds through the inlet and the outlet, that is, precipitation on the pond surface, evaporation, exfiltration and infiltration was considered negligible. Inflow to the model was determined by the measured inflow time series, while its temperature was assumed equal to the measured air temperature. The inflow was simulated as having a fixed pH, which was determined from model simulations (Table 3). The inflow was assumed saturated with O2 at the respective temperature. DIC of the inflow was obtained from model calibration. The concentrations of nitrogen and phosphorus in the inflow were estimated based on the measured data series. Both total and dissolved phosphorous were measured, however only on lumped samples covering up to 2 weeks. The dissolved fraction could have been affected by this storage and is hence associated with an unknown uncertainty. For this reason a fixed fraction of the total phosphorous was used as input parameter. The International Stormwater BMP Database (www.bmpdatabase.org) contains a large number of measurements of pollutants in stormwater runoff, here among

Fig. 3. Schematic diagram of the eutrophication model. Wide arrows indicate interactions with the surrounding environment (water and energy). Narrow arrows indicate transformations and gain or losses of chemical compounds and diamonds are sub-parameters used to calculate the main parameters shown in the squares, which were again compared to measurements by the sensors indicated as circles.

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

total and dissolved phosphorous. According to Leisenring et al., 2010 the median of total phosphorous registered in this database was 0.17 mg-P L−1 (3538 samples) while the median of dissolved phosphorous was 0.07 mg-P L−1 (1421 samples). The fixed fraction between dissolved and total phosphorous was hence assumed to be 0.41. In the present study, the median of total phosphorous in the inflow to both the pond in Aarhus and the pond in Odense was 0.24 mg-P L−1 (22 lumped samples for each pond). For nitrogen a similar approach was chosen. Collins et al. (2010) report that for 3765 samples contained in the National Stormwater Quality Database, NSQD, the median of ammonia, nitrate and nitrite was 1.04 mg-N L−1 while the median for total nitrogen was 2.0 mgN L−1 . The fixed fraction between dissolved and total nitrogen was hence assumed to be 0.52. In the present study, the median of the total nitrogen entering the pond in Aarhus was 2.2 mg-N L−1 , while it was 3.0 mg-N L−1 for the pond in Odense. The outflow was simulated as a short pipe connected to a water brake. In other words, up to a certain cut-off level, the outflow increased as a constant multiplied by the square root of the water depth. An overflow became active at high water levels in the pond. The outflow carried phosphorus, nitrogen, suspended algae, DO and DIC out of the pond, corresponding to the concentrations in the pond at that time step. Energy entered the model as radiation through the surface of the ponds. The input radiation was determined by the measured time series of meteorological data. Absorption and dispersion causes radiation to decline through the water column. The radiation penetrating the water column (I), or better, the light intensity at a given depth, was in the model described by Eq. (1) (Jørgensen and Bendoricchio, 2001). I = I0 e−(˛+ˇXPP )z

(1)

The parameters in this and the following equations are explained in the nomenclature. O2 and CO2 entered and left the modeled pond over the water surface due to gas/liquid mass transfer (reaeration), where the driving force was the difference between the saturation concentration and the actual (modeled) concentration. However, the gas/liquid mass transfer is also highly influenced by wind, and several equations describing this interaction have been proposed (Ro and Hunt, 2006; Ro et al., 2007; Wanninkhof, 1992). In the present study, Eq. (2), suggested by Ro and Hunt (2006) was applied. Ro and Hunt (2006) determined the constant  KL to 170.6 and the constant KL to 1.81. This equation was chosen over others due to a possible high influence of the wind speed, as high influence of wind speed on DO and pH was suspected in the measured data series. −0.5 KL = KL ScO U KL /DIC 10 2

  0.5 a

W

T −20 wg

(2)

2.3.2. Primary producers Primary producers were subdivided into three groups: PP (phytoplankton), MP (rooted macrophytes) and BA (benthic algae including epiphytes). The different behavior of these groups was modeled by different growth, decay and respirations kinetics. Furthermore, macrophytes were modeled as capable of taking up a fixed ratio of their nutrients directly from the sediment. The two other groups of primary producers were modeled to assimilate nutrients from the water phase only. Additionally, only phytoplankton was washed out from the pond by the discharged water. PP and MP performed photosynthesis in the entire water column, whereas the BA only existed in the bottom layer. Growth of primary producers and thereby biomass performing photosynthesis was described by Eq. (3), where X and max

605

represent the parameters for BA, MP and PP, respectively. max ×

I I + KI



+

(1 − )SN SN + KN



SDIC X T −20 pHinhib SDIC + KDIC grd

+

(1 − )SP SP + KP



(3)

It is well known that within these three groups of primary producers, maximum growth rates, half saturation constants, light dependencies, and nutrient requirements are highly diverse (Jørgensen et al., 1991; Carpenter and Lodge, 1986; Sand-Jensen and Borum, 1991; Schladow and Hamilton, 1997). However, as the development and abundance of different species in the ponds had not been studied, only these three groups were chosen and no consideration was given to individual species. The respiration and decay of the primary producers was described by Eqs. (4) and (5), respectively. X, kR and bD represent the parameters for BA, MP and PP, respectively. T −20 kR grd XpHinhib

(4)

T −20 X bD grd

(5)

2.3.3. Dissolved oxygen Oxygen was produced by photosynthesis (3) and used by respiration (4) and by mineralization of the sediment. The latter process, the sediment oxygen demand (SOD) was described by Eq. (6). The flux of oxygen into the sediment was assumed instantaneously, that is, no diffuse boundary layer and sediment diffusion processes were explicitly included in the model. T −20 RSOD SOD

SO2

Apb

SO2 + KO2 Vp

pHinhib

(6)

Moreover, oxygen enters or leaves the system by gas–liquid mass transfer (reaeration) across the pond surface, Eq. (7). KL,O2 a(SOS − SO2 )

(7)

2.3.4. Inorganic carbon Inorganic carbon was assimilated by photosynthesis (3) and produced by respiration (4) and by mineralization of sediment (6). Furthermore, CO2 was exchanged over the pond surface by gas–liquid mass transfer, Eq. (8). KL,CO2 a(SCS − SCO2 )

(8)

2.3.5. pH The pH of the water was calculated from the DIC assuming that only carbonate species affect the total alkalinity of the water. This assumption is justified when other species that may affect the alkalinity are present in low concentrations compared to DIC. In the ponds, pH was almost always above 7. Assuming CO2 -equilibrium between water and atmosphere at a pH of 7 yields a corresponding DIC concentration of 67 ␮mol L−1 . A main candidate for another compound affecting the alkalinity in stormwater ponds is phosphate. In the water phase of the ponds phosphate was typically below 0.1 mg L−1 (3 ␮mol L−1 ), i.e. the concentration of DIC at equilibrium was typically significantly higher than the concentration of phosphate. Similar conditions exist for other compounds such as ammonia, or organic acids, which also could potentially affect the alkalinity.

606

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Table 2 Matrix formulation of the stormwater pond eutrophication model. XPP Growth of PP Respiration of PP Decay of PP Growth of BA Respiration of BA Decay of BA Growth of MP Respiration of MP Decay of MP Sediment oxygen demand Gas–liquid mass transfer, O2 Gas–liquid mass transfer, CO2

XBA

XMP

1 −1 −1 1 −1 −1 1 −1 −1

2.3.7. Process matrix A process matrix is a systematic arrangement of stoichiometry and process kinetics, which is routinely applied when communicating models of activated sludge, anaerobic sludge digesters, and transformations in sewer networks (e.g. Henze et al., 1987; Hvitved-Jacobsen, 2001; Nielsen et al., 2006). This arrangement is adapted to the present study, as it gives a ready and clear overview of processes and compounds included in the model, Table 2. In the matrix, each row represents a process (e.g. growth of primary producers) and each column represents a model component (e.g. phytoplankton biomass (XPP )). The matrix of processes and components is multiplied by the process rates described in the previous text and stated in the column furthest to the right. 2.4. Calibration The model was calibrated applying an automated calibration algorithm varying 16 of the model parameters one at a time. The parameters were chosen as those being the most sensitive to the simulation result. Each parameter was varied systematically until the absolute difference between modeled and measured oxygen and modeled and measured pH was minimized. Oxygen and pH were given equal weight in the fitting algorithm. Upon varying all parameters, the procedure was repeated until all parameters had reached a stable value. The initial value of each parameter was chosen within the ranges found in literature (Table 3). 2.5. Sensitivity analysis A sensitivity analysis was carried out to evaluate the response of the model to input parameter uncertainty. Several methods for sensitivity analysis exist (e.g. Jørgensen and Bendoricchio, 2001; Spitz and Moreno, 1996; Brito and Newton, 2011). In the present study the method presented by Brito and Newton (2011) was applied. In this method, each parameter, one at a time, was changed by ±30% with base in its value obtained by the calibration. For rainwater pH it was the H+ concentration value that was changed ±30%. Sensitivity indexes were calculated by applying Eq. (9). | hi |/hi | P|/P

SPsol

SNsol

SDIC

Process rate

YO2 :C −YO2 :C

−YP:C YP:C

−YN:C YN:C

−1 1

YO2 :C −YO2 :C

−YP:C YP:C

−YN:C YN:C

−1 1

YO2 :C −YO2 :C

−YP:C YP:C

−YN:C YN:C

−1 1

−1 +1

YP:O2

YN:O2

YC:O2

Eq. (3) Eq. (4) Eq. (5) Eq. (3) Eq. (4) Eq. (5) Eq. (3) Eq. (4) Eq. (5) Eq. (6) Eq. (7) Eq. (8)

+1

2.3.6. Nutrients Phosphorus and nitrogen were assimilated during photosynthesis and produced during respiration and mineralization of the sediment. The assimilation and production were modeled as a fixed ratio to the assimilation and production of inorganic carbon, respectively.

Si =

SO2

(9)

For each state variable and for each parameter, the median of Si has been calculated over a number of days of model output, which was then used to describe the sensitivity of a model parameter. 3. Result and discussion The data obtained by the on-line sensors were of good quality, and gave reliable readings. It revealed much information about the dynamics in the stormwater ponds, both regarding long-term and short-term variability. Figs. 4 and 5 show the variation of pH, DO, and temperature, as well as the accumulated inflow, from May 1 to December 31, 2008. The gray lines in the graphs show the short-term variation from minute to minute, while the black lines show the 2-day running average. During the period covered in Figs. 4 and 5, pond Aarhus received a total inflow of 47,871 m3 , corresponding to a mean hydraulic residence time in the pond of 35 days. The accumulated nutrient load was 87 kg total-N and 13.5 kg total-P. Pond Odense received 49,207 m3 , corresponding to a mean hydraulic residence time of 9.9 days. The accumulated nutrient load was 140 kg total-N and 20.5 kg total-P. Comparing the actual received water volumes to the design values (Table 1), it is seen that pond Aarhus received less runoff than expected in the design, while pond Odense received more than it was designed for. This resulted in the significant difference in hydraulic retention times of the two ponds. Furthermore, pond Aarhus received slightly less nutrients compared to pond Odense. Long-term as well as short-term variation of DO and pH were seen to be more significant in pond Odense than in pond Aarhus. For pond Aarhus, the average DO concentration was below oxygen saturation most of the time, indicating that the oxygen consuming processes tended to exceed the oxygen production by photosynthesis. The period of zero DO in the beginning of June might have been caused by an illicit discharge, as this occurred during and after a documented discharge of 1600 m3 of wastewater. For pond Odense, the oxygen produced by photosynthesis during the growth season tended to exceed the oxygen consuming processes. At the same time, the short-term fluctuations in pH and DO were more pronounced for pond Odense than for pond Aarhus. A reason for this difference could be the significant difference in hydraulic residence time of the two ponds (35 days versus 9.9 days) in combination with pond Odense receiving higher nutrient loads. Another factor could be the illicit discharges of sanitary wastewater to pond Aarhus and of industrial wastewater to pond Odense. For pond Odense, the average pH tended to follow the trend of the average DO concentration: when DO was high, so was pH. To a lesser degree, this was also the case for pond Aarhus. Zooming in on a few days shows that on this time scale DO and pH were clearly coupled (Fig. 6). The reason is believed to be photosynthesis producing oxygen during the day and respiration consuming oxygen during the night – over-layered by a continuous

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

607

Table 3 Literature values and ranges of input parameters. Parameters obtained by calibration for the two ponds. Note that parameter units might differ from the nomenclature. Parameter

Unit

bD,BA bD,MP bD,PP DICrunoff KDIC KI KN KO2 KP kR,BA kR,MP kR,PP pHoptimum pHrunoff RSOD YN:C YN:O2 YO2 :C YP:C YP:O2 ˛ ˇ   KL KL  grd  SOD  wg max,BA max,MP max,PP

d−1 d−1 d−1 M gC m−3 J m−2 s−1 mgN m−3 gO2 m−3 mgP m−3 d−1 d−1 d−1 – – gO2 m−2 d−1 gN gC−1 gN gO2 −1 gO2 gC−1 gP gC−1 gP gO2 −1 m−1 m2 g−1 – – – – –

a b c d e f g h i j k l m n o p q r s t

d−1 d−1 d−1

Literature values and ranges

Applied value

0.001c 0.001–0.125e , 0.88i 0.83–5.3g , 0.36–9h 4.8–18.1b , 7.2d , 6.7f , 35.7f , 35.4l , 2.2l 25d , 1.4–980c , 0–500e , 47l 1.4a , 5.9a , 0.5b , 0.2–1500c , 3d , 1–30e , 11l 0.018c 0.03–0.92c , 0.001–0.171e , 0.08–0.1m 6.31–6.84n , <6.5o , 8p 7.37–8.07q , 4.5–10.1r 0.02–50c 0.25d , 0.18k , 0.15l 2.67–4.21k , 3.1c 0.025d , 0.008j , 0.019j 0.024k 0.2d , 0.03–15.1e

170.6t 1.81t 1.02–1.14c , 1.066m 1.045c 1.024s 0.08–0.2c 0.5d , 0.1–11b , 1.3–3.63e , 2.90f , 0.1–5.65m

Model initiation

Pond Aarhus

Pond Odense

0.001 0.01 0.01 0.005 1 2 25 0.5 4 0.2 0.018 0.2 7.5 7.5 1.5 0.18 0.08 2.67 0.015 0.005 0.4 0.3 0.5 170.6 1.81 1.07 1.05 1.024 2 0.15 2

<0.001 <0.001 0.218 0.007

<0.001 <0.001 0.039 0.019

<0.001 0.103 1.150

0.010 0.032 0.308

7.30 6.02

6.10 0.74

33.8 1.17

26.8 1.06

<0.01 0.254 2.62

<0.01 0.250 3.01

Holm and Armstrong (1981). Auer and Forrer (1998). Asaeda and Van Bon (1997) and references therein. Mulderij et al. (2007) and references therein. Hamilton and Schladow (1997) and references therein. Storey et al. (1993). Caperon and Smith (1978). Clark and Flynn (2000). Brun et al. (2001). Thingstad et al. (1996). Fraga et al. (1998). Jørgensen et al. (1991) (average of relevant values). Jørgensen et al. (1991). Mayo (1997). Necchi and Zucchi (2001). Graham et al. (1996). Dierberg et al. (2002). Kayhanian et al. (2007). Hvitved-Jacobsen (2001). Ro and Hunt (2006).

bacterial respiration in the pond sediments. At the same time the photosynthesis consumes carbonate, which raises the pH, and respiration produces carbonate, which decreases the pH. The periods selected in Fig. 6 also show the influence of stormwater runoff entering the ponds. In both cases the runoff caused decreases in DO and pH. Temperature also decreased but this might well have been an effect of a general cooling of the air. At pond Aarhus the average air temperature dropped from 21 to 16 ◦ C, while it at pond Odense dropped from 16 to 14 ◦ C. For pond Aarhus the runoff events of August 2–5 conveyed 2750 m3 of runoff, corresponding to 40% of the volume of the permanent wet pool. In addition to the general reduction in DO and pH, there was a reduction of amplitude of the diurnal variation of these parameters upon that runoff event. Whether this was due to flushing out of algae, increased turbidity resulting in less light penetrating into the water, or clouds allowing less solar radiation is not quite clear. The

turbidity meter was unfortunately inoperative during this period so this parameter could not be evaluated. The solar radiation of August 4 was around 20% of what it was on the 3 preceding days. On the following days the solar radiation again increased somewhat, but only to about half of what it was on August 1–3. On August 8–9 the solar radiation was back to the same level as before the event. With exception of the DO and pH increase on August 6, this corresponds well with what could be expected due to variations in photosynthesis caused by differences in solar radiation. The runoff event around midnight between August 7 and 8 was of nearly the same volume as the preceding event, however, it did not affect pH nor DO in the pond. This leads to the conclusion that wash-out of algae probably played a minor role in pond Aarhus, and that solar radiation was the most significant parameter. Prior to the runoff in the afternoon of September 2 the pond Odense, was super-saturated with oxygen to an average level of

608

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Fig. 4. Dissolved oxygen, pH, temperature, and accumulated inflow to pond Aarhus.

nearly 200% and had a pH above 9 (Fig. 6). It took 2–3 days for the DO to rise from 70% of saturation measured, and the pH from 7 to 9. At the same time, the solar radiation had increased continuously to a comparatively high level. On September 1 the solar radiation dropped to about half of the preceding day, and a slight reduction in DO and pH were seen. During the first part of the following day, DO and pH were further slightly reduced. In the afternoon a runoff event caused some 900 m3 of stormwater runoff to enter the pond, corresponding to 45% of its permanent wet pool. This caused DO and pH to drop rapidly. The runoff event could in principle cause a significant fraction of the suspended algae to be washed out. The turbidity as measured by the turbidity meter was, however, not significantly affected by the inflow. This might, on the other hand, have been due to a combination of increased turbidity from runoff particles and a decreased turbidity from less suspended algae. After the event, the solar radiation stays more or less constant until September 5, where there is some increase. During this period the DO and pH increase slightly. Even though wash-out of algae may have occurred, solar radiation was therefore still playing a major role. For both events the availability of nutrients after the events might have been higher than before the events as the stormwater runoff contains significant amounts of both phosphorous and nitrogen. However, the measured data series do not indicate such to be the case.

3.1. Calibration For each pond, the model was calibrated to the 9 days of measurement shown in Fig. 6. The values obtained by the calibration

as well as literature values and ranges for other input variables are shown in Table 3. Several publications address algae and plant dynamics in eutrophic lakes and reveal mechanisms describing interactions between such groups of primary producers (Sand-Jensen and Borum, 1991; Carpenter and Lodge, 1986; Barko and Smart, 1981; Madsen and Cedergreen, 2002; Xie et al., 2005). Based on this literature it was chosen to assign the same initial values to parameters quantifying respiration and decay of BA and of PP. The growth rate of BA was chosen slightly lower than that of PP. For MP, the ratio between uptake of phosphorus and nitrogen from the water column and from the sediment depends both on plant species, on concentration of nutrients in the water column and on the age of the plants. For the simulations it was assumed that MP took up 50% of their phosphorus and 30% of their nitrogen from the sediment (Mulderij et al., 2007), which would give them a significant advantage when nutrient limitations occurred in the water column. It was possible to achieve a reasonable calibration of the model to the two 9-day intervals (Figs. 7 and 8). However, the calibration to pond Aarhus was better than the calibration to pond Odense, which most like was due to the a lesser variability in DO and pH in pond Aarhus compared to pond Odense. A major difference in the parameter values obtained by the calibration was the sediment oxygen demand (RSOD ). In comparison did Belanger (1980) for a hyper-eutrophic, shallow lake report an average sediment oxygen uptake rate of 1.6 gO2 m−2 d−1 . For sewer sediments at DO concentrations of 6–8 gO2 m−3 , Vollertsen and Hvitved-Jacobsen (2000) found rates in the range of 1–8 gO2 m−2 d−1 . The more degraded the sediments were, the lower the rates became. In light of those studies, the high sediment oxygen demand of pond Aarhus was probably caused by the wastewater being illicitly discharged to this

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

609

Fig. 5. Dissolved oxygen, pH, temperature, and accumulated inflow to pond Odense.

pond, increasing the amount of biodegradable organic matter in the pond sediments. For both calibrations the water–gas mass transfer due to wind, as governed by the model parameters  KL and KL , were found to be significantly less than what was suggested by Ro and Hunt (2006). The parameters were lower for pond Odense than for pond Aarhus, which could be caused by pond Odense being located at a more sheltered location than pond Aarhus. For both ponds the calibration tended to suppress the activity (growth rate) of BA in favor of PP and MP, resulting in BA becoming nearly absent. The growth rates for the three groups of primary producers were approximately the same for the two calibrations, however, the respiration and decay rates for PP and MP in pond Aarhus were lower than for pond Odense. The effect here of is seen in Figs. 9 and 10 where the development of BA, PP, and MP is shown from May to December. 3.2. Validation Figs. 7 and 8 show the calibration to the 9-day periods and the detailed validation on the 26 following days. For both validations the trends of DO and pH were reasonably well simulated, even though pond Aarhus showed a better agreement of the absolute values. Even though the validation of pond Odense tended to overshoot or undershoot both DO and pH, the overall trend of an increased pH and DO around September 12–15, the trend toward a drop in both parameters around September 18–24, and the tendency to subsequent increase and following decrease were captured by the model. The diurnal variations were also simulated reasonably well, even though the simulated amplitude of the diurnal DO variation was somewhat too large. The validation of pond Aarhus showed good agreement between measured and simulated

DO and pH. Diurnal variations were generally well simulated, even though increased levels of DO and pH around August 16–20, around August 22–27, and around August 31 were somewhat underestimated. Figs. 9 and 10 show the simulation of the whole period in terms of 2-day running averages and compare those to the 2-day running averages of measured DO and pH. The simulation of pond Aarhus prior to the calibration is poor. However, this is probably partly due to the large illicit discharge of wastewater to the pond of in the beginning of June. Other illicit discharges may also have occurred in this period, obscuring the validity of the simulation result. From October and on, both DO and pH are overestimated, probably due to an underestimation of the sediment oxygen uptake rate. In terms of reproducing the large fluctuations in DO and pH, the long-term simulation of pond Odense was somewhat better. Many of the decreases and increases of DO and pH were reproduced satisfactory, and the general trend to reproduce periods with high and low values could be simulated. Comparing to the simulated amount of primary producers, it becomes clear that the capability of the model to simulate these variations is strongly linked with its capability of reproducing the amount of PP present. In pond Odense, PP was dominating until October, after which MP took over, while MP was dominating most of the time in pond Aarhus. This observation was supported by measurements of chlorophyll from start May to end August the following year in pond Odense. 11 samples equally distributed over that period showed an average chlorophyll concentration of 288 ± 136 ␮g L−1 . In 2009, chlorophyll was also measured in pond Aarhus. However, as iron had been added to enhance treatment, these data were not usable in the present context. The simulated PP biomass in pond Odense was in the range from 10 to 30 gC m−3 (Fig. 10).

610

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Fig. 6. Dissolved oxygen, pH, temperature, solar radiation, and accumulated inflow for 9 consecutive days for both ponds.

Assuming a carbon to chlorophyll ratio of 50 mgC mgChl−1 (Jørgensen et al., 1991), the simulated PP corresponded to a chlorophyll concentration of 200–600 ␮g L−3 , i.e. in good agreement with the measurements made the following summer. Even though these simulated and measured chlorophyll concentrations correspond to two different growth seasons, the agreement in magnitude is deemed to support the simulation results. The model tends to show high concentrations of MP in the early winter – i.e. it does not sufficiently account for the die-back of the macrophytes.

calculated on simulation of the period from August 1 to September 10. Fig. 11 shows the 12 most influential parameters determined by the analysis. The most influential parameters identified were the growth and respiration rates. For the pH simulation, the pH of the runoff was also rather influential. Interpreting the results of the analysis it is, however, important to note that model parameters are generally not independent, and that some parameters, for example those related to BA, were suppressed by the model and for that reason showed up as being insensitive.

3.3. Sensitivity analysis

3.4. Application of the model

A sensitivity analysis was carried out for all parameters included in the model. The parameter set from the calibration of pond Aarhus was used for this analysis, and it was carried out with respect to both DO and pH. The model sensitivity parameter Si (Eq. (9)) was

Regardless of a high nutrient load on the ponds, no significant oxygen depletion or harmful high pH caused by eutrophication were observed. That is, assuming that the periods of oxygen depletion in pond Aarhus were caused by illicit wastewater

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Fig. 7. Calibration and validation of the eutrophication model to part of the data series of pond Aarhus.

Fig. 8. Calibration and validation of the eutrophication model to part of the data series of pond Aarhus.

Fig. 9. Simulated and measured levels of DO and pH in pond Aarhus.

611

612

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

Fig. 10. Simulated and measured levels of DO and pH in pond Odense.

Fig. 11. Sensitivity analysis based on dissolved oxygen and pH as output parameters. Parameters were changed ±30%. pH in runoff was changed ±1 unit.

discharges and not by primary producers, an assumption which is supported by the model results. The lack of oxygen depletion due to eutrophication could be explained by the, compared to natural lakes, high exchange rate of water in the stormwater ponds, causing the phytoplankton to be washed out before it caused significant oxygen depletion. This explanation was supported by the simulated algae biomass in the ponds. Here it was seen that the PP biomass decreased when stormwater runoff was introduced to the pond, i.e. when exchange of water took place (Figs. 9 and 10). For example did large amounts of runoff in Aarhus in primo August cause significant reduction in PP biomass within a few days. The PP biomass was not able to recover from this event. In pond Odense, a succession of runoff events reduced the PP concentrations, but here the PP biomass did recover, and first became insignificant during November. Besides explaining observations made on primary producers in stormwater ponds, the developed model has potential as a

planning and design tool for new stormwater ponds. Even though the model is not perfect, as it for example does not include illicit discharges such as were seen in both ponds, it is in principle capable of predicting how different load patterns and land uses affect the presence of primary producers. It is also capable of assessing the sensitivity of primary producer growth on pond layout. It can for example answer questions on how an increase or decrease in water depth will tend to affect DO and pH in the pond. The deeper the pond, the less the footprint and hence, the less the land cost of the pond. Today, stormwater ponds are seldom designed deeper than 1.5 m as engineers want to avoid low DO (HvitvedJacobsen et al., 2010). It is probably a rather conservative choice and water depths could in many situations be increased, leading to decreased investment costs. Similarly, other design characteristics could be assessed by applying the model, for example the effect of outlet flow restriction or the effect of compartmentalizing the pond.

T. Wium-Andersen et al. / Ecological Engineering 61P (2013) 601–613

4. Conclusion The eutrophication model could be successfully calibrated to two stormwater ponds. The model parameters obtained by the calibration differed somewhat, but most differences could be explained by one of the ponds receiving significant loads of illicitly discharged wastewater. The model showed that the intermittent flow pattern and comparative short hydraulic residence times strongly affected the presence of different groups of primary producers. For example did a washout of phytoplankton in one of the ponds allow a late autumn bloom of macrophytes. The model is deemed a potential tool for improving the design of stormwater ponds by taking the behavior of the plant ecosystem into account. It has the potential to yield information on parameters critical for design, for example levels of pH and dissolved oxygen. Also the prediction of phytoplankton is an issue as stormwater ponds not only serve as treatment facilities but also as recreational water bodies in the city. References Asaeda, T., Van Bon, T., 1997. Modelling the effects of macrophytes on algal blooming in eutrophic shallow lakes. Ecol. Modell. 104 (2-3), 261–287. Auer, M.T., Forrer, B., 1998. Development and parameterization of a kinetic framework for modelling light- and phosphorus-limited phytoplankton growth in Cannonsville reservoir. Lake Reserv. Manage. 14 (2–3), 290–300. Barko, J.W., Smart, R.M., 1981. Sediment-based nutrition of submersed macrophytes. Aquat. Bot. 10 (4), 339–352. Belanger, T.V., 1980. Benthic oxygen demand in Lake Apopka, Florida. Water Res. 15 (2), 267–274. Brito, A.C., Newton, A., 2011. The role of microphytobenthos on shallow coastal lagoons: a modelling approach. Biogeochemistry 106 (2), 207–228. Brun, R., Reichert, P., Künsch, H.R., 2001. Practical identifiability analysis of large environmental simulation models. Water Resour. Res. 37 (4), 1015– 1030. Caperon, J., Smith, D.F., 1978. Photosynthetic rates of marine algae as a function of inorganic carbon concentration. Limnol. Oceanogr. 23 (4), 704–708. Carpenter, S.R., Lodge, D.M., 1986. Effects of submersed macrophytes on ecosystems processes. Aquat. Bot. 26 (3-4), 341–370. Clark, D.R., Flynn, K.J., 2000. The relationship between the dissolved inorganic carbon concentration and growth rate in marine phytoplankton. Proc. Biol. Sci. 267 (1447), 953–959. Collins, K.A., Lawrence, T.J., Stander, E.K., Jontos, R.J., Kaushal, S.S., Newcomer, T.A., Grimm, N.B., Ekberg, M.L.C., 2010. Opportunities and challenges for managing nitrogen in urban stormwater: a review and synthesis. Ecol. Eng. 36 (11), 1507–1519, Danmarks miljoeportal http://arealinformation.miljoeportal.dk/distribution/modified July 1 2013. Dierberg, F.E., DeBusk, T.A., Jackson, S.D., Chimney, M.J., Pietro, K., 2002. Submerged aquatic vegetation-based treatment wetlands for removing phosphorus from agricultural runoff: response to hydraulic and nutrient loading. Water Res. 36 (6), 1409–1422. Fraga, F., Ríos, A.F., Péres, F.F., Figueiras, F.G., 1998. Theoretical limits of oxygen:carbon and oxygen:nitrogen ratios during photosynthesis and mineralisation of organic matter in the sea. Sci. Mar. 62 (1-2), 161–168. Gelbrecht, J., Lengsfeld, H., Pöthig, R., Opitz, D., 2005. Temporal and spatial variation of phosphorus input, retention and loss in a small catchment of NE Germany. J. Hydrol. 304 (1-4), 151–165. Graham, J.M., Arancibia-Avila, P., Graham, L.E., 1996. Effects of pH and selected metals on growth of the filamentous green alga Mougeotia under acidic conditions. Limnol. Oceanogr. 41 (2), 263–270. Hamilton, D.P., Schladow, S.G., 1997. Prediction of water quality in lakes and reservoirs. Part I. Model description. Ecol. Modell. 96 (1–3), 91–110. Henze, M., Grady Jr., C.P.L., Gujer, W., Marais, G.v.R., Matsuo, T., 1987. Activated Sludge Model No. 1. Scientific and Technical Report No. 1. International Association on Water Pollution Research and Control. Hipsey, M.R., Romeo, J.R., Antenucci, J.P., Hamilton, D., 2006. Computational Aquatic Ecosystem Dynamics Model: CAEDYM v2, v2.3 Science Manual. Contract Research Group, Centre for Water Research, University of Western Australia. Holm, N.P., Armstrong, D.E., 1981. Role of nutrient limitation and competition in controlling the populations of Asterionella formosa and Microcystis aeruginosa in semicontinuous culture. Limnol. Oceanogr. 26 (4), 622–634. Hvitved-Jacobsen, T., 2001. Sewer Processes – Microbial and Chemical Process Engineering of Sewer Networks. CRC Press. Hvitved-Jacobsen, T., Johansen, N.B., Yousef, Y.A., 1994. Treatment systems for urban and highway run-off in Denmark. Sci. Total Environ. 147, 499–506. Hvitved-Jacobsen, T., Vollertsen, J., Haaning-Nielsen, A., 2010. Urban and Highway Stormwater Pollution – Concepts and Engineering. CRC Press.

613

Isteniˇc, D., Arias, C.A., Vollertsen, J., Nielsen, A.H., Wium-Andersen, T., HvitvedJacobsen, T., Brix, H., 2012. Improved urban stormwater treatment and pollutant removal pathways in amended wet detention ponds. J. Environ. Sci. Health A 47, 1446–1477. Jørgensen, S.E., Bendoricchio, G., 2001. Fundamentals of Ecological Modeling, 3rd ed. Elsevier Science. Jørgensen, S.E., Nielsen, S.N., Jørgensen, L.A., 1991. Handbook of Ecological Parameters and Ecotoxicology. Elsevier Science Publishers. Kayhanian, M., Suverkropp, C., Ruby, A., Tsay, K., 2007. Characterization and prediction of highway runoff constituent event mean concentration. J. Environ. Manage. 85 (2), 279–295. Leisenring, M., Clary, J., Stephenson, J., Hobson, P., 2010. International stormwater best management practices (BMP) database pollutant category summary: nutrients. Report of the International Stormwater BMP Database., pp. 36, http://www.bmpdatabase.org Madsen, T.V., Cedergreen, N., 2002. Sources of nutrients to rooted submerged macrophytes growing in nutrient-rich stream. Freshw. Biol. 47 (2), 283–291. Madsen, H., Vollertsen, J., Hvitved-Jacobsen, T., 2007. Modelling the oxygen mass balance of wet detention ponds receiving highway runoff. Highway and Urban Environment, Bookseries: Alliance for Global Sustainability, vol. 12. Springer, The Netherlands, pp. 487497. Magaud, H., Migeon, B., Morfin, P., Garric, J., Vindimian, E., 1997. Modelling fish mortality due to urban storm run-off: interacting effects of hypoxia and unionized ammonia. Water Res. 31 (2), 211–218. Mayo, A.W., 1997. Effects of temperature and pH on the kinetic growth of unialgal Chlorella vulgaris cultures containing bacteria. Water Environ. Res. 69 (1), 64–72. Mehta, S.K., Gaur, J.P., 2005. Use of algae for removing heavy metal ions from wastewater: progress and prospects. Crit. Rev. Biotechnol. 25 (3), 113–152. Mitsch, W.J., Reeder, B.C., 1991. Modelling nutrient retention of a fresh-water coastal wetland: estimating the roles of primary productivity, sedimentation, resuspension and hydrology. Ecol. Modell. 54 (3-4), 151–187. Mulderij, G., Van Nes, E.H., Van Donk, E., 2007. Macrophyte–phytoplankton interactions: the relative importance of allelopathy versus other factors. Ecol. Modell. 204 (1-2), 85–92. Necchi, O., Zucchi, M.R., 2001. Photosynthetic performance of freshwater Rhodophyta in response to temperature, irradiance, pH and diurnal rhythm. Phycol. Res. 49, 305–318. Nielsen, A.H., Vollertsen, J., Hvitved-Jacobsen, T., 2006. Kinetics and stoichiometry of aerobic sulfide oxidation in wastewater from sewers – effects of pH and temperature. Water Environ. Res. 78 (3), 275–283. NURP, 1983. Final Report of the Nationwide Urban Runoff Program – Volume 1. Water Planning Division, U.S. Environmental Protection Agency, Washington, DC 20460. Pettersson, T.J.R., German, J., Svensson, G., 1999. Pollutant removal efficiency in two stormwater ponds in Sweden. In: Joliffe, I.B., Ball, J.E. (Eds.), Proceedings of the 8th International Conference on Urban Storm Drainage. Sydney, Australia, August 30–September 3, pp. 866–873. Ro, K.S., Hunt, P.G., 2006. A new unified equation for wind-driven surficial oxygen transfer into stationary water bodies. Trans. ASABE 49 (5), 1615–1622. Ro, K.S., Hunt, P.G., Poach, M.E., 2007. Wind-driven surficial oxygen transfer. Crit. Rev. Environ. Sci. Technol. 37 (6), 539–563. Sand-Jensen, K., Borum, J., 1991. Interactions among phytoplankton, periphyton, and macrophytes in temperate freshwaters and estuaries. Aquat. Bot. 41 (1-3), 137–175. Schladow, S.G., Hamilton, D.P., 1997. Prediction of water quality in lakes and reservoirs. Part II. Model calibration, sensitivity analysis and application. Ecol. Modell. 96 (1-3), 111–123. Spitz, K., Moreno, J., 1996. A Practical Guide to Groundwater and Solute Transport Modelling. John Wiley & Sons, Inc. Storey, M.L., Auer, M.T., Barth, A.K., Graham, J.M., 1993. Site-specific determination of kinetic coefficients for modelling algal growth. Ecol. Modell. 66 (3-4), 181–196. Thingstad, T.F., Riemann, B., Havskum, H., Garde, K., 1996. Incorporation rates and biomass content of C and P in phytoplankton and bacteria in the bay of Aarhus (Denmark) June 1992. J. Plankton Res. 18 (1), 97–121. Torno, H.C., Marsalek, J., Desbordes, M., 1985. Urban Runoff Pollution. Series G: Ecological Sciences, vol. 10. Springer. Vollertsen, J., Hvitved-Jacobsen, T., 2000. Resuspension and oxygen uptake of sediments in combined sewers? Urban Water 2 (1), 21–27. Vollertsen, J., Åstebøl, S.O., Coward, J.E., Fageraas, T., Madsen, H.I., Nielsen, A.H., et al., 2007. Monitoring and modeling the performance of a wet pond for treatment of highway runoff in cold climates. In: Highway and Urban Environment, Bookseries: Alliance for Global Sustainability. Springer, The Netherlands, pp. 499–509. Wanninkhof, R., 1992. Relationship between wind speed and gas exchange over the ocean. J. Geophys. Res. 97 (C5), 7373–7382. Wium-Andersen, T., Nielsen, A.H., Hvitved-Jacobsen, T., Vollertsen, J., 2012. Modeling nutrient and pollutant removal in three wet detention ponds. Urban Environment: Alliance for Global Sustainability Series, vol. 19. Springer. Wong, J.W.C., Yang, C.L., 1997. The effect of ph and redox potential on the release of nutrients and heavy metals from a contaminated marine sediment. Toxicol. Environ. Chem. 62 (1), 1–10. Xie, Y., An, S., Yao, X., Xiao, K., Zhang, C., 2005. Short-time response in root morphology of Vallisneria natans to sediment type and water-column nutrient. Aquat. Bot. 81 (1), 85–96.