Urban stormwater is regarded as a key input of faecal contamination in receiving water bodies and therefore, a major concern for health risks associated with aquatic recreation. Wastewater leakages, cross connections and overflows, together with faeces washed from surfaces during rainfall events, are possible origins of faecal contamination which enter these water bodies through stormwater drains. This paper applies conceptual models to a case study of the Yarra River estuary to understand the relative importance of fluxes derived from an urban creek and the 219 urban stormwater pipes which drain directly to the estuary as compared with other inputs, such as the Yarra River itself. Existing hydrologic-microorganism models were used for the estimation of the inputs from riverine and urban stormwater fluxes. These predictions were applied as boundary conditions for a new, highly simplified, model which accounts for the transport and survival of faecal microorganisms in the estuary. All models were calibrated using a rich dataset, containing over 2,000 measured Escherichia coli concentrations. Mass balances from the riverine and stormwater models indicate the limited influence of urban stormwater drains on the estuary during dry weather; less than 0.05% to 10% (5th and 95th percentile; median 0.5%) of the total daily E. coli load entering the estuary was derived from urban stormwater drains. While wet weather contributions from stormwater drains could be more significant (2% to 50%; 5th and 95th percentile), the average contribution remained marginal (median 10%). Sensitivity testing of the estuarine microorganism model by switching off stormwater boundary conditions resulted in minimal model efficiency reduction; this may reflect the low average daily contribution from urban stormwater drains. While these results confirm previous studies which show that E. coli loads derived from stormwater drains are dwarfed by other inputs, it is essential to note that these results also demonstrate that some conditions reveal the opposite; high proportions from stormwater are possible when combined with low riverine inputs and high urban rainfall. Furthermore, this study focuses on the overall impacts of direct urban stormwater inputs on the faecal contamination levels within the estuary, and localized impacts would certainly require further investigation.

INTRODUCTION

Urban estuaries around the world are highly valued assets to the local community, as they provide aesthetics, improved microclimate and recreational opportunities (Mallin et al. 2000). Like many other urban estuaries, the Yarra River estuary has elevated levels of faecal contamination (Daly et al. 2013), which is of public health concern for recreational users. Faecal microorganisms have been identified as the leading cause of pollution of environmental waters (Lipp et al. 2001; Burton & Pitt 2002; Ortega et al. 2009).

Urban stormwater has been recognized as an important input of faecal contamination to these waterways (Burton & Pitt 2002; McCarthy et al. 2011). As such, increased efforts have been made towards mitigating the impacts of direct stormwater inputs (i.e. the stormwater drains that discharge directly into the estuary), including the Yarra River estuary (e.g. Melbourne Water 2013). However, despite these efforts, minimal improvement in compliance figures was observed for this particular system, implying that there may be other, more significant, inputs which require mitigation.

Effective management of faecal contamination in urban estuaries requires a firm understanding of the inputs of pollution and its transport and fate within the system. Integrated modelling tools which fully account for both the input and the estuarine microorganism dynamics are absent from the literature. Indeed, most models found in the literature poorly represent the microbial dynamics, and instead (1) are calibrated and tested on a small number of measured data points (Kashefipour et al. 2002; Gao et al. 2011; de Brauwere et al. 2011), (2) represent inputs using a constant flux value, (3) predict inputs using simple relationships with flow (Garcia-Armisen et al. 2006; Liu & Huang 2012), and/or (4) predict inputs using a sediment–microbe correlation (Ghimire & Deng 2013). Use of these approaches might mask the true importance of a particular input, which in turn can significantly influence the results of estuarine microorganism model and misinform mitigation.

The aim of this study was to create an integrated conceptual-level Escherichia coli model for an estuarine catchment; we did this by linking models which already exist for riverine E. coli prediction (Haydon & Deletic 2006) and stormwater E. coli prediction (McCarthy et al. 2011) to a newly developed estuarine microorganism model. This integrated model was then used to assess the importance of the various inputs into the estuary. Of particular importance was whether the stormwater flow from the urban creek and the 219 urban stormwater drains, which directly enter the estuary, are a significant input of E. coli. In addition, there are other stormwater inputs entering the estuary indirectly through upstream river inflow. These are not assessed separately but are considered as part of riverine input. The models were calibrated on an extensive dataset, containing over 2,000 samples analysed for the most commonly used faecal indicator, E. coli. The major hypothesis of this work was that the importance of direct urban stormwater was minimal during dry weather periods, but increased during urban wet weather periods, especially when lower riverine flow rates were combined with higher amounts of urban rainfall. The impact of direct wet weather stormwater inputs could be important even in the case of uniformly distributed rainfall across a whole catchment, as stormwater could be entering the estuary much sooner than the riverine input due to the higher imperviousness and shorter time of concentration that characterize urbanized areas.

METHODS

The estuary and monitoring sites

The Yarra River estuary (Melbourne, Victoria, Australia) is a highly stratified, salt-wedge estuary (Beckett et al. 1982) and extends for about 22 km from Port Philip Bay to Dights Falls – a weir which represents the upper boundary of the estuary. Monitoring sites were selected and established for data collection (Figure 1). Two of the sites were within the estuary: Abbotsford at the very beginning of the estuarine section of the Yarra River (represents the region with little influence from the salt-wedge, but still impacted by tidal changes) and Morell Bridge, located in the lower part of the estuary (highly impacted by the salt-wedge). Both sites were equipped with refrigerated automated samplers and depth sensors and had continuous measurements of electrical conductivity (EC) and temperature (T) at 100 mm below the surface. The Morell Bridge site was also equipped with an Acoustic Doppler Current Profiler (ADCP) for 3D measurements of velocities at 1-minute intervals.

Figure 1

Monitoring stations in the Yarra River catchment (stations: Heidelberg and Coldstream (rain data) and Viewbank and Melbourne airport (climate data) are positioned outside the figure boundary). Shaded area represents the urban estuary catchment with the biggest 20 of the 216 modelled drains shown.

Figure 1

Monitoring stations in the Yarra River catchment (stations: Heidelberg and Coldstream (rain data) and Viewbank and Melbourne airport (climate data) are positioned outside the figure boundary). Shaded area represents the urban estuary catchment with the biggest 20 of the 216 modelled drains shown.

Monitoring of upstream river inputs was conducted at Kew (Figure 1), where only grab samples were taken and water levels and flow rates were measured at 6-minute intervals by Melbourne Water (the local water management authority).

Monitoring of stormwater inputs was done at Gardiners Creek, a heavily channelized creek which is the largest input of water other than the Yarra River upstream of Dights Falls. The site has been equipped with an automated sampler, EC/T sensors and a depth/velocity probe. Climate data were obtained from Australian Bureau of Meteorology and Melbourne Water for different locations in the Yarra River catchment (Figure 1). Gardiners Creek is considered to be an open channel stormwater drain because its catchment is completely developed with total impervious fraction of 47%. Furthermore, observed range of the E. coli concentrations (944; 6,203; 17,673 most probable number (MPN)/100mL; 5th, 50th, 95th percentile) is well within the range reported for urban stormwater (Makepeace et al. 1995; Burton & Pitt 2002).

Sample collection and analysis

Estuarine and riverine samples were taken approximately 100 mm below the surface, where the health exposure to recreational users is expected to be the highest. In the period from November 2012 to July 2013, 2,106 samples were collected; 1,500 during dry weather and 606 during wet weather conditions. All collected samples were transported to the Environmental and Public Health Microbiology (EPHM) laboratory at Monash University in coolers on ice and analysed for E. coli content using the Colilert method (IDEXX Laboratories 2013) within 24 h of collection. A large range of other indicators and reference pathogens were tested, but not reported here.

Riverine model

Hydrology of the upper Yarra River catchment (river inflow at Dights Falls into the estuary, Figure 1) was modelled using MUSIC – SimHyd, which is a spatially lumped catchment rain-runoff model (eWater 2012). The model was applied with some slight variations: (1) a linear-reservoir routing routine was employed (instead of MUSIC's standard Muskingum Cunge method) as it has been demonstrated previously that this simpler and more stable form of routing produces equivalent results (McCarthy 2008); (2) the model was employed using a constant 6-minute timestep (as opposed to MUSIC's standard method of daily simulation and subsequent disaggregation). This method improved the computational efficiency of the model, without compromising the results. Model inputs were areal averaged rainfall (Heidelberg, Kew, Kew Reservoir, Coldstream and Viewbank stations) and daily potential evapotranspiration, calculated using the Food and Agriculture Organization Penman-Monteith method (data from Coldstream, Viewbank and Heidelberg stations). The MUSIC–SimHyd model was calibrated with a Monte-Carlo approach using a least squares objective function comparing the predicted flow rates with untransformed measured flow rates at Kew. The performance of the hydrologic model was assessed using the Nash-Sutcliffe coefficient of efficiency (Nash & Sutcliffe 1970). Parameter sensitivity was also explored using the Monte-Carlo results, as per others in the literature (e.g. Dotto et al. 2010).

For the prediction of riverine microbial concentrations, a modified version of the EG pathogen-hydrologic catchment model (Haydon & Deletic 2006) was applied. The main variation was that the loss of microorganisms from the subsurface store was estimated to be inversely proportional to the soil moisture instead of directly proportional, which was originally proposed by Haydon & Deletic (2006), as many studies report extended survival of faecal microorganisms at higher soil moisture contents (Desmarais et al. 2002; Schäfer et al. 1998). The model had six parameters: one parameter described build-up, two were loss coefficients and three were related to wash-off processes. Inputs to the model were time series potential evapotranspiration and flow components as calculated by MUSIC – SimHyd. The model was calibrated against Abbotsford's E. coli concentration dataset. Although there are obvious issues with this methodology (i.e. calibrating the upstream model to a site within the estuary), it was considered adequate for the following reasons: (1) Daly et al. (2013) showed that Kew and Abbotsford have similar distributions; (2) the correlation between the E. coli from the two sites was 0.83 (Pearson correlation coefficient, p < 0.001); and (3) the Abbotsford dataset had many more calibration points (776 compared with 43 at Kew), which could allow a better calibrated model. The optimized parameter set for the EG model was obtained using a least squares objective function and by observing the Pareto front formed when calibrating using untransformed and log-transformed E. coli concentrations. Additional calibration of the model parameters was conducted using the Generalized Reduced Gradient method, without limiting the parameters and using a criterion which added the two components of the Pareto front. The model's performance was assessed by the Nash-Sutcliffe efficiency calculated using untransformed and log-transformed E. coli concentrations – and , respectively.

Stormwater model

Modelling of the urban stormwater input of Gardiners Creek was performed using Micro-Organism Prediction in Urban Stormwater, MOPUS (McCarthy et al. 2011), where the pervious component of the rain-runoff model was excluded. As shown previously by Dotto et al. (2011), the parameters which are used to model the pervious component are less sensitive than those used to model impervious areas, therefore demonstrating the importance of impervious areas in urbanized catchments. The rainfall-runoff module of MOPUS was calibrated against the untransformed flow rates measured at the Gardiners Creek monitoring station using the same procedure outlined above for the riverine model.

MOPUS's microorganism model has five model parameters: three which represent the build-up and die-off of microorganisms on the surface of the catchment, and two others which represent the same for the subsurface (i.e. in the stormwater drain). The inputs to the model include: time series of rainfall, relative humidity and vapour pressure. MOPUS was calibrated using the 383 E. coli samples taken from Gardiners Creek during dry and wet weather periods and assessed using the same procedure as the EG model.

In addition to Gardiners Creek, there are 219 stormwater drains of various sizes that drain directly into the Yarra River estuary (Figure 1 – the 20 biggest shown). MOPUS was further used to generate a time series of stormwater flow rates and microorganism concentrations for each of these stormwater inputs. This was achieved by generating 219 different parameter sets. First, the impervious area for each of the drains was estimated using an empirical relationship between impervious area and drain cross-sectional area (McCarthy 2008). Then, due to the lack of measured data, the five microorganism model parameters were obtained by random sampling within parameter ranges defined by the optimized values from Gardiners Creek Catchment (this study) together with optimized values from literature which has used the MOPUS model on four other stormwater drains in Melbourne, Australia (McCarthy et al. 2011). Finally, the MOPUS model was executed for all 219 drains, using the relevant input data: rainfall, relative humidity and vapour pressure from Melbourne Regional Office station (Figure 1).

Simplified estuary model

The whole estuary was represented as a single reservoir where all modelled flows and microbial loads from the river, Gardiners Creek and 219 stormwater drains were linearly routed and translated through the system (Table 1 for equations). The rationale behind this approach is twofold. First, the Yarra River estuary is a salt-wedge estuary (Beckett et al. 1982), which was confirmed by measurements conducted by authors (data not shown). Essentially, this means that the fresh water layer flows over the moving sea water layer (i.e. salt-wedge), with minimal mixing between the two layers. Furthermore, minutely velocity measurements obtained at Morell Bridge monitoring site using an ADCP over the October 2012 to August 2013 period showed that, on average, velocity in the downstream direction was 0.16 m/s, while the upstream velocity was 0.06 m/s with only 18% of the time velocity being negative, i.e. forming upstream flow. Therefore, the estuary can be effectively regarded as a river with a moveable bottom boundary. Secondly, this model is very simple and would form a baseline level of performance achievable with minimal data input and minimal model complexity. The benefit of further increasing complexity of the model will be assessed in the future against the performance achievable with the simple microorganism model.

Table 1

Estuarine microorganism model (calibration parameters are in bold)

Flow 
 
 
Microbial load 
 
 
Dynamic survival rate 
 
 
Microorganism concentration 
 
Flow 
 
 
Microbial load 
 
 
Dynamic survival rate 
 
 
Microorganism concentration 
 

[m3] – inflow volume stored within estuary; [m3/min] – river inflow; [m3/min] – stormwater inflows; [m3/min] – discharge exiting the estuary; M [MPN] – microorganisms stored within estuary; [MPN/min] – river load rate; [MPN/min] – stormwater load rate; [MPN/min] – load rate exiting the estuary; [-] – routing coefficient; [min] – time of concentration; [min] – time step; k [1/day] – microorganism survival rate; [1/day] – survival rate at 20°C; [%] – percentage sea water; [°C] – measured water temperature; [MJ/m2] – average daily solar radiation; [1/m] – average light attenuation coefficient over depth; [m] – depth of the water column; [mS/cm] – measured electrical conductivity at Morell Bridge; [mS/cm] – electrical conductivity of sea water; [MPN/100mL] – microorganism concentration exiting estuary; – unit conversion factor.

In addition to routing and translating microbes, the model accounts for the impact of environmental factors on the survival of microorganisms in the water column using first-order kinetics. The survival rate was modelled dynamically as a function of temperature, salinity (% sea water) and solar radiation using the expression proposed by Mancini (1978). A simple term has been added when calculating microorganism concentration to account for mixing between fresh and sea water, where sea water was assumed to be free of E. coli.

The estuarine microorganism model was calibrated against Morell Bridge's E. coli dataset (829 points), using the same methods as outlined above for the input models. Simple sensitivity testing was conducted to assess the effect of survival processes and direct stormwater inputs on the model's performance. In the first case, the model was calibrated without accounting for the survival of E. coli (i.e. there was no die-off). In the second case, both survival and stormwater volume and E. coli load were removed and the model was re-calibrated following methodology described above. Furthermore, we assessed the effect of spatial discretization on the model's performance by dividing the estuary into 33 cells of 500 m length. The model equations were applied in each cell.

Input analysis

Predicted stormwater flow rates and microorganism concentrations were used to calculate daily delivered volumes and loads to the estuary. A similar approach was taken with the riverine input, but instead of using predicted flow rates (which were substantially underestimated during base flow periods by the MUSIC model), measured data from Kew were used to achieve more realistic results. To assess the contribution of stormwater in dry and wet weather, in terms of both daily delivered volumes and loads, a ratio of stormwater over total inputs (sum of stormwater and river inputs) was calculated. Similarly, a ratio of daily delivered stormwater volume to the average estuary volume (estimated using Geographic Information System and bathymetry data to be 4 × 106 m3) was also used to assess the impact of direct stormwater inputs.

RESULTS AND DISCUSSION

Input modelling

The MUSIC–SimHyd model reproduced the observed flow pattern reasonably well ; however, during base flow periods there was substantial underestimation of flow rates (probably a result of the model being modified for urbanized catchments). There were also timing issues with the prediction of the peak flows. The stormwater rainfall-runoff model had quite high performance in prediction of flow rates for Gardiners Creek, with an efficiency of . It performed particularly well in the region of very high flow rates (>10m3/s), which was expected as the model was essentially developed and calibrated for the prediction of wet weather flows.

The efficiencies of the two microorganism input models were similar: 0.20 and 0.40. Although these are not high efficiencies, they agree well with the performance reported in the literature for similar microorganism models (McCarthy et al. 2011). The pathogen-catchment model reproduced E. coli patterns well, although there are certain peak prediction time issues similar to that described by Haydon & Deletic (2006). The MOPUS concentration predictions are better in the region of high concentrations, which are commonly observed during wet weather periods. Indeed, the current model structure was developed for modelling wet weather microbial dynamics in stormwater; hence it is expected to give better predictions during wet weather.

Input analysis

The relative contribution of stormwater discharging directly to the estuary during dry weather ranged from <0.5% to 10% (5th and 95th percentile), suggesting limited influence of stormwater on overall E. coli levels in the estuary during these periods (Figure 2(a)). As expected, wet weather stormwater proportions were higher (2% to 50%; 5th and 95th percentile), yet the average daily contribution under these conditions remained marginal (median 10%). These findings agree well with those of Daly et al. (2013), suggesting the median daily E. coli loads coming into the estuary from the three biggest drains (two of them 3 m in diameter and one 6 × 2 m) are about 1.5 orders of magnitude lower than the riverine inputs. However, it is important to note that our results also demonstrate that some conditions can produce high stormwater contributions, especially during periods of low riverine flows and high urban rainfall amounts (see Figure 2(b) and (c)). It is also possible for urban stormwater to enter the estuary much faster than riverine inputs due to the higher imperviousness and the smaller time of concentration of urban catchments. Hence, at finer temporal scales (i.e. time step <1 day), stormwater could have a significant impact on overall faecal contamination levels within the estuary. Furthermore, stormwater might be significantly influencing faecal microbe distribution locally around the drain outlets. All issues stated above would certainly require further investigation, which is not within the scope of this paper.

Figure 2

(a) Modelled daily stormwater contributions during dry/wet weather conditions as a percentage of total delivered water volume (%VOL) and E. coli load (%LOAD) to the estuary (black dots represent 5th and 95th percentiles) for the simulated period of November 2012–August 2013; (b) the relationship between percentage daily stormwater load and riverine input flowrate during dry and wet weather; (c) the relationship between percentage daily stormwater load and urban rainfall (Melbourne Regional Office station).

Figure 2

(a) Modelled daily stormwater contributions during dry/wet weather conditions as a percentage of total delivered water volume (%VOL) and E. coli load (%LOAD) to the estuary (black dots represent 5th and 95th percentiles) for the simulated period of November 2012–August 2013; (b) the relationship between percentage daily stormwater load and riverine input flowrate during dry and wet weather; (c) the relationship between percentage daily stormwater load and urban rainfall (Melbourne Regional Office station).

Estuarine modelling

Considering the simple approach adopted for modelling the estuary (i.e. neglecting estuarine hydrodynamic characteristics), as well as the accuracy of predicting input loads, the model performed reasonably well with and values of 0.37 and 0.41, respectively (Figure 3 and Table 2). Spatial discretization of the estuary into 33 cells did not have a significant effect on the model's efficiencies, which remained similar to the original model ( and ). Due to its simplicity, the model's performance is very much linked to the performance of the input models, emphasizing the effect that the inputs have on the estuarine microbial dynamics and the importance of the adequate representation of these inputs.

Table 2

Parameter ranges, distribution sampled, optimized parameter values and Nash-Sutcliffe efficiencies of the estuarine microorganism model

  Optimized calibration parameters
 
Model efficiency
 
RC TOC k20 keH ECsea EC EClog 
Range 0.001–1 0–3,600 −1.5–1.5 1–1,000 30–60   
Distribution sampleda LogU LogU   
No SW/no die-off & mixing 0.009 540 – – – 0.32 0.34 
No die-off & mixing 0.008 720 – – – 0.34 0.41 
Full model 0.008 284 −0.3 47.7 >60 0.37 0.41 
  Optimized calibration parameters
 
Model efficiency
 
RC TOC k20 keH ECsea EC EClog 
Range 0.001–1 0–3,600 −1.5–1.5 1–1,000 30–60   
Distribution sampleda LogU LogU   
No SW/no die-off & mixing 0.009 540 – – – 0.32 0.34 
No die-off & mixing 0.008 720 – – – 0.34 0.41 
Full model 0.008 284 −0.3 47.7 >60 0.37 0.41 

aU – uniform distribution; LogU – log-uniform distribution; SW – stormwater.

Figure 3

Performance of the estuarine model: (a) predicted versus measured concentrations; (b) predicted versus measured pollutograph during a wet weather event.

Figure 3

Performance of the estuarine model: (a) predicted versus measured concentrations; (b) predicted versus measured pollutograph during a wet weather event.

Initial conclusions can be drawn by relying on the small amount of sensitivity testing conducted here and by exploring the optimized parameter values (Table 2). Switching off stormwater boundary conditions resulted in minimal model efficiency reduction (as indicated by and with and without stormwater input), which may reflect the low average daily contribution from urban stormwater drains.

Furthermore, modelling E. coli die-off results in limited improvement in the model's performance. Indeed, optimized die-off calibration parameter values (Table 2) indicate that the best results are gained when there is minimal die-off. In fact, a negative indicates that there is actually growth due to temperature fluctuations instead of die-off (k20 of −0.3 represents an outlier compared with literature values for E. coli die-off from 0.48 in fresh water to 1.09 in sea water (Hipsey et al. 2008)). The optimized light attenuation coefficient of 11.9 1/m (calculated assuming average fully mixed depth ) is more than twice as high as that reported in the literature for highly turbid estuaries (Devlin et al. 2008), indicating a tendency of the model to minimize die-off by reducing the detrimental effect of sunlight on microbial survival. This is also the case for the optimized EC value of sea water. The issues described above could be related to the fact that the model is very simple and does not fully account for the hydraulic and microbial complexity of the estuarine environment. In fact, resuspension of sediments can increase microbial concentration in water systems (Pachepsky & Shelton 2011), and hence the growth observed here could be compensating for the absence of this process in the model.

CONCLUSIONS

An integrated conceptual-level model of the whole estuarine catchment was developed. Existing models for modelling faecal microorganisms in river and stormwater were linked with a new estuarine microorganism model which accounted for microbial die-off due to temperature, salinity and sunlight. The mass balance analysis using model predictions of daily faecal microorganism loads delivered to the Yarra River estuary via riverine input, Gardiners Creek and 219 stormwater drains discharging directly to the estuary revealed limited influence of urban stormwater on the estuary during dry weather. Wet weather contributions from stormwater drains were significant in some cases (95th percentile of 50%); however, the average contribution remained marginal (median 10%). Sensitivity analysis of the new highly simplified estuarine microorganism model showed minimal change in model performance when direct stormwater inputs were removed. This may reflect the low average daily contribution from urban stormwater drains. Both input analysis and sensitivity testing confirm previous studies showing that E. coli loads derived stormwater drains are dwarfed by other inputs. Nevertheless, it is essential to note that these results also demonstrate that some conditions reveal the opposite; high proportions from stormwater are possible when combined with low riverine inputs and high urban rainfall amounts. This study focuses on the overall impacts of direct stormwater inputs on faecal contamination levels within the estuary, and localized impacts require further investigation.

In spite of the very simplistic modelling approach and high likelihood of missing an important process or input (as indicated by optimized parameter values), the estuarine microorganism model performed reasonably well. This is likely due to the significant effect of inputs on microbial dynamics within the estuary itself. Therefore, appropriate representation of inputs is a requirement for modelling faecal contamination in urban estuaries. Additionally, the model's performance encourages further investigation of simple conceptual ways of modelling faecal contamination in narrow river-like urban estuaries.

ACKNOWLEDGEMENTS

The authors wish to acknowledge Melbourne Water and the Australian Research Council (LP120100718) for providing funding for this project. Our gratitude goes to C. Shang, P. Kolotelo, R. Williamson and M. Siddiqee.

REFERENCES

REFERENCES
Beckett
R.
Easton
A.
Hart
B.
Mckelvie
I.
1982
Water movement and salinity in the Yarra and Maribyrnong estuaries
.
Marine and Freshwater Research
33
,
401
415
.
Burton
G. A.
Pitt
R. E.
2002
Stormwater Effects Handbook: A Toolbox for Watershed Managers, Scientists and Engineers
.
Lewis Publisher
,
Boca Raton
.
Daly
E.
Kolotelo
P.
Schang
C.
Osborne
C. A.
Coleman
R.
Deletic
A.
Mccarthy
D. T.
2013
Escherichia coli concentrations and loads in an urbanised catchment: The Yarra River, Australia
.
Journal of Hydrology
497
,
51
61
.
de Brauwere
A.
de Brye
B.
Servais
P.
Passerat
J.
Deleersnijder
E.
2011
Modelling Escherichia coli concentrations in the tidal Scheldt river and estuary
.
Water Research
45
,
2724
2738
.
Devlin
M. J.
Barry
J.
Mills
D. K.
Gowen
R. J.
Foden
J.
Sivyer
D.
Tett
P.
2008
Relationships between suspended particulate material, light attenuation and Secchi depth in UK marine waters
.
Estuarine, Coastal and Shelf Science
79
,
429
439
.
Desmarais
T. R.
Solo-Gabriele
H. M.
Palmer
C. J.
2002
Influence of soil on fecal indicator organisms in a tidally influenced subtropical environment
.
Applied and Environmental Microbiology
68
,
1165
1172
.
Dotto
C. B. S.
Kleidorfer
M.
Deletic
A.
Fletcher
T. D.
Mccarthy
D. T.
Rauch
W.
2010
Stormwater quality models: performance and sensitivity analysis
.
Water Science and Technology
62
,
837
843
.
Dotto
C. B. S.
Kleidorfer
M.
Deletic
A.
Rauch
W.
Mccarthy
D. T.
Fletcher
T. D.
2011
Performance and sensitivity analysis of stormwater models using a Bayesian approach and long-term high resolution data
.
Environmental Modelling and Software
26
,
1225
1239
.
eWater
2012
Music V5 – User Manual. eWater
.
Gao
G.
Falconer
R. A.
Lin
B.
2011
Numerical modelling sediment–bacteria interaction processes in the Severn Estuary
.
Journal of Water Resource and Protection (JWARP)
3
,
22
31
.
Garcia-Armisen
T.
Thouvenin
B.
Servais
P.
2006
Modelling faecal coliforms dynamics in the Seine estuary, France
.
Water Science and Technology
54
,
177
184
.
Haydon
S.
Deletic
A.
2006
Development of a coupled pathogen-hydrologic catchment model
.
Journal of Hydrology
328
,
467
480
.
Hipsey
M. R.
Antenucci
J. P.
Brookes
J. D.
2008
A generic, process-based model of microbial pollution in aquatic systems
.
Water Resources Research
44
,
W07408
.
IDEXX Laboratories
2013
Colilert*-18.
IDEXX
,
Westbrook, Maine
,
USA
.
Kashefipour
S. M.
Lin
B.
Harris
E.
Falconer
R. A.
2002
Hydro-environmental modelling for bathing water compliance of an estuarine basin
.
Water Research
36
,
1854
1868
.
Lipp
E. K.
Kurz
R.
Vincent
R.
Rodriguez-Palacios
C.
Farrah
S. R.
Rose
J. B.
2001
The effects of seasonal variability and weather on microbial fecal pollution and enteric pathogens in a subtropical estuary
.
Estuaries
24
,
266
276
.
Liu
W.-C.
Huang
W.-C.
2012
Modeling the transport and distribution of fecal coliform in a tidal estuary
.
Science of The Total Environment
431
,
1
8
.
Makepeace
D. K.
Smith
D. W.
Stanley
S. J.
1995
Urban stormwater quality: Summary of contaminant data
.
Critical Reviews in Environmental Science and Technology
25
,
93
139
.
Mallin
M. A.
Williams
K. E.
Esham
E. C.
Lowe
R. P.
2000
Effect of human development on bacteriological water quality in coastal watersheds
.
Ecological Applications
10
,
1047
1056
.
Mancini
J. L.
1978
Numerical estimates of coliform mortality rates under various conditions
.
Journal of the Water Pollution Control Federation
50
,
2477
2484
.
Melbourne Water
2013
Stormwater strategy: a Melbourne Water strategy for managing rural and urban runoff
.
Melbourne Water Corporation
,
Melbourne
,
Australia
.
McCarthy
D. T.
2008
Modelling microorganisms in urban stormwater
.
PhD Thesis, Monash University
.
McCarthy
D. T.
Deletic
A.
Mitchell
V. G.
Diaper
C.
2011
Development and testing of a model for Micro-Organism Prediction in Urban Stormwater (MOPUS)
.
Journal of Hydrology
409
,
236
247
.
Nash
J.
Sutcliffe
J.
1970
River flow forecasting through conceptual models part I – A discussion of principles
.
Journal of Hydrology
10
,
282
290
.
Ortega
C.
Solo-Gabriele
H. M.
Abdelzaher
A.
Wright
M.
Deng
Y.
Stark
L. M.
2009
Correlations between microbial indicators, pathogens, and environmental factors in a subtropical estuary
.
Marine Pollution Bulletin
58
,
1374
1381
.
Pachepsky
Y. A.
Shelton
D. R.
2011
Escherichia coli and fecal coliforms in freshwater and estuarine sediments
.
Critical Reviews in Environmental Science and Technology
41
,
1067
1110
.
Schäfer
A.
Ustohal
P.
Harms
H.
Stauffer
F.
Dracos
T.
Zehnder
A. J. B.
1998
Transport of bacteria in unsaturated porous media
.
Journal of Contaminant Hydrology
33
,
149
169
.