Assessing the combined effects of urbanisation and climate change on the river water quality in an integrated urban wastewater system in the UK

Journal of Environmental Management 112 (2012) 1e9

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Assessing the combined effects of urbanisation and climate change on the river water quality in an integrated urban wastewater system in the UK Maryam Astaraie-Imani*, Zoran Kapelan, Guangtao Fu, David Butler Centre for Water Systems, College of Engineering, Mathematics and Physical Science, University of Exeter, North Park Road, Exeter, EX4 4QF Devon, United Kingdom

a r t i c l e i n f o

a b s t r a c t

Article history: Received 15 June 2011 Received in revised form 13 June 2012 Accepted 23 June 2012 Available online

Climate change and urbanisation are key factors affecting the future of water quality and quantity in urbanised catchments and are associated with significant uncertainty. The work reported in this paper is an evaluation of the combined and relative impacts of climate change and urbanisation on the receiving water quality in the context of an Integrated Urban Wastewater System (IUWS) in the UK. The impacts of intervening system operational control parameters are also investigated. Impact is determined by a detailed modelling study using both local and global sensitivity analysis methods together with correlation analysis. The results obtained from the case-study analysed clearly demonstrate that climate change combined with increasing urbanisation is likely to lead to worsening river water quality in terms of both frequency and magnitude of breaching threshold dissolved oxygen and ammonium concentrations. The results obtained also reveal the key climate change and urbanisation parameters that have the largest negative impact as well as the most responsive IUWS operational control parameters including major dependencies between all these parameters. This information can be further utilised to adapt future IUWS operation and/or design which, in turn, should make these systems more resilient to future climate and urbanisation changes. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Climate change Integrated modelling Sensitivity analysis Urbanisation Wastewater system Water quality

1. Introduction Climate change and urbanisation are among the key factors affecting the future of water quality and quantity in urbanised catchments and the ones associated with most uncertainty. Climate change modifies the natural hydrologic cycle and results in, among other things, changing rainfall patterns. Urbanisation leads to changing land use and water consumption patterns. As a result, the combined effect of the two leads to changing waste and storm water flow patterns and associated pollutant loadings which, in turn, is likely to have an impact on the aquatic life in the receiving waters, i.e. rivers. In order to respond to these future challenges, it is vital to understand the impact of various related factors on urban river water quality, including their uncertainties and how the mediating wastewater system can best respond and adjust to maintain performance. This is best carried out through the lens of an integrated urban wastewater system (IUWS) perspective whereby the urban wastewater system is considered as a whole comprising sewer system, wastewater treatment plant (WWTP) and the recipient. This * Corresponding author. Tel.: þ44 (0)1392 723600; fax: þ44 (0)1392 217965. E-mail addresses: [email protected], [email protected] (M. AstaraieImani). 0301-4797/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jenvman.2012.06.039

integrated approach to wastewater systems modelling has been demonstrated in the academic literature (Schütze et al., 2002; Butler and Schütze, 2005), although there are few practical examples of any scale (Butler and Davies, 2011). It is, however, firmly in line with the precepts of the EU Water Framework Directive (Kallis and Butler, 2001). So far this approach has typically been applied with the motivation of improving the performance of the whole system by real time control, based on achieving appropriate receiving water quality criteria (Vanrolleghem et al., 2005; Fu et al., 2008). The potential impacts of climate change on water systems have been and are being widely studied. In this specific domain for example, Mimikou et al. (2000) assessed the impacts of climate change on river water quality based on two climate change scenarios. Wilby et al. (2006) used an integrated framework to model climate change impacts on river water quality and quantity in the UK. Whitehead et al. (2006), Delpla et al. (2009) and Park et al. (2010) investigated the impacts of climate change on river water quality. The impact of urbanisation on wastewater systems and the receiving water quality has been investigated in this study. Tong and Chen (2002) revealed that there was a significant relationship between land use and in-stream water quality at a regional scale in the State of Ohio, USA. He et al. (2008) investigated the response of surface water quality to urbanisation in Xi’n, China. Fu

2

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

et al. (2009) quantified the impact of new developments on river water quality in the integrated wastewater system context. Jacobson (2011) provided a review on the impacts of urban imperviousness on hydrological systems in diverse areas. The combined or relative affects of climate change and urbanisation have not gained so much attention in the context of urban wastewater management. There are a few exceptions. For example, Semadeni-Davies et al. (2008a, 2008b) studied the potential impacts of climate change and continued urbanisation on storm water flows in a suburban stream and combined sewer system, Helsingborg, Sweden. Tu (2009) analysed the combined impact of climate and land use changes on water quality in watersheds of eastern Massachusetts. Notwithstanding the uncertainties in emissions scenarios, climate models and population projections, uncertainty also arises from modelling of the urban drainage system. In order to scientifically analyse and assess the future changes in environmental quality and their uncertainties, there is a need to quantitatively characterise the uncertainties of environmental impacts by systematically identifying the relationships between the future changes and the environmental quality (Zhou et al., 2010). To achieve this, Mannina and Viviani (2010) and Thorndahl et al. (2008) used the Generalised Likelihood Uncertainty Estimation (GLUE) method to evaluate uncertainty. Zhou et al. (2010) developed an integrated assessment method based on accounting for the uncertainty of water quality. Dessai and Hulme (2007) provided an assessment framework that allows the identification of adaptation strategies that are robust (i.e. insensitive) to climate change uncertainties in a water resources management framework in the UK. Sin et al. (2009) focused on uncertainty analysis of WWTP models. The aim of this study is to identify the input and operational control parameters that have the most significant impact on the receiving water quality in an IUWS context under the combined effects of climate change and urbanisation. An integrated urban wastewater model is used to simulate receiving water quality in an urban river represented by dissolved oxygen (DO) and ammonium (AMM). The input parameters considered are from three categories, i.e., climate change, urbanisation and system operational control parameters. Two sensitivity analysis methods, one local and one global, together with correlation analysis are used to analyse the sensitivity of river DO and AMM concentrations to above parameters. 2. Methodology 2.1. Integrated urban wastewater system modelling The simulation model of the IUWS studied here is based on the SIMBA5 simulation tool, developed by IFAK (2005). SIMBA5 consists of a library of modelling blocks (in the Matlab/Simulink environment) to simulate different individual components in the IUWS. The sewer system is modelled using a hydrological approach. The hydrological processes and losses considered include surface runoff from both pervious and impervious areas, wash-off, flow and pollutant transport in sewers and storage tanks. Flow between subcatchments (i.e. pipe flow) is modelled using a Nash Cascade model. The Activated Sludge Model No.1 (Henze et al., 1986) is used for modelling the WWTP. The IWA Task Group used this model as the Benchmark Simulation Model No.1 (Copp, 2002). The river is simulated using the EPA Storm Water Management Model (SWMM) (Huber and Dickinson, 1988). The river water quality model and the relevant conversions described by Schütze et al. (2002) are used in this study. The water quality in the river is

Table 1 Urbanisation parameters’ nominal values and their value ranges. Parameter

Unit

Nominal value

Range

POP IMP PCW NHþ 4

% % litre/person/day mg/l

0 0 180 27.7

[4.5, 15] [5, 15] [80, 260] [20, 30]

measured by AMM and DO. As each of three subsystems has its own characteristics and is simulated using different sets of pollutants, a factor-based conversion method is used to convert three subsystems models into a single integrated model (Rauch et al., 2002; Vanhooren et al., 2003). 2.2. Sensitivity analysis methods Saltelli et al. (2006) classified the Sensitivity Analysis methods into three categories: screening methods, local sensitivity analysis (LSA) and global sensitivity analysis (GSA). The local, one-factor-ata-time method (OAT), and the global, Regional Sensitivity Analysis method (RSA) are used here. 2.2.1. Local sensitivity analysis The LSA is used here to screen the significance of various climate change, urbanisation and operational control parameters and their potential impact on the receiving water quality. This was done by perturbing one parameter value at a time while keeping the values of other input parameters fixed at their nominal (i.e. ‘Base Case e BC’) values. Once this is done, only the most sensitive parameters filtered this way are then considered in the GSA. The BC represents the ‘business as usual’ in the existing IUWS, i.e. with no climate nor urbanisation changes, with all IUWS parameters set to their default values. The Tornado type graphs are used to visualise the LSA results. These graphs were prepared as follows: (1) Run the IUWS model with all sensitivity parameters set equal to their nominal values (BC); (2) Select one IUWS model input parameter and change its value from default to upper (i.e. maximum) or lower (i.e. minimum) value in the considered range. Keep the other input parameter values at their nominal values (see Table 1 and Table 2); (3) Run the IUWS model and evaluate the analysed IUWS model outputs; (4) Calculate the relative (i.e. percent) change for the IUWS model outputs relative to the BC and (5) Rank the relative differences obtained in a descending order. Note that Tornado graphs obtained include both negative and positive values. The positive values represent an increase relative to the BC and vice versa. 2.2.2. Global sensitivity analysis In GSA, the analysed IUWS model input parameters are varied simultaneously and the effects of their possible interactions are taken into account. Therefore, unlike LSA, the GSA takes into account both local and global effects of IUWS model inputs on outputs. RSA has been widely applied in the fields of environmental Table 2 Operational control parameters’ nominal values and their value ranges. Operational control parameter

Unit

Nominal value

Range

Qmaxout Qmaxin Qtrigst Qempst QRAS

m3/d m3/d m3/d m3/d m3/d

5  DWF 3  DWF 24,192 12,096 14,688

[3  DWF, 8  DWF] [2  DWF, 5  DWF] [16416, 31,104] [6912, 24,192] [6912, 24,192]

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

and hydrological modelling (McIntyre et al., 2003; Cox and Whitehead, 2005), and has been used here because of its capability to perform a GSA for highly-dimensional and non-linear models (Hornberger and Spear, 1981) such as the IUWS model used here. The main idea behind the RSA is a division of the IUWS model output space into behavioural (B) or non-behavioural (NB) regions in terms of a priori defined criteria. The RSA procedure has the following principal steps: (1) Identify the most important sources of uncertainty in the IUWS model input parameters using the LSA method described in Section 2.2.1; (2) Characterise the aforementioned uncertainties by assigning a probability density function (uniform distribution used here); (3) Generate multiple input parameter sets by using the Latin Hypercube Sampling technique; (4) Run the IUWS model to evaluate the IUWS model outputs of interest (AMM and DO) and classify each input parameter set as either ‘behavioural’ or ‘non-behavioural’, based on the a priori defined criterion. For a multi-objective analysis, this classification is conducted for each objective separately; (5) Assign each of the parameter sets likelihood estimated from its relevant output value for behavioural and non-behavioural groups of samples. The marginal cumulative distribution of behavioural and nonbehavioural parameter sets is then derived for each output. The difference between the two marginal cumulative distributions can be estimated by using a two-sample KolmogoroveSmirnov (KS) test. The statistic Dm,n is determined as the maximum vertical distance between the cumulative distributions for behavioural and non-behavioural as below:

Dm;n ¼ supjSB ðxÞ  SNB ðxÞj

(1)

where SB and SNB are the empirical distribution functions for n behavioural and m non-behavioural samples, respectively. The larger the value of Dm,n, the more significant the corresponding parameter.

3

3. Case study 3.1. Description The case study used here is a semi-real case study (see Fig. 1). This case study has been used for various purposes, including real time control (Butler and Schütze, 2005; Schütze et al., 2002), system RTC potential analysis (Zacharof et al., 2004) and system impact analysis (Lau et al., 2002). The IUWS is divided into the sewer subsystem, the WWTP and the river. The sewer system analysed here is an example sewer system used by ATV (1992). It has 7 sub-catchments with a total area of 725.8 ha. The average Dry Weather Flow (DWF) is approximately 27,500 m3/d. A storage tank is located at the WWTP inlet to control the Combined Sewer Overflows (CSOs) by storing the wastewater in excess to the WWTP pumping capacity. The storage tank release capacity is controlled by means of thresholds on maximum outflow rates. Lessard (1989) provided a dataset for concentration of different water quality indicators for dry and storm water flows which are applied for the sewer system in this study. The nitrifying activated sludge plant in Norwich (UK) is used here as the basis of the treatment plant simulations. The system has a treatment capacity of 27,500 m3/d which is equal to DWF. It consists of a storm tank, primary clarifiers, activated sludge reactor and secondary clarifiers. The storm tank is controlled by the maximum inflow rate to the primary clarifiers. The tank is emptied at a certain pump rate, as soon as the inflow rate to the plant drops below a pre-specified threshold value. The WWTP model has been previously calibrated and validated (Lessard and Beck, 1993). The river system is a hypothetical, 45 km long stretch divided into 45 equal reaches. The runoff generated by rainfall on the upstream catchments enters the system as an additional inflow into the river at reach 1. The CSOs are assumed to discharge at reach 7 with the storm tank overflows and treatment plant effluent at reach 10. The above IUWS operation was assessed using a six day rainfall event during 7the13th February 1977 with a total depth of 27 mm.

2.3. Correlation analysis 3.2. IUWS model input parameters Correlation analysis is used here to indicate a predictive relationship between sensitivity parameters analysed. Scatter plots of most important input parameters are presented as they can be considered as global measures of correlation (Helton, 1993). In addition, the Pearson correlation coefficient R is used to indicate the strength of the linear relationship between the two IUWS model variables (Saltelli et al., 2000).

Sewer System

Wastewater Treatment Plant CSO

SC2

flow

(Tank)

SC4

Pump 1

(Tank)

SC7 (Tank)

SC5

SC6 (Tank)

POP: Used to represent percentage increase in population over a given time period. This parameter is related to the DWF but

CSO discharge

Reach 7

Inflow Primary Clarifier

Storm Tank

Reactor

Secondary Clarifier

Effluent

SC3

-

Return Flow

SC1

3.2.1. Urbanisation parameters Urbanisation can, in principle, be represented by a number of different parameters, but the following are used in this study:

Return Sludge Pump 2

Waste Sludge

Dispose

Discharge

River

Fig. 1. Schematic diagram of the case study of IUWS.

Reach 10

4

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

-

-

-

may also influence the DWF quality (Butler and Davies, 2011). Population growth has also been shown to be related to other urbanisation indices such as housing density and occupancy (Environment Agency, 2006; Jefferies, 2005; Office for National Statistics, 2010). In the UK, population growth by 2030 is predicted to be between 4.5% (as a minimum in Table 1) and 15% (as a maximum in Table 1) (Department for Communities and Local Government (DCLG), 2010). PCW: Defined as a daily Per Capita Water consumption. This factor has a key role in influencing domestic wastewater quantities. Butler and Davies (2011) indicate that 95% of the water consumed in the UK is returned to the sewer system as wastewater. PCW is primarily influenced by factors such as changing household demographic composition, the changing structure of the UK economy and the water charging policies (Environment Agency, 2006) but may also be influenced by climate change via increased air temperature. The possible range of this parameter is set according to Defra (2006) as shown in Table 1. IMP: IMP represents the percentage increase in impervious surfaces which has a direct influence on the rate of storm water runoff in urban areas (Butler and Davies, 2011). The growth of impervious surfaces, by definition, implies a change from natural catchment surface status as influenced by rates and types of development in urban areas. The rate of increase of imperviousness, in a given urban area is defined as the urban creep, and has increased significantly over recent years (UKWIR, 2010). The UK government policy is currently being changed (DCLG, 2010) hence it is difficult to estimate future urban creep and the resulting amount of imperviousness. The range of the IMP values considered here (see Table 1) is based on the values reported by CIWEM (2009). þ NHD 4 : NH4 is defined as the concentration of Ammonium in DWF. This is a rather different parameter to the above ones as it assumes the possible roll out of urine separation toilets. This is unlikely in the UK in the short-term, but is a possible in the long-term. Urine separation toilets reduce both the hydraulic and nutrient loads in the sewer systems. Achileitner et al. (2007) and Semadeni-Davies et al. (2008b) estimated that urine separation in new homes could reduce the specific load of nitrogen by some 25%. This has been used as a basis for the range of values considered (see Table 1).

-

yt ¼ maxð0; x þ ðxt  x  cÞ  dÞ

-

RD: This parameter represents the Rainfall Depth in mm. The 30 percent increase in 2080 is possible for winter rainfall in the UK under the aforementioned emissions scenario. Based on this, the 10, 20 and 30 percent increases in rainfall depth were considered here (denoted as RD10, RD20, and RD30). These RD increases have been applied to the base rainfall intensity to increase the total rainfall depth without modifying the duration of the base rainfall.

(2)

where yt is the value of rainfall intensity at time t, xt is the value of base rainfall intensity at time t, xis the average of rainfall intensity in the base rainfall, d is a scaling factor and c is a weighting factor that acts as the adjustor of x. As shown in Equation (2) the values of yt have a minimum value of zero to avoid generating negative rainfall rates which can result from the total rainfall depth increases. To generate a new rainfall series, an initial value is assigned to the factor c, and the new rainfall series are calculated using Equation (2). Then the total rainfall depth of the new series is computed and compared with the base rainfall depth value. If the two depth values are not equal the factor c is used to adjust the difference by trial and error. This process is repeated until the two series have the same rainfall depth. 3.2.3. IUWS operational control parameters The following IUWS operational control parameters are analysed here: -

-

-

-

-

3.2.2. Climate change parameters In this study, rainfall has been selected as the indicator of climate change. Hulme et al. (2002) and IPCC (2000) estimated the future pattern changes in rainfall for the UK indicating a future with wetter winters and drier summers for some regions under certain climate change scenarios. The MediumeHigh IPCC emissions scenario has been adopted as a basis for the analyses done here. According to this scenario, an increase in rainfall depth and/or intensity is likely under the future climate change. Rather than undertake a detailed regional climate model study, a simplified approach has been adopted in this study. Here, the rainfall increase is represented by using the following two parameters (as in Hulme et al., 2002):

RI: This parameter represents the Rainfall Intensity in mm/h. Three cases of 10, 20 and 30 percent increases to the base rainfall intensities (denoted as RI10, RI20, RI30) are analysed here. The increase is achieved by applying a fixed percentage increase to rainfall intensities across the entire event/timeseries whilst maintaining the same cumulative depth as follows:

Qmaxout: This parameter is the maximum outflow rate from the storage tank in the sewer system. This parameter controls CSO discharges into the river and the WWTP inflow. Qmaxin: This parameter represents the maximum inflow to the WWTP. It controls the inflows to the primary clarifiers and has an impact on the rate of storm tank overflows into the river. Qtrigst: This parameter defines the threshold at which the storm tank emptying is triggered. This parameter controls the storm tank operation in terms of river overflows. Qempst: The filling of the storm tank is controlled by the maximum inflow rate to the primary clarifiers. The tank is emptied at a certain pump rate, as soon as the inflow rate to the plant drops below a pre-specified threshold value. QRAS: The return activated sludge is taken from the secondary clarifiers and is pumped back into the aerator head. The waste sludge is set constant to 660 m3/d.

The nominal values of the above parameters and their possible ranges are given in Table 2.

3.3. IUWS model outputs The following two IUWS model outputs are used to represent water quality in the receiving water: -

-

Minimum Dissolved Oxygen Concentration in the river (DO): DO is one of the most important water quality factors that can affect the aquatic life in the rivers if it drops too low. Maximum Ammonium concentration in the river (AMM): Ammonium is a component of DWF that affects the quality of water in the rivers by consuming the DO.

The critical threshold of 4 mg/l is used here for both DO and AMM based on the Urban Pollution Management (UPM) manual (FWR, 1998).

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

RD30

RI30

RD30 Qmaxout PCW Qmaxin RD20 RI30 RD10 RI20 RI10 NH4 POP IMP QRAS Qtrigst Qempst

RD30 PCW Qmaxin RD20 RI30 POP IMP Qmaxout RD10 RI20 RI10 Qtrigst NH4 QRAS Qempst

a

-100

-80

-60

-40

-20

0

20

40

Qmaxout RD30 Qtrigst PCW NH4 QRAS Qmaxin Qempst RD20 RD10 RI20 RI10 POP RI30 IMP -10

0

10

20

30

40

50

-80

-60

-40

-20

0

20

Relative variation of DO to the Base Case for maximum values of System Control Parameters (%) Qmaxin PCW RD30 POP NH4 RD20 RD10 IMP QRAS RI10 RI20 RI30 Qtrigst Qempst Qmaxout

c -20

b

-100

Relative variation of DO to the Base Case for minimum values of System Control Parameters (%)

-30

5

60

Relative variation of AMM to the Base Case for minimum values of System Control Parameters (%)

d

-10

0

10

20

30

40

50

60

70

Relative variation of AMM to the Base Case for maximum values of System Control Parameters (%)

Fig. 2. LSA results.

4. Results and discussion

impact on the AMM in the river as it controls the storm tank overflows (see Fig. 2c).

4.1. Local sensitivity analysis results The LSA results are summarised in Tornado graphs shown in Fig. 2. Fig. 2a and b depict the sensitivity of DO to the analysed IUWS model inputs and Fig. 2c and d depict the corresponding results in the AMM case. The following points are deduced from these figures: -

-

-

The most significant climate change parameter is RD for both DO and AMM. As shown in Fig. 2a and b, the 30 percent increase in rainfall depth results in a more severe deterioration of DO than any other input parameter. Similar deterioration is observed for AMM in Fig. 2c and b. This deterioration is a consequence of increasing the CSOs and storm tank overflows with excess rainfall volume. PCW has the most significant impact on river water quality of all urbanisation parameters considered. The minimum value of PCW improves the DO and AMM (see Fig. 2a and c) while the opposite happens when the PCW is set to the maximum value (see Fig. 2b and c). As shown in Fig. 2, Qmaxout and Qmaxin are the two most significant operational control parameters with the highest impact on the river water quality. This is because these two parameters directly control the CSOs and storm tank overflows which, in turn, affect the quality of water in the river (Butler and Davies, 2011). In addition, Qtrigst also has a significant

Based on the LSA results shown in Fig. 2, the following significant parameters are selected for further investigation in the GSA: RD and RI (with 30% increase only), PCW, POP, IMP, Qmaxout, Qmaxin and Qtrigst. Even though the RI parameter did not show as significant impact on the recipient water quality as the RD, it was investigated further in the GSA for possible interactions with other parameters. 4.2. Global sensitivity analysis results The above IUWS model input parameters were used to generate corresponding samples by using the methodology described in Table 3 KS statistic of each parameter for the model outputs.

Urbanisation parameters

Operational control parameters

Model outputs

DO

AMM

Climate change parameters

RD30

RI30

RD30

RI30

IMP POP PCW NHþ 4 Qmaxout Qmaxin Qtrigst Qempst QRAS

0.68 0.20 0.76 0.33 0.93 0.29 0.28 0.20 0.14

0.27 0.11 0.50 0.07 0.64 0.26 0.07 0.07 0.06

0.16 0.11 0.41 0.15 0.59 0.13 0.13 0.04 0.06

0.12 0.07 0.40 0.18 0.56 0.20 0.10 0.06 0.06

6

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

D-line 1

B

Rainfall Depth (RD30)

Rainfall Intensity (RI30)

1

1

2 0.5

0.5

0 1.05

Cumulative Probability

NB

1.07

1.09

1.11

1.13

1.15

0 1.05

1.07

1.09

1

1.11

1.13

1.15

IMP (%)

IMP (%) 1

3

4

0.5

0.5

0 80

100

120

140

160

180

200

220

240

260

0 80

100

120

PCW (lit/person/day)

140

160

180

200

220

240

260

220

240

260

PCW (lit/person/day)

1

1

5

6

0.5

0.5

0 80

100

120

140

160

180

200

220

240

260

0 80

100

120

PCW (lit/person/day)

140

160

180

200

PCW (lit/person/day)

Fig. 3. CDFs of the selected urbanisation parameters from GSA under two climate change parameters with regard to DO (Fig. 3-1 to 3-4) and AMM (Fig. 3-5 to 3-6) in the river, respectively. B denotes the ‘behavioural’ group and NB denotes the ‘non-behavioural’ group.

analysed water quality indicators. The corresponding KS statistics obtained in RD30 and RI30 cases are shown in Table 3.

Section 2.2.2. The pre-defined threshold value of 4 mg/l was applied for both water quality parameters to divide the obtained parameter samples into ‘behavioural’ and ‘non-behavioural’. The behavioural group of samples has a DO concentration above 4 mg/l and the AMM concentration below 4 mg/l and vice versa for the nonbehavioural groups of samples. The cumulative distribution functions (CDFs) of both groups were plotted with regard to the

4.2.1. GSA results for the urbanisation parameters The most significant sensitivities obtained for the urbanisation parameters are shown in Fig. 3. The following observations can be made:

D-line

Rainfall Depth (RD30)

B

1

1 0.5

0 3

4

5

6

7

8

1

Cumulative Probability

Rainfall Intensity (RI30)

2

0.5

0 3 1

3

4

5

6

7

8

5

6

7

8

4

0.5

0.5

0 3

4

5

6

Qmaxout(*27500, m3/d)

7

8

RD30

1

0 3

4

Qmaxout(*27500, m3/d) RD30

1

5

6

0.5

0 2

NB

1

0.5

2.5

3

3.5

4

Qmaxin (*27500, m3/d)

4.5

5

0 700

800

900

1000

1100

1200

Qtrigst (*24, m3/d)

Fig. 4. CDFs of the selected operational control parameters from GSA under two climate change parameters with regard to DO (Fig. 4-1, 4-2, 4-5 and 4-6) and AMM (Fig. 4-3 to 4-4) in the river.

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9 Table 4 Correlations between system input parameters and IUWS model outputs.

-

IUWS model output e climate change parameters

Qmaxout

PCW

DO (under RD) DO (under RI) AMM (under RD) AMM (under RI)

0.75 0.87 0.18 0.15

0.017 0.46 0.43 0.51

-

-

-

In general, departures of the urbanisation parameters from the diagonal line (D-line) under RD30 are more significant than under RI30. In other words, the sensitivity of the water quality indicators to the urbanisation parameters under RD30 is more significant than under RI30. The most significant departures from the D-line are observed for the PCW parameter (see Fig. 3-3 to 3-6). Under RD30 (see Fig. 3-3, 3-5), the PCW values leading to compliant water quality standards (see KS statistics in Table 3), are predominantly in the lower bound of the considered PCW range. In other words, reducing the PCW is important for meeting the water quality standards in the river (especially the DO standards) in a changing climate. Fig. 3-1 and 3-2 show that DO is considerably more sensitive to the IMP than the AMM and this sensitivity is more significant under RD30. The excess volume of rainfall under RD30 causes more runoff and this runoff intensifies when the imperviousness increases. The increased runoff causes CSOs and storm tank overflows which, in turn, lead to the deterioration of DO in the river. Therefore, under a changing climate, attention must be paid to controlling the imperviousness (i.e. the urban creep), although this is of less importance than the PCW.

-

-

The departures from the D-line in Fig. 4-1 and 4-2 show that the Qmaxout parameter has the most significant impact on DO. Comparing Fig. 4-1 and 4-2 with Fig. 4-3 and 4-4 indicate that the DO is more sensitive to Qmaxout than AMM. Also, as it can be observed from Fig. 4-1 to 4-4, the maximum departures (dashed lines) are obtained near the maximum value of Qmaxout in the RD30 case. This is because increased Qmaxout reduces the volume of CSO discharges into the river which, in turn, helps maintaining the water quality in the river. Note also that the sensitivity of Qmaxout to the RD is larger than the corresponding sensitivity to the RI. The DO is also sensitive to Qmaxin (but less than to Qmaxout) under both climate change parameters with no marked difference (hence Fig. 4-5 depicts this sensitivity under RD30 only). This figure also shows that most of the behaving samples (B dashed line) occur for larger Qmaxin values. This is because increasing Qmaxin has the potential to reduce the storm tank overflows which in turn, improves the DO. The sensitivity of DO to Qtrigst is observed only in the RD30 case, as shown in Fig. 4-6. The maximum departure from the D-line occurs near the lowest Qtrigst values. This indicates that for maintaining the DO quality in the river, the control scheme reduces the storm tank overflows. Fig. 4-6 shows that, under the RI30, AMM is not sensitive to Qtrigst.

4.3. Correlation analysis results Although the KS statistic provides an insight into the distinction between behavioural and non-behavioural distributions, it may not identify the regional sensitivity hidden by high correlation between the analysed IUWS model parameters. Therefore the results of the RSA should be interpreted in conjunction with the parameter covariance or correlation matrix (McIntyre et al., 2003). As the sensitivity results showed, PCW and Qmaxout are the two most important parameters. Table 4 shows the correlation coefficients for PCW and Qmaxout. The significant correlation, i.e. relationship is observed between Qmaxout and DO for both RD30 and RI30 as both values (0.75 and 0.87

4.2.2. GSA results for the operational control parameters The most significant sensitivities obtained for the operational control parameters are shown in Fig. 4. Fig. 4-1 to 4-4 show the sensitivity of DO to Qmaxout and Fig. 4-5 to 4-6 show the sensitivity of DO to Qmaxin and Qtrigst respectively. The following observations can be made from these figures:

RD30

RI30 6

a

DO Concentartion (mg/l)

DO Concentartion (mg/l)

5 4 3 2 1 0 3

4

5

6

7

5

b

4 3 2 1 0 3

8

4

5

24

c

16 12 8 4 0 80

130

180

PCW (lit/person/day)

6

7

8

Qmaxout (m3/d)

AMM Concentartion (mg/l)

AMM Concentartion (mg/l)

Qmaxout (m3/d) 20

7

230

260

20

d 16 12 8 4 0 80

130

180

230

260

PCW (lit/person/day)

Fig. 5. Scatter plots between parameters (PCW and Qmaxout) and water quality indicators (DO and AMM) under climate change parameters (RD30 and RI30).

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9

Cumulative Probability

8

1 0.8

1 99 % 88 %

a

0.6

b

0.8 0.6

49 %

0.4

55% 43%

0.4

Base Rainfall RD30

21% 0.2 0

0.2 1

2

3

4

5

0

RI30 2

4

DO Concentration (mg/l)

6

8

10

12

14

16

18

AMM Concentration (mg/l)

Fig. 6. Empirical CDFs of water quality indicators in the river under climate change parameters.

respectively) are relatively large, i.e. closer to one than other values. The positive values obtained also indicate that increasing Qmaxout lead to increasing DO values, i.e. improved system performance. The reason for this is that the increasing Qmaxout is leading to reduced CSOs. Fig. 5a and b show the relationship between Qmaxout and DO under RD30 and RI30. In these figures, with increasing Qmaxout, the number of samples which have DO above the threshold of 4 mg/l (dashed line) is reduced by a changing climate and it is worse under RD30. Therefore, it is difficult to maintain the quality of DO in the river without increasing Qmaxout and consequently reducing CSOs. Regarding the AMM (see Table 4), the maximum correlation coefficients are observed for the PCW. In this study, the correlation between AMM and PCW is more significant than correlation to any of other parameter under the climate change. As can be observed in Fig. 5c and d, more non-behavioural samples are observed near the maximum value of PCW for AMM showing that the increase in PCW under the climate change has the potential to increase DWF and consequently leads to deterioration of water quality in the river. Fig. 6 shows the empirical CDF of water quality indicators under climate change based on the generated samples. Fig. 6a indicates that the probability of DO failure (i.e. DO below the 4 mg/l threshold) increases from 49% to 88% and nearly 99% assuming RI30 and RD30 respectively. Therefore, RD30 has a greater negative impact on DO than RI30 because of the increased CSO discharges into the river. The probability of AMM failure (i.e. AMM above the 4 mg/l threshold) increases with increasingly worsening climate. The probability of failure increases by 12% for the RI30 and 34% for the RD30 (see Fig. 6b). The AMM failure is a consequence of the increased CSO discharges i.e. discharges of untreated wastewater into the river under worsening climate conditions. 5. Conclusions This paper investigated the combined impact of urbanisation and climate change on the receiving water quality in rivers in the context of an integrated urban wastewater system in the UK, together with the associated potential for improved operational system control. The main conclusions can be summarised as follows. Based on the case study results obtained, it can be concluded that climate change is likely to lead to deterioration of water quality in rivers. The potential increase in rainfall depth is likely to have a more significant impact on the river water quality than the increase in rainfall intensity. The likelihood of breaching the 4 mg/l thresholds for ammonium and dissolved oxygen is likely to increase for both ammonium and dissolved oxygen and the breach for the latter is likely to be more frequent and significant than for the former. With regard to the urbanisation parameters, assuming the climate change, the per capita water consumption is the most

significant parameter. This is due to the fact that increased water consumption leads to increase dry weather flow and combined sewer overflows which, in turn, results in the increased ammonium concentration in the river. The per capita water consumption is also the most important urbanisation parameter influencing the dissolved oxygen in the river because it increases dry weather flow and consequently the ammonium concentration which leads to depletion of dissolved oxygen in the river. The maximum outflow rate from the sewer system is the most significant operational control parameter in terms of complying with the dissolved oxygen standard in the river under the climate change. The maximum inflow rate to the wastewater treatment plant is the most important operational control parameter in complying with the ammonium standard in the river. These two operational control parameters are the most significant ones as they affect the river water quality by controlling the discharge frequency of the combined sewer overflows and storm tank overflows. The results obtained provide a valuable insight into the key urbanisation and climate change parameters impacting on the receiving water quality in the integrated urban wastewater system context in the UK. This information can be further utilised to adapt future operation (and, if necessary, design) of these systems which, in turn, should make these systems more resilient to future climate and urbanisation changes. References Achileitner, S., Möderl, M., Rauch, W., 2007. Urine separation as part of a real-time control strategy. Urban Water J. 4 (4), 233e240. ATV (Abwassertechnische Vereinigung e.V.), 1992. Richtlinien für die Bemessung und Gestaltung von Regenentlastungsanlagen in Mischwasserkanälen. ATVArbeitsblatt A128, Gesellschaft zur Förderung der Abwassertechnik, St. Augustin. Butler, D., Davies, J.W., 2011. Urban Drainage, third ed. Spon Press, London. Butler, D., Schütze, M., 2005. Integrating simulation models with a view to optimal control of urban wastewater systems. Environ. Model. Softw. 20, 415e426. CIWEM, 2009. Integrated Urban Drainage Modelling Guide. URBAN DRAINAGE GROUP, WAPUG. http://www.ciwem.org/media/44495/WaPUG_IUD_ Modelling_Guide_Draft_Rev1_v28_(June_09)_v01e001.pdf. Copp, J.B., 2002. The COST Simulation Benchmark Description and Simulator Manual. Office for Official Publications of the European communities, Luxembourg, ISBN 92-894-1658-0. Cox, B.A., Whitehead, P.G., 2005. Parameter sensitivity and predictive uncertainty in a new water quality model Q2. J. Environ. Eng. ASCE 131, 147e157. Dessai, S., Hulme, M., 2007. Assessing the robustness of adaptation decisions to climate change uncertainties: a case study on water resources management in the East of England. Glob. Environ. Change 17, 59e72. Defra, 2006. BNWAT22: Domestic Water Consumption in Domestic and Nondomestic Properties. Defra’s Market Transformation Programme, UK. Delpla, I., Jung, A.V., Clement, M., Thomas, O., 2009. Impacts of climate change on surface water quality in relation to drinking water production. Environ. Int. 35, 1225e1233. Department for Communities and Local Government, 2010. Estimating Housing Need, ISBN 978-1-4098 26262. Environment Agency, 2006. Using Science to Create a Better Place. Environment Agency scenarios 2030. Science Report SC050002/SR1, ISBN: 1844325717. Fu, G., Butler, D., Khu, S.T., 2008. Multiple objective optimal control of integrated urban wastewater systems. Environ. Model. Softw. 23, 225e234.

M. Astaraie-Imani et al. / Journal of Environmental Management 112 (2012) 1e9 Fu, G., Butler, D., Khu, S.T., 2009. The impact of new developments on river water quality from an integrated system modelling perspective. Sci. Total Environ. 407, 1257e1267. FWR, 1998. Urban Pollution Management Manual, second ed. Foundation for Water Research, FR/CL0009, UK: Marlow. He, H., Zhou, J., Wu, Y., Zhang, W., Xie, X., 2008. Modelling the response of surface water quality to the urbanization in Xi’an, China. J. Environ. Manag. 86 (4), 731e749. Helton, J.C., 1993. Uncertainty and sensitivity analysis techniques for use in performance assessment for radioactive waste disposal. Reliab. Eng. Syst. Safe. 42, 327e367. Henze, M., Grady Jr., C.P.L., Gujer, W., Marais, GvR., Matsuo, T., 1986. Activated Sludge Model No.1. LAWQ Scientific and Technical Report 1. IAWQ, London. Hornberger, G.M., Spear, R.C., 1981. An approach to the preliminary analysis of environmental systems. J. Environ. Manag. 12, 7e18. Huber, W.C., Dickinson, R.E., 1988. Storm Water Management Model. User’s Manual Version IV. US Environmental Protection Agency. Hulme, M., Jenkins, G.J., Lu, X., Turnpenny, J.R., Mitchell, T.D., Jones, R.G., Lowe, J., Murphy, J.M., Hassell, D., Boorman, P., McDonald, R., Hill, S., 2002. Climate Change Scenarios for the United Kingdom: the UKCIP02 Scientific Report. Tyndall Centre for Climate Change Research, School of Environmental Sciences, University of East Anglia, Norwich, UK, 120 p. IFAK, 2005. SIMBA 5.0: Simulation of Wastewater Systems, User’s Guide. Institut für Automation und Kommunikation e.V., Magdeburg, Germany. IPCC, 2000. In: Nakicenovic, Neboisa, Swart, Rob (Eds.), Emission Scenarios: Special Report of Working Group III to the IPCC. Cambridge University Press, UK, 570 p. Jacobson, C.R., 2011. Identification and quantification of the hydrological impacts of imperviousness in urban catchments: a review. J. Environ. Manag. 92 (6), 1438e1448. Jefferies, J., 2005. Chapter 1: The UK Population: Past, Present and Future. Focus on People and Migration: 2005. Kallis, G., Butler, D., 2001. Implications of the EU water framework directive. Water Policy 3, 125e142. Lau, J., Butler, D., Schütze, M., 2002. Is combined sewer overflow spill frequency/ volume a good indicator of receiving water quality impact? Urban Water 4 (2), 181e189. Lessard, P., 1989. Operational Water Quality Management: Control of Stormwater Dischages. PhD thesis. Department of Civil Engineering; Imperial College of Science, Technology and Medicine, University of London, London. Lessard, P., Beck, M.B., 1993. Dynamic modelling of the activated sludge process: a case study. Water Resour. 27 (6), 963e978. Mannina, G., Viviani, G., 2010. An urban drainage stormwater quality model: model development and uncertainty quantification. J. Hydrol. 381, 248e265. McIntyre, N.R., Wagener, T., Wheater, H.S., Chapra, S.C., 2003. Risk-based modelling of surface water quality: a case study of the Charles River, Massachusetts. J. Hydrol. 274 (1e4), 225e247. Mimikou, M.A., Baltas, E., Varanou, E., Pantazis, K., 2000. Regional impacts of climate change on water resources quantity and quality indicators. J. Hydrol. 234, 95e109.

9

Office for National Statistics (ONS), 2010. UK Population Approaches 62 Million. http://www.statistics.gov.uk/pdfdir/popnr0610.pdf. Park, J.H., Duan, L., Kim, B., Mitchell, M.J., Shibata, H., 2010. Potential effects of climate change and variability on watershed biogeochemical processes and water quality in Northeast Asia. Environ. Int. 36, 212e225. Rauch, W., Bertrand-Krajewski, J.L., Krebs, P., Mark, O., Schilling, W., Schütze, M., Vanrolleghem, P.A., 2002. Mathematical modelling of integrated urban drainage systems. Water Sci. Tech. 45 (3), 81e94. Saltelli, A., Ratto, M., Tarantola, S., Campolongo, F., 2006. Sensitivity analysis practices: strategies for model-based inference. Reliab. Eng. Syst. Saf. 91 (10e11), 1109e1125. Saltelli, A., Chan, K., Scott, E.M., 2000. Sensitivity Analysis. Wiley Series in Probability and Statistics, ISBN 0-471-99892-3. Semadeni-Davies, A., Hernebring, C., Svensson, G., Gustafson, L.G., 2008a. The impacts of climate change and urbanization on drainage in Helsingborg, Sweden: suburban stormwater. J. Hydrol. 350, 114e125. Semadeni-Davies, A., Hernebring, C., Svensson, G., Gustafsson, L.G., 2008b. The impacts of climate change and urbanisation on drainage in Helsingborg, Sweden: combined sewer system. J. Hydrol. 350 (1e2), 100e113. Schütze, M., Butler, D., Beck, M.B., 2002. Modelling Simulation and Control of Urban Wastewater Systems. Springer, Berlin. Sin, G., Gernaey, K.V., Neumann, M.B., van Loosdrecht, M.C., Gujer, W., 2009. Uncertainty analysis in WWTP model applications: a critical discussion using an example from design. Water Res. 43, 2894e2906. Thorndahl, S., Beven, K.J., Jensen, J.B., Schaapur-Jensen, K., 2008. Event based uncertainty assessment in urban drainage modelling, applying the GLUE methodology. J. Hydrol. 357, 421e437. Tong, S.T.Y., Chen, W., 2002. Modelling the relationship between land use and surface water quality. J. Environ. Manag. 66 (4), 377e393. Tu, J., 2009. Combined impact of climate and land use changes on streamflow and water quality in eastern Massachusetts, USA. J. Hydrol. 379, 268e283. UKWIR, 2010. Impact of Urban Creep on Sewerage Systems. 10/WM/07/14, ISBN: 1 84057 558 1. Vanhooren, H., Meirlaen, J., Amerlinck, Y., Claeys, F., Vangheluwe, H., Vanrolleghem, P.A., 2003. WEST: modelling biological wastewater treatment. J. Hydroinform. 5 (1), 27e50. Vanrolleghem, P.A., Benedetti, L., Meirlaen, J., 2005. Modelling and real-time control of the integrated urban wastewater system. Environ. Model. Softw. 20 (4), 427e442. Wilby, R.L., Whitehead, P.G., Wade, A.J., Butterfield, D., Davis, R.J., Watts, G., 2006. Integrated modelling of climate change impacts on water resources and quality in a lowland catchment: river Kennet, UK. J. Hydrol. 330, 204e220. Whitehead, P.G., Wilby, R.L., Butterfield, D., Wade, A.J., 2006. Impacts of climate change on in-stream nitrogen in a lowland chalk stream: an appraisal of adaptation strategies. Sci. Total Environ. 365, 260e273. Zacharof, A.L., Butler, D., Schütze, M., Beck, M.B., 2004. Screening for real-time control potential of urban wastewater systems. J. Hydrol. 299 (3e4), 349e362. Zhou, J., Liu, Y., Chen, J., 2010. Accounting for uncertainty in evaluating water quality impacts of urban development plan. Environ. Impact Asses. Rev. 30, 219e228.