A decision support system (DSS) tool for the assessment of intervention strategies in an urban water system (UWS) with an integral simulation model called ‘WaterMet2’ is presented. Lists of intervention options and performance indicators are exposed by the DSS for the user to define intervention strategies and metrics for their comparison. The quantitative and risk-based metrics are calculated by WaterMet2 and risk modules, while the qualitative metrics may be quantified by external tools feeding into the DSS. Finally, a multi-criteria decision analysis approach is employed in the DSS to compare the defined intervention strategies and rank them with respect to a pre-specified weighting scheme for different scenarios. This mechanism provides a useful tool for decision makers to compare different strategies for the planning of UWS with respect to multiple scenarios. The suggested DSS is demonstrated through the application to a northern European real-life case study.

INTRODUCTION

Urban water systems (UWS) face the long-term perspective of constraints and challenges associated with climate change and the availability of natural resources. This prospect requires the adaptation of the operation and infrastructure of UWS to meet uncertain future scenarios through the adoption of mitigating technologies in the water industry. However, it is suggested that the impact on the UWS of these technologies, prior to their practical implementation, is best evaluated by a decision support system (DSS). This approach has attracted attention by practitioners and researchers in recent years, leading to the development of tools such as AQUACYCLE (Mitchell et al. 2001), UWOT (Makropoulos et al. 2008), UVQ (Mitchell & Diaper 2010), City Water Balance (Mackay & Last 2010) and WaterMet2 (Behzadian & Kapelan 2015a). Despite a plethora of DSS being developed in recent years, relating to the integrated modelling of UWS, there remain outstanding issues which need to be addressed in this framework. The principal concern relates to simultaneously covering the whole range of sustainability dimensions in the performance indicators (PIs), including both quantitative and risk-based ones. Ideally, the PIs should reference all facets of sustainability including social, environment, economic, governance and assets (Alegre et al. 2012).

This paper presents a DSS which implements a tool which is able to quantify the impact of different sets of interventions/technologies on the performance of the UWS, including associated risks and costs, by evaluating a wide variety of sustainability PIs under different scenarios. The WaterMet2 model (Behzadian et al. 2014), which undertakes the simulation of the integrated modelling of UWS, is employed in the DSS presented. In the following section, a brief description of the DSS configuration is followed by a review of WaterMet2. The principal stages of the DSS are mapped through four steps including (1) problem definition, (2) metric calculation/decision matrix population, (3) ranking and (4) result viewing/modification/re-evaluation. The capabilities of the developed DSS are demonstrated on a real-life UWS in northern Europe. By way of the real case study, the paper presents a walk-through for each stage, presenting a list of the scenarios, intervention strategies and metrics used. The values obtained after running the WaterMet2 model and the risk module are shown, along with how those outputs are used in the population of the multi-criteria decision analysis (MCDA) decision matrices.

DSS METHODOLOGY

The DSS developed seeks to support long-term, strategic-level planning of UWS at the city/system level. This is achieved through a new methodology for comparison and selection of alternative solutions, within the framework of long-term transition paths, and amidst multiple decision criteria (Morley et al. 2012). The support offered to the decision maker takes several forms and guides the user through the description of alternative intervention options and scenarios. The structure of the classes in the DSS engine is split into three principal modules (Figure 1): (1) ‘Environment’, (2) ‘Performance’ and (3) ‘MCDA’. These modules are described below in further detail.
Figure 1

Flowchart of the suggested DSS framework.

Figure 1

Flowchart of the suggested DSS framework.

Environment module

The ‘Environment’ part manages the specifications of the analysis, including timing, intervention strategies, PIs, scenarios and customised model input. Basically, this module assists the user in defining the Environment configuration – i.e. the outline definition of the problem to be analysed. More specifically, the following steps are taken in this module:

  • Defining a time horizon for the analysis, along with the intermediate times at which Interventions may take place.

  • Defining scenarios which comprise varying input parameters to the analysis tool (i.e. WaterMet2 model) or to custom metrics defined outside of WaterMet2. Note that analysis of the UWS over some planning horizon in the DSS is the basis of a pre-specified scenario. Each scenario can influence a number of specific variables in WaterMet2.

  • Selecting the performance metrics of interest to be used for evaluation from the list of available PIs, including those supported by the WaterMet2 model and those supported by other tools outside the DSS.

  • Defining intervention strategies based on the list of available intervention options. Each intervention strategy comprises a set of individual interventions, including technologies and their operation on different parts of the UWS, each of which is assumed to occur at a specific time over a defined planning horizon. The DSS supports an existing library of individual intervention options quantified by WaterMet2 based on different components in the UWS.

Performance module

The ‘Performance’ part undertakes the responsibility of evaluating the performance metrics, which are split into two categories: (1) performance assessment of quantitative metrics, including non-risk-based and risk-based metrics calculated by the WaterMet2 and Risk section, respectively; (2) qualitative metrics of the aforementioned types, defined within the DSS and quantified by external tools outside the immediate scope of the DSS. The PIs calculated or supported by the WaterMet2 directly, such as risk-based metrics, are automatically populated in the DSS, whilst others evaluated outside the DSS need to be supplied manually by the user. Through repeated execution of the WaterMet2 model, each intervention strategy is evaluated to determine its effect on urban water cycle system performance. This is achieved by, firstly, applying each scenario defined in the Environment configuration in turn and also applying each intervention strategy in turn – at the appropriate time step. This process results in a series of metric values, for each time step and scenario, representing the performance of the system.

As a part of the built-in simulation model in the DSS, the WaterMet2 model is used to calculate all non-risk-based performance metrics and support risk-based performance metrics (Behzadian & Kapelan 2015b). In fact, the WaterMet2 model is used in the DSS for metabolism simulation of UWS for a long-term planning horizon. WaterMet2 is a conceptual and metabolism-based model which simulates water-related and other resource flows (e.g. energy, materials and environmental categories) throughout the UWS components (Behzadian & Kapelan 2015a). WaterMet2 is able to model three main subsystems of UWS (i.e. water supply, stormwater and wastewater) in different spatial (i.e. indoor, local, subcatchment and system area levels) and temporal (i.e. daily, weekly, monthly and annual) scales. WaterMet2 adopts a daily mass-balance-based approach for modelling water flows and other fluxes. Once water-related flows are tracked down for each UWS component by simulating the integrated UWS in each time step, other fluxes (e.g. greenhouse gas (GHG) emissions) can be quantified in the UWS spatially and temporally based on associated impact coefficients. Thus, the physical metabolism of this integrated UWS is quantified through some PIs using the WaterMet2 model. Further details of the principal flows and storages modelled in WaterMet2, as well as descriptions of the components and their functionality, can be found in Behzadian et al. (2014).

The risk assessment is calculated based on the likelihood of occurrence and severity of consequences. The likelihood is assumed here as the probability of the scenario under analysis and is scaled in five levels, each associated with a specified probability range (Table 1). The likelihood scale needs to be as objective as possible. The probability of a risk event is assumed here to follow corresponding scenarios. For example, the probability of a risk event related to population growth is simplified as equivalent to the probability of the scenario of population growth. The scenario of high and low population growth was selected here to illustrate the methodology. For example, it is assumed that the probability of exceeding the upper limit of the 90% confidence interval is equal to 5% ([100%–90%]/2). This means that there is a 5% probability that the population will follow the strong growth or higher, and a probability of between 1 and 2% of being lower than the lower profile. Therefore, assuming Table 1 as reference scale, the probability level is set to 4 ‘likely’ and 3 ‘moderate’, respectively.

Table 1

Risk matrix for quantifying risk-based metrics

    Consequence level
 
 Probability range 
Likelihood level 
 5 Almost certain P > 10% 5E–Med. 5D–Med. 5C–High 5B–High 5A–High 
 4 Likely 2% < P ≤ 10% 4E–Low 4D–Med. 4C–Med. 4B–High 4A–High 
 3 Moderate 1% < P ≤ 2% 3E–Low 3D–Med. 3C–Med. 3B–Med. 3A–High 
 2 Unlikely 0.2% < P ≤ 1% 2E–Low 2D–Low 2C–Med. 2B–Med. 2A–Med. 
 1 Rare P ≤ 0.2% 1E–Low 1D–Low 1C–Low 1B–Low 1A–Low 
    Consequence level
 
 Probability range 
Likelihood level 
 5 Almost certain P > 10% 5E–Med. 5D–Med. 5C–High 5B–High 5A–High 
 4 Likely 2% < P ≤ 10% 4E–Low 4D–Med. 4C–Med. 4B–High 4A–High 
 3 Moderate 1% < P ≤ 2% 3E–Low 3D–Med. 3C–Med. 3B–Med. 3A–High 
 2 Unlikely 0.2% < P ≤ 1% 2E–Low 2D–Low 2C–Med. 2B–Med. 2A–Med. 
 1 Rare P ≤ 0.2% 1E–Low 1D–Low 1C–Low 1B–Low 1A–Low 

Considering that consequences are established as deviations from the sustainability objectives, with corresponding criteria, metrics and targets, the consequence scale consists of levels defined by ranges of deviations from the set targets. Consequences are defined here as five levels (A–E) of deviations of absolute value of risk event from a specified sustainability target value (Table 1). The absolute value of the consequences is estimated based on the PIs obtained from the UWS simulation in the WaterMet2 model. Note that the level of deviations for each metric needs to be converted to the summary scale as well (i.e. from A to E, in which A is the most severe consequence level and E is the least severe). Finally, the risk level can be estimated based on the assessment of likelihood and consequence levels for each event using a selected risk matrix, as shown in Table 1. The risk matrix was built in collaboration with the water utility, comparing the levels of risk estimated during the risk analysis with the risk criteria established. The criteria for risk evaluation have been described and specified for each level of risk, it being usual to consider that the levels of low, medium and high risk are acceptable, tolerable and intolerable, respectively. The levels for each combination of probability and consequence levels have been selected for the metrics here presented, resulting in the matrix of Table 1.

MCDA module

Having created two or more intervention strategies, the principal role of the DSS is to undertake an automatic ranking of the strategies using a MCDA technique. The MCDA module applies a user-configured ranking approach to the specified intervention strategies for the purposes of scoring and ranking them for each scenario and user preference combination. Two well-known MCDA techniques are implemented in the DSS for the purpose of ranking intervention strategies: compromise programming (CP) (Zeleny, 1973) and the analytic hierarchy process (AHP) (Saaty, 1980), although the design does not preclude adding other techniques, including optimization. The two methods were selected because of their widespread use but also because they use different ranking technologies and, also, allow users to express their preferences in a different way. In the CP method, user preferences are specified as multiple evaluation criteria weights, making this method more suitable for use by less experienced users. In the AHP method, user preferences are specified via the pairwise criteria-importance comparisons. This requires more experience to configure and employ the method. The DSS will enable the user to select the method to use when solving a particular problem, including the possibility of using both methods on the same problem and then comparing results (e.g. to see if there is an alternative solution that is ranked highly regardless of the MCDA method used). Further details of these techniques can be found in Behzadian et al. (2015).

The ranking is performed according to the Metrics that have been identified in the Environment configuration, and is repeated for each combination of scenario and user preferences defined. Following the ranking, the decision maker is supported in interactively modifying the intervention strategies and submitting them for the evaluation and rankings to be revised. Any number of intervention strategies can be created by the DSS, and existing Strategies can be cloned and modified to assist in ‘what-if?’ analysis, allowing variations of Strategies to be analysed in a straightforward fashion to investigate their influence on the overall strategy rankings.

CASE STUDY

Introduction

The UWS of a northern European city is used here as a reference city for the case study, combined with assumptions when necessary. The existing UWS comprises two main surface water resources connected to corresponding water treatment works (WTW), service reservoirs, distribution mains and eventually water consumption points. The existing sewer network represents a mix of combined and separate sewers which are connected to two wastewater treatment works (WWTWs). Treated wastewater from WWTWs and untreated wastewater/stormwater from combined sewer overflows of sewer networks are discharged into a single receiving water body (i.e. downstream sea). The DSS is demonstrated here for conditions of likely future population growth. Hence, increased water demand as a result of highest foreseen population growth is the key driver which is likely to impose significant strains on the UWS performance. In the first instance, the DSS needs to have specified scenarios, intervention strategies and metrics and associated target/goals and preferences, described in the following sections.

Scenarios

In order to demonstrate the efficacy of the DSS, the following two scenarios related to high (Scenario 1) and low (Scenario 2) population growth are considered in this case study. In this instance, the WaterMet2 parameters changed in these population growth scenarios are the different water demand categories (i.e. household/population growth, industrial/commercial growth and irrigation growth).

Intervention strategies

Three types of intervention options are employed in this case study:

  1. Addition of a new water resource along with two WTWs.

  2. Increased annual rehabilitation rate for pipes.

  3. Addition of rainwater harvesting (RWH) and grey water recycling (GWR) schemes.

Based on the above individual intervention options, the metabolism model is analysed in this demonstration based on the following seven alternative UWS intervention strategies against a 30 year planning horizon (2011–2040). Note that the intervention strategies numbered 3 to 7 start from 2015.

  1. Business as usual.

  2. Addition of a new water resource along with two WTWs starting from 2020.

  3. One percent additional annual pipe rehabilitation starting from 2015.

  4. Addition of RWH and GWR systems at a local level by 25% of households starting from 2015.

  5. Addition of RWH and GWR systems at a local level by 25% and 50% of households, respectively, starting from 2015.

  6. Addition of RWH and GWR systems at a local level by 50% of households starting from 2015.

  7. Addition of RWH and GWR systems at a local level by 25% of households and 0.5% additional rehabilitation annually starting from 2015.

Performance metrics

Six metrics according to the performance criteria of sustainability dimensions of water systems (Alegre et al. 2012) are considered for the purposes of this case study. These metrics include three quantitative criteria (C1–C3), two quantitative risk-based criteria (C4, C5) and a single qualitative example. The quantitative metrics are directly calculated by WaterMet2 and risk modules, respectively. The qualitative metric (C6) is quantified by relevant experts and the quantified values incorporated in the DSS. Instead of using qualitative categories (linguistic terms) for metric C6, these are rated as scoring on a scale of acceptance ranging from 1 to 10, being: extremely low (1–2), low (3–4), medium (5–6), high (7–8) and extremely high (9–10). Furthermore, for the risk-based metrics, failure times shorter than the time step in the simulation model (i.e. daily in the WaterMet2 metabolism model) cannot be captured by the DSS. A brief description of these metrics is outlined below:

  1. Reliability of water supply (C1): the ratio of water delivered to customers to the total water demand.

  2. Total cost (C2): annual average of the discounted initial capital investment of interventions plus discounted value of the fixed and variable costs in different UWS components to the first year with a specific discount rate.

  3. GHG emissions (C3): annual average of the aggregated greenhouse gas emissions, as Global Warming Potential (GWP100) measured in units of carbon dioxide equivalents (CO2-eq) from all components of the UWS.

  4. Days with restrictions to water service (C4): the risk of the annual days with restriction (water supply failure) being greater than the target value.

  5. Prolonged hydraulic failure (C5): the risk of annual expected value for the time length of hydraulic failure being greater than a target value.

  6. Social acceptance (C6): the extent to which an intervention strategy would be supported by society, especially water consumers, in order to fulfil the water demands with respect to a number of factors, especially safety and health issues.

RESULTS AND DISCUSSION

The results are presented in the following two parts: (1) calculation of the quantitative and risk-based metrics for each intervention strategy; (2) ranking the intervention strategies using MCDA. The expert-quantified values for the single qualitative metric are directly populated in the decision matrix.

The time-series of the quantitative metrics (C1–C3) over the planning horizon are calculated by the DSS by running the WaterMet2 model with respect to each scenario and intervention strategy. The single value for each of these metrics is calculated and populated in Table 2 for each of the two scenarios.

Table 2

MCDA decision matrix and rankings for Scenarios 1 and 2

Criteria/metrics
 
Reliability of water supply
 
Total cost
 
GHG emissions
 
Risk of restriction to service
 
Risk of hydraulic failure
 
Social acceptance
 
 
Units
 
%
 
M Euros/year
 
103 Tons/year
 

 

 

 
  
Weights
 
1.00
 
1.00
 
1.00
 
1.00
 
1.00
 
1.00
 
   
Goal Maximize Minimize Minimize Minimize Maximize Maximize Rank  
Scenario 
Strategy 1 94 99 53 52 95 89 
Strategy 2 100 100 74 72 99 90 
Strategy 3 96 100 58 57 96 89 
Strategy 4 98 100 62 61 90 83 
Strategy 5 98 100 63 62 89 82 
Strategy 6 99 100 71 69 89 81 
Strategy 7 98 100 64 63 90 83 
Criteria/metrics
 
Reliability of water supply
 
Total cost
 
GHG emissions
 
Risk of restriction to service
 
Risk of hydraulic failure
 
Social acceptance
 
 
Units
 
%
 
M Euros/year
 
103 Tons/year
 

 

 

 
  
Weights
 
1.00
 
1.00
 
1.00
 
1.00
 
1.00
 
1.00
 
   
Goal Maximize Minimize Minimize Minimize Maximize Maximize Rank  
Scenario 
Strategy 1 94 99 53 52 95 89 
Strategy 2 100 100 74 72 99 90 
Strategy 3 96 100 58 57 96 89 
Strategy 4 98 100 62 61 90 83 
Strategy 5 98 100 63 62 89 82 
Strategy 6 99 100 71 69 89 81 
Strategy 7 98 100 64 63 90 83 

The risk-based metrics (C4, C5) are calculated based on the following sequential steps (Ugarelli et al., 2014): (1) likelihood of risk event; (2) consequence levels from the PIs calculated by WaterMet2 for each scenario; (3) risk estimation. The likelihood of risk events is assumed to correspond to the probability of scenarios, i.e. 4 ‘likely’ and 3 ‘moderate’ for the high and low population rate scenarios (1 and 2), respectively. Assuming a target value of 1% and 100%, respectively, for risk events of water supply failure (C4) and prolonged hydraulic failure (C5), the consequence scales of deviation value in Table 1 are defined as follows, respectively: 
formula
Given the maximum value experienced being used to aggregate the risk-based metrics over the planning horizon, the consequence levels of risk events can be calculated as shown in Table 3. With the given likelihood and consequence levels, the risk is then estimated according to the risk matrix of Table 1 for each intervention strategy and scenario, the results of which are further illustrated in Table 3.
Table 3

Deviation values and risk estimation for risk-based metrics; H = high, M = medium, L = low

  Scenario 1
 
Scenario 2
 
  C4 [%]
 
C5 [%]
 
C4 [%]
 
C5 [%]
 
Strategy Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk 
99 100 97 99 
83 80 
99 100 77 97 
73 88 80 
55 88 83 
25 84 83 
60.4 88 90 
  Scenario 1
 
Scenario 2
 
  C4 [%]
 
C5 [%]
 
C4 [%]
 
C5 [%]
 
Strategy Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk Dev. Prob. Cons. Risk 
99 100 97 99 
83 80 
99 100 77 97 
73 88 80 
55 88 83 
25 84 83 
60.4 88 90 

Ranking results

The aforementioned metric values calculated for each intervention strategy are used to populate the corresponding MCDA decision matrix, as per Table 2, for each of the two scenarios, respectively. As the qualitative risk levels reported in Table 3 cannot directly be used for a quantitative comparison between the intervention strategies, they are rated on a scale between 1 and 3 as: high (3), medium (2) and low risk (1).

Following the population of the decision matrices, the ranking of intervention strategies is undertaken by means of the CP method (Zeleny 1973). The outputs of this ranking can be seen in the two right-most columns of Table 2. In this table, equal metric weights have been used to rank the strategies.

To further analyse the sensitivity of the ranking to the metric weights of the metrics, two further weighting schemes, including Water Company and Consumer perspectives, have been ranked by the MCDA (Table 4).

Table 4

Weights of the metrics from different perspectives

  Reliability of supply Total costs GHG emissions Risk of restriction to service Risk of hydraulic failure Social acceptance 
Equal weight 
Consumer 
Water company 
  Reliability of supply Total costs GHG emissions Risk of restriction to service Risk of hydraulic failure Social acceptance 
Equal weight 
Consumer 
Water company 

Given the three weighting schemes and two scenarios, a total of six groups of ranking for the intervention strategies are obtained, illustrated in Table 5. Naturally, there are several ways that these rankings can be merged together to achieve a final ranking for each intervention strategy. In this instance, the sum of the ranks of each strategy is used for determining final ranking, as shown in the last column for each scenario in Table 5.

Table 5

Summary of rankings of intervention strategies and final ranking

  Scenario 1
 
Scenario 2
 
Strategy Equal weight Consumer Water company Sum of rankings Final ranking Equal weight Consumer Water company Sum of rankings Final ranking 
18 20 
11 19 
16 11 
16 
14 
13 
  Scenario 1
 
Scenario 2
 
Strategy Equal weight Consumer Water company Sum of rankings Final ranking Equal weight Consumer Water company Sum of rankings Final ranking 
18 20 
11 19 
16 11 
16 
14 
13 

As can be seen, Strategy 2, which has been consistently ranked highly, is selected in the top strategy for both scenarios. However, it is further seen that if there is low population growth (Scenario 2), Strategy 7 is ranked first due to its consistent high rank when seen from all perspectives. Strategy 1 has the lowest final rank because it has been identified as the worst strategy for several scenario/weighting combinations. Therefore, while Strategies 2 and 7 are recommended as the best strategies to adopt in this simple example, Strategy 1 is clearly not to be recommended. However, further analysis will be required to fully cover and test different criteria for these strategies.

CONCLUSIONS

A new DSS was developed to facilitate decision-making for the long-term city metabolism planning problem. The new DSS methodology comprises creation of alternative intervention strategies within the framework of long-term transition paths and then comparison and selection of strategies accommodating multiple decision criteria. The DSS is able to deal with uncertain future scenarios and differing stakeholder perspectives. The DSS is able to employ a metabolism based model (WaterMet2) and risk modules to evaluate quantitative metrics of intervention strategies in an UWS over a long-term planning horizon under different scenarios (e.g. population growth). The levels of detail required for modelling and simulating alternative intervention strategies are less complex than similar, physically based models and the previously developed conceptually based models; hence the UWS itself and the relevant intervention strategies can be easily modelled. For each intervention strategy in the decision matrix, the DSS can also include qualitative metrics (e.g. social acceptability) quantified by experts outside the DSS. The DSS is able to rank intervention strategies using the CP MCDA method operating over different weighting schemes, allowing the consideration of the case study from different stakeholder perspectives. The strategies which are consistently ranked highly are identified as being the likely appropriate strategies to be implemented. These strategies can be considered for further analysis, such as inclusion of other metrics. The results obtained in the case study demonstrate how the DSS developed and presented here can be used to effectively and efficiently assist planners in making better, more objective and strategic level decisions with respect to meeting the future long-term goals and performance targets in their UWS/city.

ACKNOWLEDGEMENTS

This work was carried out as part of the ‘TRansition to Urban water Services of Tomorrow’ (TRUST) project funded in the EU 7th Framework Programme under Grant Agreement No. 265122. The authors also wish to thank NTNU, Oslo VAV, LNEC and Addition (TRUST project partners) for their collaboration. Readers interested in receiving the DSS and WaterMet2 model for research purposes may contact the authors.

REFERENCES

REFERENCES
Alegre
H.
Cabrera
E.
Jr
Hein
A.
Brattebø
H.
2012
Framework for Sustainability Assessment of UWCS and Development of a Self-assessment Tool. Deliverable D31.1. TRUST Project
.
Behzadian
K.
Kapelan
Z.
2015a
Modelling metabolism based performance of an urban water system using WaterMet2
.
Resources, Conservation and Recycling
99
,
84
99
.
Behzadian
K.
Kapelan
Z.
Venkatesh
G.
Brattebø
H.
Sægrov
S.
2014
WaterMet2: a tool for integrated analysis of sustainability-based performance of urban water systems
,
Drink. Water Engineering Science Discussion
7
,
1
26
.
Behzadian
K.
Morley
M. S.
Vitorino
D.
Coehlo
S.
Ugarelli
R.
Kapelan
Z.
2015
Final report with DSS methodology, software and case study from a pilot city, Deliverable 54.3. TRUST Project
.
Mackay
R.
Last
E.
2010
SWITCH city water balance: a scoping model for integrated urban water management
.
Reviews in Environmental Science and Bio/Technology
9
(
4
),
291
296
.
Makropoulos
C. K.
Natsis
K.
Liu
S.
Mittas
K.
Butler
D.
2008
Decision support for sustainable option selection in integrated urban water management
.
Environmental Modelling & Software
23
(
12
),
1448
1460
.
Mitchell
V. G.
Diaper
C.
2010
UVQ User Manual (urban water balance and contaminant balance analysis tool), Version 1.2, CMIT Report No. 2005–282. CSIRO
.
Mitchell
V. G.
Mein
R. G.
McMahon
T. A.
2001
Modelling the urban water cycle
.
Environmental Modelling & Software
16
(
7
),
615
629
.
Morley
M. S.
Kapelan
Z.
Savić
D. A.
2012
Integrated Decision Support Framework. Deliverable D54.1. TRUST Project
.
Saaty
T. L.
1980
The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation
.
McGraw-Hill
,
New York
,
USA
.
287
pp.
Ugarelli
R.
Ceu Almeida
M.
Behzadian
K.
Liserra
T.
Smeets
P.
Kapelan
Z.
Sægrov
S.
2014
Sustainability risk based assessment of integrated urban water system
. In:
11th International Conference on Hydroinformatics
,
HIC 2014
,
New York City
.
Zeleny
M.
1973
Compromise programming
. In:
Multiple Criteria Decision Making
(
Cochrane
J.
Zeleny
M.
, eds).
University of South Carolina Press
,
Columbia, South Carolina
.