Department Urban Water Management

by Christoph Egger and Max Maurer

Predictions of the structural condition of urban drainage networks enables us to estimate future investment costs under various rehabilitation strategies. This facilitates proactive, far-sighted sewer asset management, which, in turn, contributes to a better balance between expenses and system performance. The latter is related to, among other things, the system’s ability to drain large quantities of stormwater under extreme precipitation with as little damage as possible. The design of system adaptions must therefore incorporate additional information on potential future urban land use as well as the changing precipitation regime.

Deterioration modelling of urban drainage networks

Successful calibration of sewer deterioration models is often hindered by the following factors,:

• Limited information content of the data, as in case of small sample sizes
• If sewer pipes have been renovated, repaired or replaced by new ones corresponding condition ratings prior the rehabilitation measure are often overwritten or discarded. In this case the remaining data distort the observed deterioration process. 

Detailed information on methods to overcome these issues can be found under the following link. See also (Egger et al., 2013).

http://apollo.emp-eaw.ch:9999/forschung/sww/gruppen/swip/eng_wastewater/2nd_revision_open_source_egger

Uncertainties as a result of climate change and variability

Hydraulic sewer system design under conditions of an changing climate is challenging due to the following reasons:

• Some climate projections suggest that the probability of heavy precipitation events will increase. However, the spacial and temporal resolution of these projections are too coarse to be directly used as input data for sewer system design.
• The currently used rain data series are too short and therefore not representative enough.

Current design practice relies on historical precipitation data. Even though observation periods cover usually several decades these data are not representative with respect to precipitation extremes. Stochastic models for rainfall simulation and disaggregation are appropriate tools to overcome these issues which enables us to take account of uncertainties resulting from climate change and variability. Further information on this can be found here. An investigation is currently underway that highlights the significance of these uncertainties – also with regard to additional investment costs for the deliberate incorporation of safety precautions into the systems.

Outlook

The methods described above are currently being applied to concrete cases under a variety of socio-economic scenarios. Thereby, we investigate the following issues:

• The predicted future investment costs depend strongly on the size and quality of the available condition data. Low data availability increases the significance of the prior knowledge and increase uncertainty. In these cases great care is needed to identify adequate prior information. This also highlights the importance of keeping historic data.
• How significant are the individual factors (i) the need for system adaptions in order to assure their serviceability in the future, (ii) sewer deterioration and (iii) urban development with regard to future investment needs? And in considering these, what are the uncertainties do we have to deal with?

Further information

• Aqua Urbanica
• Egger, C., Scheidegger, A., Reichert, P. and Maurer, M., 2013. Sewer deterioration modeling with condition data lacking historical records. Water Research 47 (17), 6762-6779. Postprint Publisher's Version

 by Lisa Scholten, Judit Lienert, Max Maurer

Our centralized water supply systems are aging. Especially small utilities (altogether servicing more than half the Swiss population; SVGW, 2009) often lack the institutional, financial, and personnel means to anticipate the long-term performance of the system and to proactively plan their water supply systems into an uncertain future. The aim of this sub-project thus is the development of approaches for

1. the prediction of the future condition of centralized pipe systems,
2. strategic rehabilitation planning, and
3. water infrastructure decision analysis that acknowledge uncertainties of predictions, future development, and stakeholder preferences.

Lifetime and failure prediction of water supply pipe networks

To assess the future condition of water supply networks, predictive pipe failure and pipe lifetime models are needed. The basis for these is the knowledge about expected pipe lifetimes and failure occurrence. In practice, failure and replacement data are often either absent or only cover short time windows, e.g. recording of the last few years. In addition, the amount of data available in small water networks does usually not suffice to robustly calibrate prediction models. This challenge is overcome in two ways:

1. Development of prediction models which take into account the absence of failure and replacement data (i.e. left censoring, right truncation, selective survival)
2. Combination of prior knowledge with locally available data to calibrate the models (Bayesian parameter inference).

The prior knowledge can be obtained from different sources. In this work, an approach to elicit pipe service life estimates from experts and its use for the calibration of pipe survival models was developed. This is presented in detail in Scholten et al. (2013a). Experts proved to be a useful source of knowledge which allowed to obtain differentiated estimates of the survival curves for different pipe groups (e.g. by material and vintage). The incorporation of in-between experts’ variance permitted to acknowledge different environmental and operational framework conditions, an important aspect in pipe survival modeling.

In another study, prior knowledge was obtained from recorded data of three mid-size to large water networks in Switzerland. This was combined with local data (by Bayesian parameter estimation) to calibrate a novel pipe failure model for a small water utility, see Scholten et al. (2013b). The model is presented in Scheidegger et al. (2013) where its ability to deal with the common data situation is also demonstrated. Similar modeling approaches were developed for sewerage Systems (Sub-project 2)

Strategic pipe rehabilitation planning

In light of pipe aging and high replacement values, pipe failure models are increasingly used to support water asset management decisions. Besides costs, different fundamental objectives play a role in these decisions. In this sub-project, the fundamental objectives of costs, reliability, and intergenerational equity were considered and traded against each other using multi-criteria decision analysis (MCDA) to define most desirable long-term rehabilitation strategies. This desirability depends on the long-term performance of a strategy and the preference of the decision maker(s) regarding these objectives.

Thereto, the failure model was combined with an existing strategic asset management software to model the outcomes of 18 rehabilitation strategies under four future scenarios for a small water utility. Model parameter uncertainty was propagated to the model outcomes and considered during evaluation. Different preferences were assumed to compare these alternatives. The analysis for a single case study utility revealed that, in this case, the common purely reactive rehabilitation strategy is not recommendable and that an annual replacement rate of 1–2% of the network by pipe condition could be a good long-term strategy. The ranking of alternatives differed the most under a development scenario with massive population and socio-economic growth (“Boom scenario”) as compared to the Status Quo or less extreme growth/ recession scenarios. For more details please see Scholten et al. (2013b).

Multi-criteria decision analysis under uncertainty – Good water supply infrastructure for the “Mönchaltorfer Aa”

To achieve a “good water supply infrastructure” in the long term, not only the technical pipe condition and pipe rehabilitation play a role. Together with stakeholders, 30 lower-level fundamental objectives for drinking water supply were identified by Lienert et al. (2014). These are quantified by 30 attributes. In addition, eleven water supply alternatives characterized by different organizational forms, geographical extent, management strategy, and technical configuration were developed and evaluated. Thereto, the outcomes of all alternatives regarding these attributes were predicted under four future scenarios to account for uncertainties about the future development. Based on a stakeholder and social network Analysis (, ten stakeholders were selected for individual MCDA interviews (Lienert et al., 2013). Their preferences were elicited and modeled using an approach which includes the imprecision of the stated preferences as well as uncertainties of preference parameters which were not elicited (the aggregation model, marginal value functions, risk attitude, and scaling factors). This finally resulted in obtaining probability distributions of the ranking of alternatives for each single stakeholder. Future dynamics and uncertainties could be incorporated by combining decision making and modeling with scenario planning, besides the quantitative consideration of uncertainties in making predictions, and evaluating the results. The results are currently prepared for publication. Sub-project 1)

Research on different preference elicitation formats is being conducted in another MCDA for identifying good wastewater infrastructures (sub-project 4)

References and more information

• Lienert, J., Schnetzer, F., Ingold, K. 2013. Stakeholder analysis combined with social network analysis provides fine-grained insights into water infrastructure planning processes. Journal of Environmental Management 125: 134–148. Postprint / Publisher's version
• Lienert, J., Scholten, L., Egger, C., Maurer, M. (2014) Structured decision making for sustainable water infrastructure planning and four future scenarios. EURO Journal on Decision Processes, special issue on Environmental Decision Making. Publisher's version

• SVGW, 2009. Statistische Erhebungen der Wasserversorgungen in der Schweiz Betriebsjahr 2008. Zürich, Schweizer Verein des Gas- und Wasserfaches. Link
• Scholten, L., Scheidegger, A., Reichert, P., Maurer, M., 2013a. Combining expert knowledge and local data for improved service life modeling of water supply networks. Environmental Modelling & Software 42 1-16. Postprint / Publisher's version
• Scholten, L., Scheidegger, A., Reichert, P., Maurer, M., Lienert, J., 2013b. Strategic rehabilitation planning of piped water networks using multi-criteria decision analysis. Water Research 49: 124-143. Postprint / Publisher's version
• Scheidegger, A., Scholten, L., Maurer, M., Reichert, P., 2013. Extension of pipe failure models to consider the absence of data from replaced pipes. Water Research 47(11) 3696-3705. Postprint / Publisher's version
• Scholten, L., Schuwirth. N., Reichert, P., Lienert, J. Tackling uncertainties in multi-criteria decision analysis - An application to water supply infrastructure planning. European Journal of Operational Research. In press. View at publisher | Postprint.

by Jun Zheng, Judit Lienert

The importance of incorporating stakeholders in public and environmental decisions is increasingly recognized. Involving stakeholders allows them to see the broader picture of the decision problem and move beyond single-interest concerns, which ultimately helps to reduce potential conflicts. The transparency of the process enhances the acceptance of the final recommendation, thus resulting in more willingness to implement it.

With help of the stakeholder and social network analysis (Sub-project1, Lienert et al., 2013), ten stakeholders were identified, who are influential on wastewater infrastructure planning or who might be affected by the decision. These stakeholders were deeply involved in the whole decision analysis process, from the initial problem structuring phase to the final recommendation. The focus of this sub-project is to quantify the stakeholders’ preferences which are to be integrated in a Multi-Attribute Utility model by its preferential parameters.

Aim

The aim of this sub-project was to elicit each stakeholder’s preferences in a quantitative way. A preference elicitation process consists of a sequence of questions and answers in which stakeholders progressively express their preferences. Preference elicitation is a rather challenging task of MCDA. There are various reasons for these difficulties:

1. The decision problem is relevant to the stakeholders, but they do not necessarily have adequate knowledge to understand the objectives and attributes. In some cases, especially the attributes (= indicators, performance measures) are inevitably technical. Often, stakeholders may only have expertise concerning a specific field (e.g. cost of wastewater infrastructure or management aspects), but do not know much about other topics (e.g. removal of micropollutants in wastewater).
2. The trade-offs themselves can be very difficult, as they may invoke moral, inner conflicts. Trade-offs have to be made if not all objectives can be reached at the same time. For instance, it may not be possible to remove all micropollutants from wastewater and to achieve the lowest-possible costs as well. In the preference elicitation process, decision makers are explicitly asked to make such difficult trade-offs.
3. Quantifying the preferences is per se not easy. Many people find it difficult to clearly express their preferences, especially if they have not been able to think about them thoroughly so far. If the preferences are not clear to a person, they are constructed during the elicitation process, in a numeric way.
4. We aimed at designing a structured elicitation procedure which enables the stakeholders to express their preferences with as much support as possible (Figure 1). At the same time we aimed at applying a methodologically correct and meaningful procedure from an MCDA point of view. We are also interested in the potential influence of different elicitation methods on the elicitation process.

Methods

Generally speaking, there are two paradigms of elicitation methods (Figure 2). Direct elicitation approaches directly ask the stakeholders for the preferential parameters so that a (mathematical) preference model can be constructed. On the other hand, indirect elicitation approaches deduce an aggregation model from a subset of alternatives, or in the language of artificial intelligence, to learn a preference model on the basis of a learning set.

The two different elicitation methods may trigger different activities in people’s mind, and ultimately reveal various aspects of preference information. We employed both approaches to elicit preferences from the same stakeholders, in order to test their possible influences on the elicitation results. We were also interested in investigating the stability of the stakeholders’ preferences.

Direct elicitation of preferential parameters was implemented by an online questionnaire and face-to-face interviews. The weights of objectives and attributes, the value functions, risk attitude, and aggregation methods were elicited from individual stakeholders. Uncertainties were dealt with by explicitly allowing stakeholders to answer the questions with a range rather than a precise number, or by requesting them to state an uncertainty level. We were also concerned about the common biases known from the literature, such as the “splitting bias” and “goal-directed bias”, and developed corresponding remedies to avoid them as much as possible.

The indirect elicitation approach was also realized by face-to-face interviews. The questions asked comprised pairwise comparisons of some hypothetical alternatives, the strength of preferences, and trade-off questions. Imprecise preference models are inferred from all this information by solving linear program problems.

In both methods, the stakeholders were encouraged to “think aloud” in the interviews, i.e. they were asked to explain the rationale when they made judgments. Furthermore, when changes of answers were detected between the interviews, a discussion followed to understand whether the change was caused by the different elicitation method, or by a real change of preferences.

Main results

As the project is still ongoing there are no final results yet.

More information

• Lienert, J., Schnetzer, F., Ingold, K. (2013) Stakeholder analysis combined with social network analysis provides fine-grained insights into water infrastructure planning processes. Journal of Environmental Management 125: 134–148. Postprint / Publisher's Version
• Lienert, J., Scholten, L., Egger, C., Maurer, M. (2014) Structured decision making for sustainable water infrastructure planning and four future scenarios. EURO Journal on Decision Processes, special issue on Environmental Decison Making: in press.
• Zheng, J., Egger, C., Lienert, J. (2014a) Multi-criteria decision analysis for wastewater infrastructure planning incorporating stakeholders’ preferences (working title, in preparation).
• Zheng, J., Lienert, J. (2014b) Stakeholder preference elicitation and modeling for wastewater infrastructure planning: a comparative evaluation of two elicitation methods (working title, in preparation).