Group decision making in infrastructure safety planning

Safety Science 42 (2004) 325–349 www.elsevier.com/locate/ssci

Group decision making in infrastructure safety planning Nils Rosmullera,*,1, Giampiero E.G. Beroggib,1 a

Dutch Institute for Fire Service and Disaster Management, PO Box 7010, 6801 HA, Arnhem, The Netherlands b Swiss Distance University of Applied Sciences, PO Box 689, 3900 Brig, Switzerland Accepted 7 July 2003

Abstract Infrastructure planning involves a multitude of concerns, where safety considerations generally range behind economic issues. We hypothesize that safety issues are insufficiently considered in infrastructure planning due to the lack of a shared view among the different safety experts, and that a carefully designed participatory group decision making method can support safety experts in reaching a shared view on the problem. To test our hypothesis, we developed a participatory methodology that helps infrastructure providers, spatial planners and emergency responders converge their views on safety in infrastructure planning. The methodology integrates dynamically and interactively risk and deterministic safety analysis and deliberation processes; the methodology has been integrated in a mobile multimedia group decision network system. We discuss the application of this method for the Northeastern Connection of the Betuweline: a freight railway linking Rotterdam Harbor to the German Ruhrgebiet. Structured questionnaires, audio tapes, observations and interviews with the participants confirm the method’s merit in reaching a shared view and consensus for safety aspects in infrastructure planning. The results of this study provide evidence for our hypothesis and indicate ways for how to consider more efficiently and effectively safety issues in infrastructure planning. # 2003 Elsevier Ltd. All rights reserved. Keywords: Infrastructure planning; Probabilistic safety assessment; Group decision making; Emergency response

* Corresponding author. Tel.: +31-26-3764127; fax: +31-26-3764144. E-mail address: [email protected] (N. Rosmuller). 1 The research was conducted during the authors’ affiliation with Delft University of Technology, Faculty of Technology, Policy and Management. 0925-7535/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0925-7535(03)00046-8

326

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

1. Introduction The Dutch Ministry of Transport, Public Works and Water Management decided in 1991 to sharpen its road-safety policy to reduce the death toll by 50% by the year 2010. This policy is referred to as ‘‘intrinsic safety’’. A major pillar of the intrinsic safety policy is to pay more attention to an integral safety approach, by involving decision makers also outside the area of road safety. Van Uden and Heijkamp (1995) point out that one of the major issues of intrinsic safety is to make sure that infrastructure decisions will be most effectual with respect to safety. To arrive at safety-effective infrastructure decisions, it will be indispensable that the different safety experts arrive at a shared view of how they perceive safety, before coordinating their safety objectives with objectives of other stakeholders involved in infrastructure planning, who are mainly focusing on economic aspects. The necessity and means of how to arrive at a shared view about safety is still subject of research. Various models of how to configure participative policy making have been proposed, with the Vroom–Yetton model as the most prevalent one (Daniels et al., 1996). Depending on (1) quality requirements, (2) amount of available information, (3) structure of the problem, (4) expected public acceptance, (5) decision competence of the Ministry of Transport, (6) goals, and (7) expected conflict, public involvement must be considered to different degrees in the decision making process. An overview of important issues in citizen participation can be found in Renn et al. (1995). The issues refer to fairness, discourse techniques, problems of legitimization, citizen juries, regulatory negotiation, environmental mediation, voluntary siting of systems and compensation, and direct participation. Mumpower (2001) presents an overview of selecting and evaluating tools and methods for public participation. Based on the work of these authors, a participatory approach to safety assessment was proposed by Beroggi (1999a). 1.1. Hypothesis Basically all proposals on how to consider safety issues more effectively in participatory decision making for infrastructure planning have raised several questions about the process and the indicators that should be considered. These concerns are summarized in the following hypothesis. H. An integral approach comprising multiple safety indicators and stakeholders contributes to a shared view of safety aspects for alternative line infrastructure plans. The motivation of this research was the need to develop a methodology that supports different safety experts in developing a shared view on safety for transportation infrastructure planning and decision making. The core of the methodology consists of a discursive aggregation mechanism, which supports safety experts to arrive at a shared view of safety in transportation infrastructure planning. This

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

327

approach will be applied to a real life infrastructure project in the Netherlands to test the above-formulated hypothesis . The reminder of the paper is organized as follows. In Section 2 we discuss existing preference aggregation models and their limitations, from which we derive a participatory approach of generating a shared view among different safety experts in Section 3. In Section 4 we introduce the infrastructure project in the Netherlands, which we used to test our hypothesis. In Section 5, we present the analysis of a one-day session conducted with real safety experts in the field, followed by conclusions, which are presented in Section 6.

2. Participatory infrastructure planning 2.1. General considerations Preference aggregation in participatory infrastructure planning entails two aspects. The first aspect refers to the procedural mechanisms focusing on the participants and the structure of the process (Renn et al., 1995; Daniels et al., 1996; Beroggi, 1999a; Mumpower, 2001). The second aspect refers to the computational preference aggregation models. A variety of preference aggregation mechanisms for expert assessments in risk management have been proposed in the literature (Cooke, 1991; Beinat et al., 1994; DeWispelare et al., 1995; Sandri et al., 1995; Myung et al., 1996; Beroggi and Wallace, 2000). They can be further classified as behavioral and mechanical procedures (DeWispelare et al., 1995). Behavioral procedures are based on either structured discussions or on non-interactive approaches such as the DELPHI Method (Linstone and Turoff, 1975). Mechanical procedures consist of mathematical formulae and algorithms, of which a large body of literature has emerged over the past two decades (Cooke, 1991). Apostolakis and his colleagues conducted a 2-year study focusing on analytic/ deliberative decision-making processes in environmental restoration decisions that involve multiple stakeholders, multiple objectives, and multiple alternatives (Zio and Apostolakis, 1998; Apostolakis and Picket, 1998; Accorsi et al., 1999; Bonano et al., 2000). Following the analytical/deliberative decision-making process as described in a report of the National Research Council of the USA (NRC, 1996), the prime aspect is to involve stakeholders throughout the whole process of decision making. Rankings of the alternatives are generated for each stakeholder, using a utility theoretic approach, where the assessment of the weights of the indicators, which was done by each stakeholder, is based on an analytical hierarchy processing (AHP). All stakeholders evaluate alternatives based upon all objectives. They do not aggregate evaluations in a way a ranking originates for the group stakeholders. Instead, they try to rank alternatives for the group stakeholders by discussing the individual rankings. The aggregation of ordinal and cardinal preferences across multiple decision makers is known to be critical. Arrow (1951) showed that there is no aggregation mechanism for aggregating at least two experts’ ordinal preference rankings for at least three alternatives, which complies with what he refers to as the social choice axioms.

328

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

The social choice axioms, namely transitivity, Pareto optimality, independence, and no dictatorship, apply also for safety experts; for numerical examples discussing the limitations of aggregation with regard to the social choice axioms we refer to Beroggi (1999a). The aggregation of cardinal preference values knows similar limitations, mainly pertaining to the difficulty of interpersonal comparisons. Utility theory requires that the assessments comply with the axioms of preferential independence and ordinal positive association, and calls for a benevolent supra decision maker who verifies the assumptions and determines the scaling constants (Keeney and Raiffa, 1993). Moreover, utility-based methods also suffer for not complying with basic social choice axioms, such as the Pareto optimality axiom, which requires that if all experts agree on a preference that the aggregated preference order should be the same. The weighted arithmetic mean and the geometric mean method to aggregate cardinal assessments derived with a ratio scale have also their limitations, for they do not comply with the independence and the Pareto optimality axioms, respectively (Beroggi, 1999a). Ramanathan and Ganesh (1994) proposed an eigenvector method to derive stakeholders’ weights, circumventing the need for a supra decision-maker. In this method, the stakeholders are regarded as ‘indicators’ for which each stakeholder must assess the importance. The interested reader is referred to Ramanathan and Ganesh (1994) for the mathematical implications of the eigenvector method proposed. Beroggi (1999a) states that their method generally complies with the four social choice axioms proposed by Arrow (1951). These theoretical limitations with regard to aggregation and the limitations of deliberation processes mentioned by Apostolakis and Pickett (1998) motivated us to identify principles for an integrated analytic–deliberative process. Several principles, which we defined are in accordance with those defined by Apostolakis and Picket:  Different safety stakeholders must be involved in the process: safety is a concern for planners, fire brigades, emergency responders, infrastructure management, etc., all having different views on safety.  Probabilistic risk analysis (PRA) and deterministic safety analysis must be integrated in the process: employing a deliberation process cannot mean to ignore data and risk analysis studies during the deliberation; however, the results of PRA must be presented in a way which is comprehensible to all safety experts. In addition, deterministic scenarios have to be included in the evaluation as well.  Analytic and discursive elements must be integrated recursively in the process: separating the analysis from the deliberation part precludes experts to reinvestigate PRA results.  Preferences of the various stakeholders have to be aggregated in order to create insights in the preference of the group of stakeholders.  Preference aggregation across the experts can only serve as stimulant for deliberation and not as prescriptive measure: due to the limitations of all aggregation methods, any proposed aggregated preference order should be seen as a proposal which must be accepted by all experts.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

329

Approaches to participatory expert decision making in safety planning have been addressed in the literature (Beroggi and Wallace, 1998; Rosmuller and Beroggi, 1999, Timmermans and Beroggi, 2000). Based upon these authors, we defined in addition, three principles that are different from Apostolakis and Picket:  Each expert should only address aspects of safety for which s/he is competent for: for example, asking a fire brigade officer to include PRA results in his/her preference assessments for alternatives might result in objection or disinterest in the process; instead s/he should be free to choose the safety indicators with which s/he wants to assess the alternatives and for which s/he is an expert.  Preference aggregation across the experts must be done from different point of views: instead of using just one aggregation method, several methods should be used and the experts should be confronted with the possibly different results.  The ranking of the alternatives by the group of stakeholders is discussed rather than the rankings of individual stakeholders; it makes less sense to let the group of stakeholders discuss a ranking of an individual stakeholder knowing that this ranking is exclusively based upon that particular stakeholder’s safety expertise. Rather, it is more useful to discuss aggregated rankings of the group of stakeholders to find out which alternative plans are from a safety point of view fruitful for further analysis. A consequence of these principles is that a flexible preference structure should be devised, which enables quick real-time processing of the stakeholders’ preferences as part of sensitivity analysis. To this end, a central role in participatory group decision-making takes on the facilitator (Anson et al., 1995). The facilitator, for our purposes an expert in infrastructure safety, guides the whole safety evaluation process by combining technical knowledge with skills of moderating participatory meetings. The tasks of the facilitator include identifying the relevant decisionmakers and experts, conducting or monitoring risk analysis studies, processing the results to meaningful information for less analytically skilled decision makers, and organize and guide the meetings where consensus decisions should be reached. 2.2. A Participatory safety evaluation approach for infrastructure planning Four elementary activities have to be conducted as a part of participatory transportation risk analysis (Fig. 1). They consist of (1) hazard identification, (2) safety assessment, (3) safety support, and (4) safety evaluation. The shaded rectangles in this figure indicate the contributions we made, in a broader context (see Rosmuller, 2001) than this paper, to traditional transportation risk analysis (see for example, Rhyne, 1994 or CCPS, 1995). This paper is focused at the safety evaluation part of the participatory transportation risk analysis. Still, the remaining building blocks, i.e., hazard identification, safety assessment and safety support, were performed in our case study; these are the elementary input for the safety evaluation. For this reason we give a brief

330

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

Fig. 1. Participatory safety evaluation approach for infrastructure planning.

description of these building blocks; a comprehensive description of the complete participatory transportation risk analysis can be found in Rosmuller (2001). 2.2.1. Hazard identification As a part of hazard identification, one must do the following things:  appoint a facilitator;  identify the elements of decision and systems analysis including decision makers, the objectives, the safety indicators, the alternative plans, and the uncertainties (Beroggi, 1999b);

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

331

 develop a brief process architecture in cooperation with selected decisionmakers (V&W, 1998);  propose alternative line infrastructure plans;  specify safety information needs; and  identify relevant stakeholders.

2.2.2. Safety assessment The identified alternative line infrastructure plans are assessed with the identified safety indicators by the facilitator. This process includes aspects of probabilistic safety assessment and deterministic scenario analysis, but also the assessment of costs, environmental impacts, emergency response opportunities and economic considerations. The experts and decision makers and their organizations are encouraged to support or even to participate in these assessments. 2.2.3. Safety support The results of the probabilistic risk assessment and deterministic safety analysis must be integrated in a decision support system (DSS), which will be used during the participatory evaluation process with the safety stakeholders. The DSS should be developed in a multimedia environment, in either a centralized or decentralized group decision support setting, on a local area network or on the Internet. Relevant information, which will be processed by the experts during the participatory meeting, must also be integrated in the DSS. Special emphasis in the development of the DSS must be placed on the intuitively sound user interface, which can handle all aspects of multimedia information processing, including animation, audio, video, text, and analytic reasoning. Advanced systems rely on Internet technology, which can be used in centralized and decentralized settings. 2.2.4. Safety evaluation The safety experts and decision-makers engage in a discursive safety evaluation and decision making process, in a centralized or decentralized setting. The facilitator explains the goal of the session, and clarifies the way of working in the session. The first task of the participants is to evaluate the alternative plans, considering the offered safety information that is being provided by the DSS. Each of the decisionmakers evaluates the alternatives using exclusively his/her transport safety indicators. Subsequently, they present their evaluations and rankings of alternative plans to the other experts. A numerical evaluation of alternatives by multiple experts using multiple indicators is never as precise as the numerical values might suggest. Therefore, sensitivity analysis must be conducted. Sensitivity analysis in analytic decision making refers to two aspects, model sensitivity and numerical sensitivity (Beroggi, 1999a). Model sensitivity refers to the well-known phenomena that different models can yield different recommendations, even if the same numbers are being used. However, if different models arrive at the same solutions, we have a robust modeling approach. Numerical sensitivity refers to the variation of numerical evaluations within a band

332

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

of assessment-imprecision. If changes within this band do not affect the decision recommendation then we have robust numerical assessment. To summarize, the following analytic activities are assumed to be effective for this purpose: 1. Each stakeholder determines an order of importance of his/her safety indicators (prioritizing). 2. Each stakeholder evaluates alternative plans using qualitative scores for his/ her indicators. 3. The group of stakeholders generates an order of importance of all indicators used by the group of stakeholders (prioritizing). 4. A ranking of alternatives for the group of stakeholders is generated using the results of 2. and 3. 5. Various multi-criteria strategies (basically: different evaluation rules) are employed to generate additional rankings of alternatives, in order to analyze the robustness of the rankings. To overcome the criticism that the ranking of alternatives heavily depends upon one specific rule applied to aggregate evaluations, it is suggested to apply multiple rules (methods) for aggregation (Voogd, 1982). A multi-method approach enables an analysis of the robustness of the results. To indicate the robustness of the generated ranking for the group of stakeholders, two strategies can be employed:  Non-compensatory strategies: where only the order of importance of indicators is adjusted;  Compensatory strategies: where scores of an alternative on an indicator can be compensated by the scores of the same alternative on other indicators.2 We propose to apply the two strategies and to compare the results. The application of the non-compensatory strategy has already been elucidated before (steps 1– 5). Below, a compensatory strategy is described, in which the following analytical steps are taken (steps a–d): (a) Each stakeholder pairwisely assigns importances to all stakeholders (importance). (b) A single weight per stakeholder is calculated (using AHP) based upon the output from (a). (c) A value per alternative per stakeholder is calculated based upon the multiplication of the rank order values of alternatives (assumed to be on an interval scale and obtained in step 2. of the non-compensatory strategy described above) and with the importance per stakeholder (the output from b). 2

Later in the evaluation session compensatory techniques are assumed to be appropriate, because the stakeholders’ task complexity is reduced significantly due to their earlier experiences with the evaluations of infrastructure plans. After the initial evaluations namely, stakeholders are familiar with the alternative plans, the indicators and the support environment.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

333

(d) a ranking of alternatives for the group of stakeholders is generated based upon the summation of the weighted values per alternative per stakeholder (the output from c). It should be reminded that these aggregation strategies serve mainly the goal of facilitating discussion between the various stakeholders. The decision-makers discuss the rankings with regard to their interests, their evaluations and the weights as assigned to criteria. The session could be concluded with a ranking of line infrastructure alternatives, eventually accompanied by a shared view on this ranking. The participatory session must not necessarily end with a shared view on the ranking of alternatives. Most important is that insights are generated in safety aspects of various alternative line infrastructure plans and the understanding of decision-makers of relevant arguments of the other decision-makers. An indication of safety aspects of alternative line infrastructure plans could be a new starting point for further consideration of safety aspects in the line infrastructure development process. Referring to Fig. 1, system analysis could be conducted to proceed to the process of transportation risk analysis and the values of the transport safety indicators. We applied the above described analytic discursive evaluation approach to an important freight railway project in The Netherlands: the Betuweline, in particular for a part of this railway, the Northeastern Connection. In this paper, we will focus on the participatory safety evaluation session. This means that we will in some cases refer to background reports for in-depth in-sights in the employed methods and techniques to generate the transportation risk analysis for the various safety indicators. We prepared the safety information for the various alternative plans and decision support environment and took on the role of the facilitator ourselves. We invited eight persons who in their daily practice are related to the Northeastern connection and the related safety issues and asked them to perform a participatory evaluation process with regard to line infrastructure safety planning.

3. Betuweline: the Northeastern connection The Netherlands are planning a dedicated high-speed freight railway from the Rotterdam Harbor area to a transfer facility called Valburg, in the eastern part of the country. A freight flow of about 17 million tons per year (of which 3.5 million tons concern hazardous materials) should be transported from Valburg northwards into North-West Europe. Initial plans indicated that a new railway called Northeastern connection would facilitate the freight transport northwards. However, in 2000 existing infrastructure alternatives are reconsidered for the transportation.3

3

While writing this article (2002), the Dutch government has decided to cancel the whole Northeastern connection of the Betuweline for economic reasons.

334

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

To make clear what information was used during the safety evaluation, the key aspects of the hazard identification, safety assessment and safety support are briefly summarized below. Subsequently, the safety evaluation is analyzed in depth. 3.1. Hazard identification The alternatives were specified solely based on transport mode (type) and their location on the map (route): type/route alternatives.4 Based upon transport modalities and routes, six alternative line infrastructure plans were identified as result of a preliminary systems analysis. Three of the six alternatives were rail alternatives for three different routes, called Railway Deventer, Railway Zutphen, and Railway New; two highway alternatives according two routes, called Highway Veluwe and Highway Achterhoek, and one was a water alternative, called Water. The six type/route alternatives had to be evaluated from a safety point of view by the stakeholders. The identification of stakeholders is essential. Based upon Perrow’s (1984) elements of potential victims (voluntariness and benefits), the Projectgroep Integraal Veiligheids Plan (1997) distinguished the people who run risks of exploiting a HighSpeedLine (HSL, a railway still to be developed) in the Netherlands. At least three stakeholders should always be present in infrastructure safety planning, namely: infrastructure providers, spatial development authorities and emergency response organizations. In this study, for practical reasons we will limit our research efforts to these three stakeholders. The safety indicators for the three stakeholders should at least include the following (Rosmuller, 2001): The infrastructure providers use as indicators risk profile of expected user fatalities and costs. Spatial planners use individual risk, societal risk and life-quality. Emergency responders use emergency response mobilization need to repress accident consequences and driving time to the accident spot of emergency response vehicles. The defined set of seven safety indicators was used to evaluate the six type/route alternatives. Each of the stakeholders used its specific safety indicators to evaluate the alternatives. Infrastructure providers were focusing on user risk profile and costs, spatial planners were focusing on individual risk, societal risk and life-quality, and emergency responders were focusing on mobilization need and driving time. 3.2. Safety assessment For the specified information needs and line infrastructure alternative plans, the seven safety indicators were assessed. The assessments of the indicators are presented per stakeholder: infrastructure providence, spatial planning and emergency response.

4 Alternative construction designs such as an embankment, an excavation or a tunnel were not specified. The construction designs could be evaluated subsequently for one or more selected type/ route-alternatives.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

335

3.2.1. Infrastructure provider: user risk profile and costs User risk profiles are normal distributions of the number of expected victims among the users of the line infrastructure per year, accompanied with confidence intervals. We used a bootstrap tool developed by Lohman (1999) to generate these profiles. Engineering judgments were used to assess the costs of the alternative line infrastructure plans. For each alternative line infrastructure plan, the infrastructure providers were presented these normal distributions of fatalities among infrastructure users and engineering judgments for costs. 3.2.2. Spatial planning: individual risks, societal risks and life-quality Individual risk5 and societal risk concern the hazards for people being near line infrastructures for example because of accidents with hazardous materials. Individual risk focuses on the probability a single individual gets killed. Societal risk focuses at multiple victims as a result of one accident. To assess individual and societal risks of the six type/route alternatives, we used the IPORBM software (AVIV, 1997). This software is recommended by the Dutch authorities in case of risk assessment of hazardous material transportation activities on various type/route alternative line infrastructure plans. The individual risk results were presented as iso-risk contours, indicating a certain probability a person in the line infrastructure’s environment gets killed. Societal risk results were presented using FN-curves, indicating the frequency (F) of a certain amount of fatalities (N) and the expected number of fatalities (the surface below the FN curves). With regard to life-quality, several images of the three types of infrastructures (highway, railway and waterway) and the environment of the planned routes (cities, natural parks, agriculture surroundings, small villages) have been shown to the experts. This set of images provided a picture of a possible situation in the future in case the type of infrastructure is constructed. 3.2.3. Emergency response: mobilization needs and driving time The mobilization need to repress accident consequences was expressed in the amount of material and the number of fire fighters necessary to adequately fight proposed accident scenarios. To assess mobilization need we made use of emergency response experts. Emergency responders got the results presented in text. The driving time of emergency response vehicles to accident spots on the alternative line infrastructures was calculated using geographic information systems (GIS). The GIS results show driving time at each segment of the alternative route in magnitudes of a minute. Emergency responders were presented colored maps on which the various colors indicated the minimum driving time.

5

In Dutch external safety policy during 2002, individual risk is renamed into location related risk. Nor the definition, neither the arithmetic has been changed.

336

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

3.3. Safety support The alternative line infrastructure plans, the safety indicator values and several aggregation mechanisms were incorporated in a multimedia decision support system. Because stakeholders used their specific safety indicators, we had to develop four computerized interfaces: one for each of the three stakeholders and one for the process facilitator to integrate the stakeholder evaluations. The computerized interfaces provided the stakeholders relevant information concerning:     

the the the the the

alternative line infrastructure plans; assessed transport safety indicators; preference of indicators; evaluation of alternatives; and ranking of alternative line infrastructure plans.

The safety indicators vary among the stakeholders, according to the stakeholders’ safety interests defined earlier. Hence, the information aspects vary among the various stakeholders. The only information aspect that is the same for all stakeholders concerns the alternative plans. As a result, the computerized interfaces were structured based upon the alternative line infrastructure plans. The assessment of indicators for the six alternative plans for the Northeastern connection and the development of the safety support environment was all preparatory work, and necessary to execute the transportation safety evaluation.

4. Safety evaluation of the Betuweline 4.1. The participants We invited eight persons who in their daily practice were related to the Northeastern connection and its safety. Below, we will describe briefly the participants’ backgrounds. One of the infrastructure providers was affiliated with Dutch National Railways. He had substantial experience with various railway safety related issues such as passenger safety, hazardous material transportation, and emergency response. At the moment of the session, he was affiliated with a group dealing with railway emergency response aspects. He was actively involved in a study that considered safety aspects of the Betuweline. The other infrastructure provider was affiliated with the ministry of Transport, directorate Rijkswaterstaat direction East. He is involved in transport safety policy for highways and waterways in the eastern part of The Netherlands, thus the part where the Northeastern connection would go through. With regard to spatial development we invited residents and spatial planners. The residents we invited were active in a local/regional group with interests in reaction of the Northeastern connection (RONA). One of the residents was the former chairwoman of

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

337

the RONA; the other was actively involved in it. The spatial planners we invited were affiliated with the Provinces of Gelderland and Overijssel who both had the safety aspects of the Northeastern connection under their supervision. A Northeastern connection would by definition go through both provinces. One of the emergency responders was affiliated with the regional fire-fighting brigade Stedendriehoek that covers the cities of Zutphen, Deventer and Apeldoorn. He was involved in evaluating safety aspects of a railway connection going through the region where his brigade was responsible for the emergency response. The other emergency responder was affiliated with the fire-fighting brigade of the city of Arnhem. He was in particular involved in medical aid aspects of a new to be developed railway from Valburg to Oldenzaal at the time the solution for this connection was limited to a ‘new railway’. Six of the eight invited participants participated in the March 2000 session in the city hall of Rheden: a small village centrally located in the region the Northeastern connection would be developed. Unfortunately, both persons affiliated with the provinces of Gelderland and Overijssel were at the last moment not able to attend the session. With the presented safety information and the participants, all the ingredients for the safety evaluation session are described. 4.2. The evaluation of the alternative plans To describe the safety evaluation session, the five steps briefly mentioned in Section 2.2 are followed below. 4.2.1. Importance of indicators per stakeholder The infrastructure providers had to evaluate the alternatives using the risk profiles as well as the construction costs (in Dutch guilders, Dfl6). In the initial order of importance of both indicators, the risk profile was considered to be more important than costs (irisk profile > icosts). The spatial planners had to evaluate the alternatives using individual risk, societal risk and life-quality. In the initial order of importance of indicators, individual risk was considered to be more important than societal risk and life-quality, and societal risk was considered to be more important than live quality (iindividual risk > isocietal risk > ilife-quality). The emergency responders had to evaluate the alternatives using mobilization need and the driving time to the accident site. In the initial order of importance of indicators, mobilization need was considered to be more important than driving time (iemerg. resp. mobilization need > idriving time).

6

1.0 Dfl is about 0.75 Euro.

338

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

4.2.2. Evaluation of alternatives per stakeholder To evaluate alternatives, infrastructure providers could assign a value between 1 (good) and 6 (bad) to the alternatives for each indicator. Table 1 summarizes the results. The far left column presents the rank order values of the alternatives presented in the column on the right of it. The columns ‘risk profile’ and ‘costs’ present the ratings per alternative as assigned by the infrastructure providers for the respective indicators. Induced by the higher importance of the risk profile as compared with costs, the resulting ranking is primarily based upon the evaluation of the risk profiles of the alternatives. It is a coincidence that for both risk profile and costs the alternatives have been evaluated identical. Railway Deventer and Railway Zutphen were evaluated good (1), based upon the risk profile. Moreover, both Railway Deventer and Railway Zutphen have been evaluated identical in terms of costs (1). The result is that both alternatives have been ranked identically. The spatial planners/residents could also assign a value between 1 (good) and 6 (bad) to the alternatives for each indicator. Table 2 summarizes the results. The far left column presents the rank order values of the alternatives presented in the column on the right of it. The columns ‘individual risk’, ‘societal risk’ and ‘life-quality’ present the ratings per alternative as assigned by the spatial planners/residents for the respective indicators. Because in the initial ranking, individual risk was considered to be more important than societal risk, and societal risk was considered to be more important Table 1 Evaluation and ranking of alternatives by infrastructure providers Ranking

Alternative

Risk profile

Costs

1 1 3 4 5 6

Railway Deventer Railway Zutphen Highway Veluwe Waterway Highway Achterhoek Railway New

1 1 3 4 5 6

1 1 3 4 5 6

1=most preferred, 6=least preferred. Table 2 Evaluation and ranking of alternatives by spatial planners/residents Ranking

Alternative

Individual Risk

Societal Risk

Life-quality

1 2 3 4 5 6

Waterway Railway New Railway Zutphen Railway Deventer Highway Veluwe Highway Achterhoek

1 2 3 4 5 6

1 2 4 5 6 5

2 3 6 6 4 3

1=most preferred, 6=least preferred.

339

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349 Table 3 Evaluations and ranking of alternatives by emergency responders Ranking

Alternative

Mobilization need

Driving time

1 2 3 4 5 6

Highway Veluwe Highway Achterhoek Railway New Railway Zutphen Railway Deventer Waterway

1 2 3 4 5 6

4 1 5 3 4 6

1=most preferred, 6=least preferred. Table 4 Initial ranking Rank Alternative

Individual Profile Mobilization Societal Time Costs Life-quality risk need risk

1 2 3 4 5 6

1 2 3 4 5 6

Waterway Railway New Railway Zutphen Railway Deventer Highway Veluwe Highway Achterhoek

4 6 1 1 3 5

6 2 4 5 1 2

1 2 4 5 6 5

6 2 3 4 4 1

4 6 1 1 3 5

2 3 6 6 4 3

1=most preferred, 6=least preferred.

than life-quality, the ranking is primarily based upon the individual risk evaluations of the alternatives. Because for individual risk the spatial planners’ evaluation of alternatives yielded a complete ranking (no alternatives were given the same number), the individual risk rating of alternatives determined the ranking of alternatives. The most preferred alternative is the Waterway. With regard to the Waterway, Table 2 shows that this alternative was also judged good (1) for societal risk and quite good (2) for life-quality. In case spatial planners had rated alternatives equally preferable in terms of individual risk, these alternatives would subsequently be ranked, based upon the evaluations of societal risk. Eventually, in case spatial planners had evaluated alternatives equally preferable in terms of both individual risk and societal risk, lifequality ratings would have determined the ranking of these alternatives. The emergency responders could assign a rating between 1 and 6 to the alternatives per indicator. Table 3 summarizes the results. In this table, the column most to the left presents the rank order values of the alternatives presented in the column on the right of it. The columns ‘mobilization need’ and ‘driving time’ present the ratings per alternative as assigned by the emergency responders for the respective indicators. Because in the initial ranking mobilization need was considered to be more important than driving time, this ranking is primarily based upon the evaluation of mobilization need. Because the evaluations of alternatives based upon mobilization need yielded a full rating from 1 to 6, the ranking based upon mobilization need

340

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

determined the ranking of alternatives. The most preferred alternative is the Highway Veluwe. This might seem odd in relation to the ratings for mobilization need and driving time of alternative ‘Highway Achterhoek’. Table 3 shows that Highway Veluwe was evaluated: 1 respectively 4, whereas Highway Achterhoek was evaluated: 2 respectively 1. This, however, is the result of an approach in which a noncompensatory strategy (elimination by indicator) is applied. The emergency responders however confirmed that the rank order was in line with their opinion. After having evaluated the alternatives, all the stakeholders were asked to give a brief presentation. These presentations were used for initial discussions. To generate insights into the implications of these rankings (per stakeholder) for the group of stakeholders, two aggregation strategies were followed: a non-compensatory strategy and a compensatory one. 4.2.3. Non-compensatory order of importance over the group of stakeholders In Table 4 we summarize the evaluations and present the ranking based upon the predefined order of importance: iir > iprofile > imob.n. > isr > itime > icosts > ilife-q. Here, iir > iprofile means that safety indicator individual risk (IR) is considered to be more important than the safety indicator risk profile. This order of importance was presented to the stakeholders and discussed. To start the aggregation over the group of stakeholders, they considered this order of importance to be meaningful. 4.2.4. Ranking alternatives for the group of stakeholders The resulting ranking of alternatives based upon group aggregation is only based upon the individual risk (IR), because of the full rating of individual risk assigned by the spatial planners/residents. The initial order of importance could easily be adjusted based upon the input of stakeholders. The facilitator could adjust the order of importance in the session in a transparent way. The order of importance was presented on a screen showing adjustments in the order of importance in real-time. Of course, the input of stakeholders was used for that purpose. Table 5 Ranking based upon adjusted order of importance of indicators Rank

Alternative

Societal risk

Individual risk

Profile

Mobilization need

Time

Costs

Lifequality

1 2 3 4 5 6

Waterway Railway New Railway Zutphen Railway Deventer Highway Achterhoek Highway Veluwe

1 2 4 5 5 6

1 2 3 4 6 5

4 6 1 1 5 3

6 2 4 5 2 1

6 2 3 4 1 4

4 6 1 1 5 3

2 3 6 6 3 4

1=most preferred, 6=least preferred.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

341

The emergency responders argued that individual risk is hardly of any interest to them. Individual risk gives no clue with regard to accident consequences, and no clue with regard to emergency response activities. They argued that in particular societal risk is important to them. Their argument was that this indicator is relevant, because it provides insights into the number of victims of a single accident, and consequently in the emergency response mobilization need. The spatial planners were also interested in the consequences of assigning the highest importance to ‘societal risk’. Hence, the emergency responders and spatial planners were interested in an additional ranking based upon the highest importance of ‘societal risk’. Infrastructure providers argued that ‘individual risk’ should always comply with the maximum-acceptable level (i.e., 1.0 E-06). Spatial planners/residents argued that except for the highway alternatives, the IR 1.0 E-06 contours are within 30 m from the line infrastructures, and the maximum-acceptable level is not exceeded. Subsequently, infrastructure providers agreed with the proposal of the emergency responders to rank the alternatives based upon the highest importance societal risk. The other indicators remained in the initial orders of importance generating the following preference order: isr > iir > iprofile > imob.n. > itime > icosts > ilife-q The result of applying this preference order is presented in Table 5. For this situation, the rank order of alternatives is (> indicates ‘preferred’): Waterway > Railway New > Railway Zutphen > Railway Deventer > Highway Achterhoek > Highway Veluwe. Compared to Table 4, Highway Veluwe and Highway Achterhoek switched places as a result of defining societal risk as the most important indicator. The rank correlation between the two rankings in Tables 4 and 5 is high (Kendall’s Tau=0.87), which means that the resulting rank order is quite robust (Kendall Tau equal to 1 implies identical rankings). 4.2.5. Compensatory strategy for the group of stakeholders To describe the activities of the compensatory strategy, the activities (a) up to (d) proposed in Section 2.2 are followed. To indicate the robustness of the ranking in Tables 4 and 5, we developed a ranking based upon the importance of stakeholders. To this end, the rank order values (interpreted in terms of values at an interval scale), e.g., those of Table 1 (infrastructure providence), Table 2 (spatial planning), and Table 3 (emergency response) are multiplied according to the weights of the respective stakeholders. The relative weight of a stakeholder is considered to be related to the safety interests s/he represents. Here, we follow the four steps (a to d) presented in Section 2.2. 4.2.5.1. Pair-wise assignment of importance to stakeholders. To indicate this importance, stakeholders evaluated their mutual importance in a pairwise comparison. Here, importance indicates dominance (in qualitative terms), i.e., one stakeholder’s assessment being more important than the one of another stakeholder, based upon the safety interests s/he represents. This means that a stakeholder compares the

342

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

safety interests of one stakeholder with the interests of other stakeholders (himself included). Saaty’s Analytical Hierarchy Process (AHP) (1980) was used to calculate the weight (in quantitative terms) per stakeholder, expressed as a number between 0 and 1. Weights are thus indirectly assigned by the assignment of orders of importance by stakeholders. 4.2.5.2. Weight per stakeholder Meanwhile, the results of the pairwise comparisons were checked for consistency using the indices consistency index (CI), consistency ratio (CR) and l.7 The interested reader is referred to Saaty (1980) for the mathematical details of calculating these indices. Table 6 summarizes the assessed weights per stakeholder. For example, the spatial planning column shows that spatial planners/residents indirectly assigned a weight of 0.714 to themselves and a weight of 0.143 to emergency responders and infrastructure providers. Spatial planners/residents consider the latter two equally important. Infrastructure providers assigned each stakeholder equally importances. It is quite remarkable that Table 6 shows that both emergency responders and spatial planners/residents assign the highest importance to spatial planning. In the discussion following the assignment, both stakeholders argued that the safety of third parties was considered to be most important because of their involuntary exposure to the transportation risks. Also remarkable is the assignment of infrastructure providers, who consider each stakeholder equally relevant. Their argument was that each safety aspect is relevant and hence equal weights should be attached. The consistency indices for spatial planners indicate perfect consistency. Infrastructure providers are perfectly consistent as well. Although not perfectly consistent, emergency responders are very consistent in their assessment of orders of Table 6 Importances per stakeholder and group weights Infrastructure providence

Spatial planning

Emergency response

Group weight

Infrastructure providence Spatial development Emergency response

0.333 0.333 0.333

0.143 0.714 0.143

0.065 0.736 0.199

0.156 0.661 0.182

l CI CR

3 0 0

3 0 0

3. 234 0.117 0.202

7

Consistency refers to the first social choice axiom being transitivity. CI and l reflect the degree of consistency of the assessments, CR indicates whether the consistency in the assessment should be accepted. In order to judge the consistency CR should, as a rule of thumb, be 20% or less (Beroggi, 1999a). The closer CI and CR approach zero, the more consistent the pairwise comparisons were conducted. The closer l approaches the number of stakeholders, the more consistent the assessment. In case the assignments are considered consistent, their results can be used in a compensatory strategy.

343

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

importance. Hence, the assigned orders of importance were useful (consistent) for the remaining part of the session and were not adjusted to gain better consistency. In Table 6, the sum of the weights in the columns equals 1. Based on Ramanathan and Ganesh (1994), the orders of importance assigned by the stakeholders have been aggregated to determine the group weights per stakeholder. In this table, the far right column presents the group weights being used in the session. The rank order values per stakeholder for the alternatives were multiplied by these group weights. 4.2.5.3. Value per alternative per stakeholder. Based upon these assigned orders of importance and the assumption that the rank order values of alternatives can be interpreted at an interval scale, the compensatory strategy was applied in the evaluation session. The results of this step were used to indicate the robustness of the non-compensatory group aggregation procedure. Based upon the evaluations of alternatives as presented in Tables 1–3, the weighted sum of the alternatives was calculated. The summation over all stakeholders yielded the weighted sum of the alternative. Table 7 summarizes the results, based upon the order of importance of indicators per stakeholder. This table shows, between brackets, the most important indicator per stakeholder. For example, the score for the alternative Highway Veluwe is 3.96: the result of {(50.661)+(30.156)+(10.182)}. The lower the score of the alternative in the column most to the right, the better the alternative. 4.2.5.4. Ranking alternatives for the group of stakeholders. The rank order of alternatives for this situation is (> indicates preferred): Waterway > Railway New > Railway Zutphen > Railway Deventer > Highway Veluwe > Highway Achterhoek. This rank order is exactly the same as the one exclusively based upon the initial order of importance (see Table 4). Hence, Kendall’s Tau, a measure for the robustness of rankings, equals 1. This rank order is slightly different from the Table 7 Scores and ranking of alternatives

Highway Veluwe Highway Achterhoek Waterway Railway Deventer Railway Zutphen Railway New

Spatial planning (Indivdual Risk)

Infrastructure providence (Profile)

Emergency response (Mobilization need)

Scores using group weights

5 6 1 4 3 2

3 5 4 1.5a 1.5a 6

1 2 6 5 4 3

3.96 5.12 2.38 3.79 2.95 2.81

1=most preferred, 6=least preferred. a Infrastructure providers ranked both railway Zutphen and railway Deventer as the best alternatives (1). The next best alternative will be ranked third. To assign a number to each of the two best alternatives, the first and second rank order numbers are summed up and subsequently divided by the two alternatives: (1+2)/2=1.5.

344

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

adjusted order of importance (highest importance of societal risk, Table 5): Highway Veluwe and Highway Achterhoek switched ranks. Still, the ranking of alternatives in Table 7 proved to be quite robust compared with the ranking presented in Table 5: Kendall’s Tau equals 0.87. The stakeholders indicated that these results supported their feeling that the alternatives Waterway, Railway New and Railway Zutphen were quite promising alternatives from a safety point of view. Although the non-compensatory strategy also put these three alternatives on top (Table 4), the difference compared to the remaining three alternative line infrastructure plans did not become clear. The difference between these three alternatives and the rest however became clearly visible with the compensatory strategy. The added value of these compensatory rankings is that it became obvious that the six feasible alternatives for the Northeastern connection were split up into two groups: a group of preferred alternatives including the Waterway, Railway New and Railway Zutphen and a group of less-preferred alternatives (Railway Deventer, Highway Veluwe and Highway Achterhoek). The stakeholders discussed these results and concluded that in spite of the fact that this approach provided new insights, additional insights were necessary. Infrastructure providers argued that it would be interesting to find out what the ranking would be in case all indicators were equally important. The spatial planners and emergency responders agreed on this proposal. This meant that the ratings (again assumed to be of an interval scale) of alternatives per stakeholder per indicator as presented in Table 4 were multiplied by the weights per stakeholder. The results are presented in Table 8. In this table, for example, the score for the alternative Highway Veluwe is the sum of:  Spatial planning (Ind.R., Soc.R., Life-Q.): (5+6+4)0.661=9.92  Emergency response (Mob. Need, Time): (1+4)0.182=0.91  Infrastructure provider (Profile, Costs): (3+3)0.16=0.94 The summation over the three stakeholders (9.92+0.91+0.94) yielded a score for Highway Veluwe of 11.76. Based upon the weighted sum, the rank order of alternatives for this situation is (> indicates ‘preferred): Waterway > Railway New > Railway Zutphen > Highway Achterhoek > Highway Veluwe > Railway Deventer. Comparing this rank order with the one presented in Table 4 (Kendall’s Tau=0.60) Table 8 Ranking alternatives based upon the same preference of indicators Rank

Alternative

Weighted sum scores

1 2 3 4 5 6

Waterway Railway New Railway Zutphen Highway Achterhoek Highway Veluwe Railway Deventer

6.08 7.23 10.18 11.36 11.76 11.87

1=most preferred, 6=least preferred.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

345

and Table 7 (Kendall’s Tau=0.60), indicates that the weighted sum rank order is quite robust. The stakeholders discussed the resulting rank order. The result of this discussion was that Railway Zutphen was no longer considered to be a preferable alternative. Therefore, Railway Zutphen was removed from the group of preferred alternatives (with a relatively low weighted sum score [Waterway (6.08) and Railway New (7.23)] to the group of less preferred alternatives with a relatively high weighted sum score [Railway Zutphen (10.18), Railway Deventer (11.87), Highway Veluwe (11.76), and Highway Achterhoek (11.36)]. All stakeholders agreed that from a safety point of view, the alternatives Waterway and Railway New were fruitful alternatives for further consideration in the infrastructure planning regarding the Northeastern connection. With these shared insights, the stakeholders and the facilitator ended the session.

5. Analysis of the safety evaluation process Data concerning the evaluation process were gathered in several ways: there were two observers in the room, we made audio tape recordings, participants filled out questionnaires before, during and after the session, and we interviewed participants by telephone a week after the session. As part of the direct assessment of the alternatives, we asked the participants, after the session, several questions. With respect to the safety evaluation process, we asked the participants ‘Does an integral approach contributes to a shared view of safety aspects of the alternatives’? Participants had to rate their score on a scale, where 1 stands for total disagreement, 5 stands for neutral, and 10 stands for total agreement with the question. The spatial planners both scored 7, just like one of the infrastructure providers and an emergency responder. The remaining infrastructure provider and emergency responder both scored 7,5. These scores indicate relative strong support for our hypothesis that an integral approach contributes to a shared view of safety aspects of the alternatives. To assess whether the analytic discursive safety evaluation process contributed to acknowledgement of stakeholders safety interests, we asked participants by questionnaire to judge how much aware they were of the safety interests of the other stakeholders. We summarized the judgment of gained insights by participants in the three schemes in Fig. 2. The structure of the three schemes is identical, only the subject per scheme is different. The upper scheme concerns the gained insights in each other interests, the scheme in the middle concerns the gained new insights and the lower scheme concerns the understanding for other stakeholders’ interests. From the three schemes, it looks like spatial planners (local residents) and infrastructure providers were less attracted to each other than each of these stakeholders was with the emergency responders. Fig. 2 indicates that in most cases the session has provided insights into the safety interests of other participants. The differences between

346

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

participants representing the same stakeholder are relatively small. Still, participants stated in the ex-post evaluation that differences could cohere with the differences in experience of the various persons and involvement in the project of the Northeastern connection. These schemes indicate that participants gained insights in safety interests of other stakeholders. Maybe just as important, they showed understanding for these interests instead of ignoring these interests. Analysis of the audiotapes and observation reports revealed that participants intensively discussed safety aspects of the alternatives. Despite intense discussions, observers called the atmosphere in which the discussions took place ‘pleasant’. The debriefings after the session confirmed that participants very much appreciated the

Fig. 2. Judgment of gained (new) insights and understanding.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

347

discussions. This provides support for our idea that bringing together the various stakeholders and their safety indicators and the proposed preference aggregation process helped the participants to reach a shared view on the infrastructure decision problem.

6. Conclusions and discussion We have proposed a participatory safety evaluation method that integrates probabilistic risk assessment, deterministic safety analysis with discursive safety evaluation by different groups of safety experts. The method was applied in the line infrastructure project of a high-speed freight railway in the Netherlands. Infrastructure providers, spatial planners and emergency responders had to evaluate six alternative line infrastructure plans using their own safety indicators. Several aggregation techniques (compensatory and non-compensatory) were employed to generate rank orders of the six alternative plans for the group of stakeholders. These rank orders facilitated discussion among the stakeholders resulting in better understanding of each others’ arguments. For the test case and based upon several aggregation methods, a waterway and a newly developed railway seemed to be particularly fruitful for further elaboration from a safety point of view. The results of an analytic discursive safety evaluation session was that the various stakeholders developed a shared view on which alternative line infrastructure plans from a safety point of view are worthwhile for further consideration and which are not. This conclusion was mainly accomplished by the integral approach in which various safety stakeholders and safety indicators were brought together for evaluating together safety aspects of several alternative line infrastructure plans. Therefore, we have the opinion that we have gathered sufficient evidence for our hypothesis, that an integral approach contributes to a shared view on safety aspects of alternative line infrastructure plans. We recognize, however, that the used safety indicators and the assessed values of these indicators for the various alternatives (some favoring the same alternative plans) could be part of the reason for participants supporting the added value of an analytic discursive safety evaluation session. However, Tables 4 and 7 show that the various stakeholders differ to a large extend in their rank orders per indicator. Another part of the reason for participants supporting the added value of an analytic discursive safety evaluation session might be the session support provided by the facilitator and the multimedia decision support environment. Participants judged the multimedia support during the session very valuable. Moreover, the role of the facilitator supporting the interaction processes during the session, was also assessed as very positive. In the debriefing session, we asked the participants to judge the role of the facilitator with regard to the progress of the session and their influence on the preference assessment of the alternatives. The most important conclusion was that the role of the facilitator is considered to be important for the progress of such a participatory evaluation session. This finding provides evidence for the procedural aspect of employing an integral safety evaluation process.

348

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

We are aware of the relatively small number (six) of participants that confirmed our hypothesis and the fact that the outcomes of the session had no implications for developments in reality. This is why we propose to employ our integral approach to several real-life line infrastructure projects for which evaluation outcomes really matter for further project developments. In such an integral approach we advise to let stakeholders evaluate alternative line infrastructure plans only using their ‘own’ specific safety indicators, using a multi-method approach for aggregating individual stakeholders rank orders to group rank orders and to discuss the group rank orders rather than the individual stakeholder rank orders. In brief, we found positive methodological support for the proposed integral approach and preference aggregation methodology, which helps safety experts reach consensus.

References Accorsi, R., Apostolakis, G.E., Zio, X., 1999. Prioritizing stakeholders concerns in environmental risk management. Journal of Risk Research 2 (1), 11–29. Anson, R., Bostrom, R., Wynne, B., 1995. An experiment assessing group support system and facilitator effects on meeting outcomes. Management Science 41/2, 189–208. Apostolakis, G.E., Picket, S.E., 1998. Deliberation: integrating analytical results in environmental decisions involving multiple stakeholder. Risk Analysis 18 (5), 621–634. Arrow, K.J., 1951. Social Choice and Individual Values. John Wiley, New York. AVIV, 1997. Interprovinciaal risicoberekeneingsmethodiek (IPO-RBM), Consultancy Agency for Industrial Safety (AVIV), Enschede. Beinat, E., Nijkamp, P., Rietveld, P., 1994. Value functions for environmental pollutants: a technique for enhancing the assessment of expert judgments. Environmental monitoring and assessment 30, 9–23. Beroggi, G.E.G., 1999a. Decision Modeling in Policy Management—An Introduction to the Analytic Concepts. Kluwer, Boston. Beroggi, G.E.G., 1999b. Safety Concepts in Land Use Engineering—The Case of Underground Infrastructures. Delft University Press, Netherlands. (ISBN: 90-407-1810-5). Beroggi, G.E.G., Wallace, W.A., 1998. Operational Risk Management–The Integration of Decision, Communications, and Multimedia Technologies. Kluwer Academic Publishers, Boston. Beroggi, G.E.G., Wallace, W.A., 2000. Multi-expert operational risk management. IEEE Transactions on Systems, Man, and Cybernetics PartC (30/1), 32–44. Bonano, E.J., Apostolakis, G.E., Salter, P.F., 2000. Application of risk assessment and decision analysis to the evaluation, ranking and selection of environmental remediation alternatives. Journal of Hazardous Materials 71, 35–75. CCPS, 1995. Guidelines for Chemical Transportation Risk Analysis, American Institute for Chemical Engineers (AIChE). Center for Chemical Process Safety (CCPS), New York. Cooke, R.M., 1991. Experts in Uncertainty: Opinion and Subjective Probability in Science. Oxford University Press, Oxford. Daniels, S.E., Lawrence, R.L., Alig, R.J., 1996. Decision-making and ecosystem-based management: applying the Vroom–Yetton model to Public participation strategy. Environmental Impact Assessment Review 16, 13–30. DeWispelare, A.R., Herren, L.T., Clemen, R.T., 1995. The use of probability elicitation in the high-level nuclear waste regulation program. International Journal of Forecasting 11, 5–24. Keeney, R.L., Raiffa, H., 1993. Decisions with Multiple Objectives. Cambridge University Press, Cambridge. Linstone, H.A., Turoff, M., 1975. The Delphi Method, Techniques and Applications. Addison-Wesley, Reading, MA.

N. Rosmuller, G.E.G. Beroggi / Safety Science 42 (2004) 325–349

349

Lohman, F.A.B., 1999. The Effectiveness of Management Information, PhD Thesis. Delft University of Technology, Delft, Netherlands. Mumpower, J.L., 2001. Selecting and evaluating tools and methods for public participation. International Journal of Technology, Policy, and Management 1/1, 66–77. Myung, I.J., Ramamoorti, S., Bailey Jr., A.D., 1996. Maximum entropy aggregation of expert predictions. Management Science 42 (10), 1420–1436. NRC, 1996. Understanding risk: Informing decisions in a democratic society. National Academy Press, Washington D.C. Perrow, C.h., 1984. Normal Accidents, Living with High-risk Technologies. Basic books, New York. Project Integraal Veiligheids Plan, 1997. Sturen op veiligheid. Dutch Ministry of Transport and Public Works, Department of Passenger Transport, Projectgroup Highspeedline-South; The Hague, Repnr. HSLV-R-97067. Ramanathan, R., Ganesh, L.S., 1994. Group preference aggregation methods employed in ahp: an evaluation and an intrinsic process for deriving members’ weightages. European Journal of Operational Research 79, 249–265. Renn, O., Webler, T., Wiedemann, P., 1995. Fairness and Competence in Citizen Participation. Kluwer Academic Publishers, Boston. Rhyne, W.R., 1994. Hazardous Materials Transportation Risk Analysis: Quantitative Approaches for Truck and Train. Van Nostrand Reinhold, New York. Rosmuller, N., 2001, Safety Analysis of Transport Corridors. Thesis, Delft University of Technology, Faculty of Technology, Policy and Management, TRAIL Research School, Delft. Rosmuller, N., Beroggi, G.E.G., 2000. Participatory multi-disciplinary safety screening of infrastructures. Safety Science 35/1–3, 151–173. Saaty, T.L., 1980. The Analytical Process. McGraw-Hill, New York. Sandri, S.A., Dubois, D., Kalfsbeek, H.W., 1995. Elicitation, assessment, and pooling of expert judgments using possibility theory. IEEE Transactions on Fuzzy Systems 3/3, 313–335. Timmermans, J.S., Beroggi, G.E.G., 2000. Conflict resolution in sustainable infrastructure management. Safety Science 35, 175–192. Van Uden, J.H.A., Heijkamp, A.H., 1995. Intrinsic road safety—a new approach. Safety Science 19/2–3, 245–252. V&W, 1998. Raamwerk Beleidsvorming Hoofdinfrastrutuur. Dutch Ministry of Transport and Public Works, and Water Management, Directorate-General of Passenger Transport, The Hague. Voogd, H., 1982. Multicriteria Evaluation for Urban and Regional Planning. Delftsche Uitgevers Maatschappij B.V, Delft. Zio, E., Apostolakis, G.E., 1998. Application of sensitivity analysis to environmental management. In: Proceedings of the Second International Symposium on Sensitivity Analysis of Model Output (SAMO), Venice, pp. 333–336.