4.3.1. Intelligence phase: determination of the network
The problem definition overlaps the decision-making intelligence phase, which refers to the structuring of the problem, the identification of the objectives and the selection of criteria or attributes to describe the degree of achievement of each objective.
The location of a new waste incinerator plant is a complex land use planning problem in which the presence of interrelated elements and conflicting aspects suggests the use of a MCA that is able to provide a rational base for the systematic analysis of the alternative options. Due to the aforementioned complexity of the problem under analysis, and with the aim of finding the most significant aspects involved in the decision and the most suitable location for the project, the ANP method has been developed according to the so-called BOCR model, which allows BOCR to be considered (Saaty and Ozdemir, 2008). Each decision in the BOCR approach can be characterized by the presence, in the short-medium term, of some favourable sure concerns (Benefits) and some unfavourable ones (Costs), and, in the long term, of some uncertain positive concerns that the decision might create (Opportunities) and some negative factors that it can entail (Risks). In the ANP–BOCR method, each of these four concerns utilizes a simple separate network structure for the decision.
In the present application, the Benefits and Opportunities, apart for the temporal dimension, have been identified as the aspects that have to be maximized in the selection of the most compatible site for the location of the new waste incinerator facility. The Costs and Risks, leaving aside the temporal dimension, have instead been identified with those factors that have to be minimized in the aforementioned decision context. Furthermore, the Benefits and Costs refer to endogenous aspects that describe the territorial system, whereas the Opportunities and Risks concern the exogenous factors that the construction of the plant could determine on the territorial system.
Starting from the overall objective of the analysis, which is the identification of the most suitable site for the location of a new waste incinerator plant, a comprehensive set of evaluation criteria that reflect all the concerns relevant to the decision problem has been identified (Table I). Taking into account the full range of aspects relevant to the decision problem enhances the quality of the final decision, allowing the totality of the effects of the transformation project to be considered and the negative externalities and the intergenerational effects to be minimized. It is necessary to put in evidence that the criteria considered in the present application reflect the requirements coming from the legislative framework in the context of Environmental Impact Assessment (first of all, the European Directive 11/97). Further information for the structure of the decision model has been derived from the specific legislation in the field of waste management both at the national and at the local level. In this sense, of particular interest is the Waste Management Plan of the Province of Torino (ATOR, 2008), which provides a list of aspects to be considered for the location of waste facilities.
Table I. List of the indicators for the ANP–BOCR model and the associated performances of the alternative sites
| || || || || || || ||Performances of the alternatives|
|BOCR||Control criteria||Clusters||Indicators||Abbreviations||Unit of measure||Assessment direction||Ivrea||Rivarolo canavese||Settimo torinese|
|Benefits||Environmental aspects||Noise and electromagnetic fields||Acoustic class||AC||Classes||III||III||VI|
| ||Socio-economic aspects||Infrastructural aspects||Infrastructural systems inside the 2 km range area||ID||km−1||6.1||3.6||6.2|
|Opportunities||Environmental aspects||Air||Emission reductions due to energy recovery||ER||Classes||Low||Low||Medium|
| || ||Soil and water||Industrial area requalification||IA||Expert judgement||Absence||Medium||High|
| || || ||Aquifer transmissivity||AT||m2/s||2.78 × 10−3||6 × 10−4||4.5 × 10−2|
| ||Socio-economic aspects||Urban planning||Coherence with the planning instruments in force||CP||Classes||Low||Medium||High|
|Costs||Environmental aspects||Soil and water||Geological risk||GR||Expert judgement||Yes||No||No|
| || || ||Surface water quality index||SW||Classes||Medium||Medium/low||Low|
| || || ||Vulnerability [% of areas with high vulnerability inside the 5 km buffer]||VA||%||91||31||32|
| || ||Air||NO2: number of cells exceeding the annual limit value (40 µg/m3)||NO||Number||0||13||29|
| || || ||PM10: number of cells exceeding the annual limit value (40 µg/m3)||PM||Number||0||4||12|
| || ||Noise and electromagnetic fields||Length of the electricity transmission network||LE||km||4.6||9.8||26.1|
| || ||Landscape and ecosystems||Biodiversity index [% of the territory inside the 2 km range area characterized by a high biodiversity index]||BI||%||39.6||35.7||3.8|
| || || ||Natural value||NV||Classes||High||High||Low|
| || || ||Landscape quality||LQ||Classes||High||Medium||Low|
| || || ||Significance of the cultural and historical heritage||SC||Classes||High||Low||Medium|
| ||Socio-economic aspects||Urban planning||Presence of constraints||PC||Presence/absence||Yes||No||No|
| || ||Infrastructural aspects||Actual traffic flows inside the 2 km range from the site||AF||Number of vehicles||12 200||16 600||27 315|
| || || ||Number of accidents per 100 km||NA||Number||25.3||43.9||39.2|
| || ||Demographic system||Density of population||DP||People/km2||931||28||449|
| || || ||Number of resident inhabitants||NR||Number||40 596||20 410||82 517|
| || || ||Rural real estate||RR||€||4 031 079||6 813 622||5 096 515|
| || || ||Residential real estate||RE||€||108 546 700||497 44 860||87 335 748|
|Risks||Environmental aspects||Air||Dispersive capacity of the area: average of the annual average concentrations of NOx evaluated on the 50 cells with the highest average concentration (ìg/m3)||DC||Number of cells||2.84||1.83||1.58|
| || || ||CO2 emitted||CO||t/year||220.6||154.3||125.1|
| || ||Landscape and ecosystems||Landscape sensitivity||LS||Classes||High||Medium||Low|
| || ||Noise and electromagnetic fields||Presence of sensitive receptors within 500 m||PS||Number||6||6||2|
| ||Socio-economic aspects||Infrastructural aspects||Distance from the nearest train station||DN||km||5.8||3.1||6.7|
| || || ||km covered in 1 year for the collection of the wastes||KM||km||1 102 782||771 393||625 354|
| || || ||Variation of the percentage of heavy lorries within a range of 2 km from the site||VH||%||0.8||0.6||0.4|
| || ||Demographic system||Number of buildings per each covered km||NB||Number/km||4.3||7.7||6.8|
Since the complex decision problem under analysis involves both environmental and socio-economic aspects, the present paper proposes a complex BOCR structure where the two aforementioned aspects have been considered as a control criteria in each of the four subnetworks. Mention should be made to the fact that, taking into account the nature of the decision problem under examination, it has been generally agreed that the environmental aspects play a very important role in the evaluation process; in any case, for a comprehensive analysis of the problem, the model has to reflect also the social dimension and the economic aspects (Morrissey and Browne, 2004). In this work, it has been decided to analyse all the aforementioned aspects, paying particular attention to the implications that the project has from the environmental point of view. For this reason, the elements of the problem have been organized according to two control criteria, namely environmental aspects and socio-economic aspects. At the same time, this choice allows the internal structure of the four subnetworks to be balanced, having the same number of elements belonging to the environmental aspects control criterion and to the socio-economic aspects control criterion.
As far as the structuring of the network is concerned, mention should be made to the fact that, as stated by Miller (1956), it is important not to have too many elements to compare through the expression of relative judgements because the decision aid tool should not be interpreted as an algorithm that is able to automatically give the desired solution but it should instead support the DM who has to make a systematic analysis of the alternative solutions and who is solely responsible for the final choice. For the sake of coherence, particular care has been taken in the present application, to have subnetworks that are easily managed by the DMs. In this sense, for each of the control criteria in the four subnetworks, the number of the elements is limited (from a minimum of 4 elements for the environmental aspects of the Benefits subnetwork to a maximum of 13 elements for the environmental aspects of the Costs subnetwork). Moreover, the number of the elements belonging to the two control criteria is almost the same for each of the fours subnetworks (4 elements for both the two control criteria of the Benefits subnetwork, 6 and 4 elements for the environmental aspects and socio-economic aspects control criteria, respectively, of the Opportunities subnetwork, 13 and 10 elements for the environmental aspects and socio-economic aspects control criteria, respectively, of the Costs subnetwork and 7 elements for both the two control criteria of the Risks subnetwork). The design of the model described allows small and balanced networks to be considered. This is particularly important from a cognitive point of view because it has been generally agreed that the DM tends to give more importance to those aspects that are described by a higher number of elements in the model (Saaty, 1980; Simon, 1960).
As described above (see Section 4.1), the present paper proposes a set of indicators as nodes of the network, since they are supported by sure information.
A measurement scale and an assessment direction (
for Benefits and Opportunities and
with reference to Costs and Risks) must be established for each indicator.
Table I presents the set of indicators that have been identified for the analysis and provides the control category (BOCR) to which each indicator belongs, the control criteria, the measurement unit, the assessment direction and the evaluation of the performance for the three alternative sites. With specific reference to the analysis of the environmental aspects of the problem, these describe the main impacts that a waste incinerator plant is likely to have on the environmental system. In this sense, under the environmental aspects control criterion four different clusters have been identified, namely air pollution, soil and water, landscape and ecosystems, noise and electromagnetic fields. Table II provides a brief description of the aforementioned clusters.
Table II. Environmental aspects of the decision problem
|Air||The criterion considers the existing air quality in the territorial system under examination and the possible variations deriving from the polluted emissions of the waste incinerator plant, with reference to both the construction phase and the operation one|
|Soil and water||The criterion takes into account the presence of possible constraints from the point of view of the characteristics of the soil (e.g. hydrogeological risk, agricultural land, etc.) that could prevent the construction of the plant; moreover, the criterion considers the existence of positive concerns that could orient the location of the waste incinerator (e.g. areas subjected to land reclamation). With reference to the water component, the criterion considers the quality of the surface water bodies and the vulnerability of the underground water; furthermore, the transmissivity of the aquifer that could positively affect the location of the plant from the point of view of the water supply is taken into account|
|Landscape and ecosystems||The criterion reflects the sensitivity of the landscape with respect to the introduction of the incinerator plant in the surrounding area, paying particular attention to the visual impacts of the project|
|Noise and electromagnetic fields||The criterion takes into account the protection of the environmental system from the negative impacts deriving from the acoustic emission generated by the incinerator. Moreover, the presence of electric lines that could negative interfere with the plant from the point of view of electromagnetic pollution is considered|
It is important to underline that the analysis dimension concerns a 2-km range area from each site. In the context of the undesirable facilities location problems, the aforementioned assumption leads to identify more negative aspects (Costs and Risks) than positive ones (Benefits and Opportunities) since the former spread more easily on the territory. Mention should also be made to the fact that social opposition-related indicators have not been considered in the structuring of the network, since they would have been equal for the three alternative sites.
4.3.2. Design phase: determination of the elements and clusters priorities
The design phase involves data collection and processing and the elicitation of the preference of the DMs. Taking into consideration the structure of the model identified for the analysis of the decision problem (Table I), the application will only be illustrated with reference to the Costs subnetwork with the aim of simplifying the explanation. The application is similar for the other subnetworks of the model (Benefits, Opportunities and Risks).
The Costs subnetwork concerns the negative aspects of the territorial context, which have to be minimized in order to obtain the most suitable solution (Figure 2). Two control criteria have been identified, namely environmental aspects and socio-economic aspects. The environmental aspects have been further divided into four clusters (soil and water, landscape and ecosystems, noise and electromagnetic fields, air), whereas the socio-economic aspects have been organized in three clusters (urban planning, demographic system, infrastructural aspects).
Figure 2. The ANP–BOCR model for the case under examination. As an example, the figure shows in detail the structure of the subnetwork costs, according to the two control criteria that have been identified.
Download figure to PowerPoint
According to the ANP methodology, once the network has been identified, it is necessary to represent the influences among the elements. It has been chosen to approach this task according to the following strategies. To start with, all the elements in the four subnetworks are supposed to have an influence on the alternatives. Secondly, in order to take into account the complexity of the decision problem, the feedbacks of the alternatives on the elements of the model are considered. Further relationships have then been identified concerning the potential influences among the elements of each subnetwork. These influences reflect the natural dynamics of the environmental and territorial systems, where link and interaction pathways exist between individual elements, which can, positively or negatively, affect each other. For example, with reference to the Cost subnetwork (Figure 2), considering the environmental aspects, it has been decided to insert in the model the influence that the element ‘water quality’ (‘soil and water’ cluster) exerts on the elements ‘biodiversity index’ and ‘natural value’ (‘landscape and ecosystems’ cluster). This relationship is represented in the model with an arrow that goes from the ‘soil and water cluster’ to the ‘landscape and ecosystem’ cluster.
After the identification of the influences among the elements of the model, it is necessary to develop the pairwise comparisons in order to determine the priorities of the elements of the problem. In this case, the judgments that have to be used in the comparison matrixes have been formulated according to the information contained in the related indicators. The degree to which the objectives are met, as measured by the indicators, is the basis for comparing alternatives. The factor scores related to the performance of the three alternatives, with reference to each indicator, have been normalized using the mode ideal normalization, as represented in the following equation (2):
The selected standardization function offers the advantage of maintaining the ratio between the normalized and the original scores. The scores obtained from the standardization have then been translated into a 1-to 9-point scale. In particular, the highest score that has been obtained through the standardization function (i.e. the score 1) is made to correspond to the highest score in the 1-to 9-point scale (i.e. the score 9); in a proportional way, the other normalized scores have been translated from the [0,1] domain into the [1,9] domain. The difference between the thus obtained scores represents the judgments that have to be used to fill in the related comparison matrix.
For example, with reference to the ‘residential real estate’ indicator, Table III shows the procedure followed for the translation of the indicator values into the 1-to 9-point scale scores, whereas Table IV represents the comparison matrix of the three alternatives with respect to the considered indicator.
Table III. Scores of the ‘residential real estate’ indicator for the three alternative plant locations
|Alternatives||Indicator value||Normalized indicator score||Indicator score on the 1-to 9-point scale|
|Ivrea||108 546 700 €||1||9|
|Rivarolo||49 744 860 €||0.46||4.14|
|Settimo||87 335 748 €||0.80||7.20|
Table IV. Pairwise comparison matrix of the alternatives with reference to the ‘residential real estate’ indicator
|Residential real estate||Ivrea||Rivarolo||Settimo||Priorities|
The procedure that has been followed for the standardization of the indicator values allows the feedbacks of the alternatives on the indicators to be assessed, as well as the interrelationships among the indicators: this is the crucial point of the ANP method. For example, with reference to the influences of the Ivrea site on the indicators of the ‘demographic system’ cluster, Tables V and VI represent the standardized scores of the considered indicators and the related pairwise comparison matrix, respectively.
Table V. Standardized scores of the indicators related to the ‘demographic system’ cluster for the Ivrea site
|Elements||Indicator value||Indicator score on the 1-to 9-point scale|
|Number of inhabitants||40 596||4.41|
|Rural real estate||4 031 079 €||5.31|
|Residential real estate||108 546 700 €||9|
Table VI. Pairwise comparison matrix of the indicators related to the ‘demographic system’ cluster with reference to the Ivrea site
|Ivrea||Density||Number of inhabitants||Rural real estate||Residential real estate||Priorities|
|Number of inhabitants||1/5||1||1/2||1/5||0.07|
|Rural real estate||1/4||2||1||1/4||0.12|
|Residential real estate||1||5||4||1||0.41|
Once all the pairwise comparison matrixes have been filled in, the totality of the related priority vectors forms the unweighted supermatrix. In order to alleviate the mathematical burden, all the calculations have been implemented using the Superdecision software (www.superdecision.com). Table VII represents the unweighted supermatrix for the control criterion of the socio-economic aspects under the Costs subnetwork. The priorities of the elements that were previously compared (Tables IV and VI) are shown.
Table VII. Unweighted supermatrix for the socio-economic aspects under the Costs subnetwork
The priorities of the clusters have been derived from expert opinions of the Provincial Administration and have been used to weight the initial supermatrix: the result is the weighted supermatrix.
4.3.4. Detailed analysis and implementation: sensitivity analysis
After obtaining a ranking of the alternatives and despite the coherence obtained in the results, it was considered useful to perform a sensitivity analysis on the final outcome of the model in order to test its robustness. The sensitivity analysis is concerned with a ‘what if’ kind of question to see if the final answer is stable when the inputs, whether judgments or priorities, are changed. It is of special interest to see whether these changes modify the order of the alternatives.
In the present paper three different sensitivity analyses have been undertaken in order to study the robustness of the model with respect to the components and interdependencies of the network. In the first one, the stability of the solution has been studied with regard to the control criteria (BOCR) priorities. In the second one, the analysis has explored the modification of the influences of the alternatives on the criteria and of the criteria on the alternatives. Finally, in the third study, the paper attempts to verify the rank reversal of the alternatives (Saaty, 2006) by eliminating one alternative at a time from each subnetwork of the model and thus studying the resulting final ranking searching for potential changes.
In the first study, while measuring the sensitivity of the alternatives to the BOCR weights, an additive formulation is used, since the meaningful changes could not be obtained by a multiplicative formulation (Tuzkaya et al., 2007). The sensitivity analysis for the four subnetworks is represented in Figure 3, where the x-axis represents the changes in the weight of the control criteria, whereas the y-axis represents the changes in the weights of the alternatives.
When the relationships between the Benefits dimension and the location alternatives are considered, it becomes clear that the share of Settimo Torinese increases and Rivarolo Canavese decreases, while the weight of the benefits increases (Figure 3(a)). This implies that Settimo Torinese has positive features and Rivarolo Canavese has negative features, in terms of benefits. The main reason for this is the higher number of infrastructures and acoustic class in Settimo Torinese and the lower level of the same indicators in Rivarolo Canavese. Although Ivrea is not sensitive to the changes in the weight of the benefits, mention should be made to the inversion in the ranking of the alternatives from Settimo Torinese–Rivarolo Canavese–Ivrea (for 0% benefit weight) to Settimo Torinese–Ivrea–Rivarolo Canavese (for 100% benefit weight), which depends on the weight the benefits are assigned.
With reference to the Costs dimension (Figure 3(b)), the sensitivity analysis shows that this is the most unstable subnetwork, since both the results and the rank of the alternatives are very sensitive to the changes in the weight of the costs. It can be noticed that the ranking of the alternatives changes from Settimo Torinese–Ivrea–Rivarolo Canavese (for 0% cost weight) to Settimo Torinese–Rivarolo Canavese–Ivrea (for 50% cost weight) to Rivarolo Canavese–Settimo Torinese–Ivrea (for 70% cost weight) to Rivarolo Canavese–Ivrea–Settimo Torinese (for 100% cost weight). On the basis of the costs factors, it may be asserted that Rivarolo Canavese and Ivrea have positive features and the increase in the weight of the costs affects them positively. On the other hand, Settimo Torinese has negative features in terms of costs.
Similar observations can be made for the other subnetworks. It is possible to see that the Ivrea and Rivarolo Canavese sites have almost equal performances in the Opportunities and Risks subnetworks (Figure 3(c and d)) and that Ivrea can be identified as being the least sensitive alternative to the changes in the weight of the control criteria.
With reference to Settimo Torinese, which is the winning alternative for all the aggregative formulas that have been considered, the sensitivity analysis highlights that the site has positive performances in terms of Benefits and Opportunities and negative performances in terms of Costs and Risks. Mention should be made to the fact that the Settimo Torinese site has also resulted to be the best solution in the study carried out by the Provincial Administration (ATOR, 2008).
In the second sensitivity analysis, the influences between the elements and the alternatives have been modified and the resulting final priority list of the alternatives has been analysed in order to see if some changes occurred.
Particularly, the priorities of the elements of the model resulting from the limit supermatrix have been considered and, for each control criterion of the four subnetworks, the highest priority alternative and the highest priority element have been taken into account.
In order to perform the analysis, the influences among the aforementioned elements have been changed in the unweighted supermatrix. The original values have been modified of ± 50% in a three-step process (Aragonés-Beltràn et al., 2010), resulting in several possible combinations which have generated new rankings of alternatives. Table XI represents the results of the analysis where, for each control criterion, the first iteration represents the original values and the corresponding final priorities of the alternatives. The following iterations consider instead the subsequent modifications of influences and the resulting new priorities.
Mention should be made to the fact that for some control criteria less iterations have been performed: this happens in those cases where there is only one element to be put in relationship with the highest priority alternative.
As it is possible to see, there are no relevant changes in the new ranking of the alternatives compared with the original one, except for the few cases which are highlighted in bold.
Finally, in order to test the robustness of the model with respect to the rank reversal of the alternatives (Saaty, 2006), the present paper proposes a third sensitivity analysis consisting in the elimination of one alternative at a time from the original model and in the evaluation of the new results. Table XII thus illustrates, for each subnetwork of the model, the original ranking of the alternatives and the results arising from the elimination of each alternative. From the observation of Table XII it is possible to conclude that the final result of the model is stable since the rank reversal of the alternatives occurs only once in the Costs subnetwork, as highlighted in bold in the table.
Table XII. Sensitivity analysis with respect to the rank reversal of the alternatives
|Eliminated alternative||Environmental aspects||Socio-economic aspects|
| ||S>R = I||S = 0.36 R = 0.07 I = 0.07||S>I>R||S = 0.28 I = 0.16 R = 0.06|
|S||I = R||I = 0.25 R = 0.25||I>R||I = 0.37 R = 0.13|
|R||S>I||S = 0.42 I = 0.08||S>I||S = 0.33 I = 0.17|
|I||S>R||S = 0.42 R = 0.08||S>R||S = 0.40 R = 0.10|
| ||Environmental aspects||Socio-economic aspects|
| ||S>R = I||S = 0.36 R = 0.07 I = 0.07||S>R>I||S = 0.35 R = 0.12 I = 0.03|
|S||R>I||R = 0.26 I = 0.24||R>I||R = 0.42 I = 0.08|
|R||S>I||S = 0.42 I = 0.08||S>I||S = 0.45 I = 0.05|
|I||S>R||S = 0.42 R = 0.08||S>R||S = 0.40 R = 0.10|
| ||Environmental aspects||Socio-economic aspects|
| ||S>I>R||S = 0.18 I = 0.16 R = 0.12||S>I>R||S = 0.19 I = 0.16 R = 0.15|
|S||I>R||I = 0.23 R = 0.22||R>I||R = 0.27 I = 0.23|
|R||S>I||S = 0.23 I = 0.22||S>I||S = 0.29 I = 0.21|
|I||S>R||S = 0.29 R = 0.21||S>R||S = 0.29 R = 0.21|
| ||Environmental aspects||Socio-economic aspects|
| ||I>R>S||I = 0.30 R = 0.15 S = 0.05||R>S>I||R = 0.19 S = 0.16 I = 0.13|
|S||I>R||I = 0.35 R = 0.15||R>I||R = 0.30 I = 0.20|
|R||I>S||I = 0.42 S = 0.08||S>I||S = 0.31 I = 0.19|
|I||R>S||R = 0.38 S = 0.12||R>S||R = 0.29 S = 0.21|