electre iii example

E L S E V I E R European Journal of Operational Research 98 (1997) 19-36

EUROPEAN JOURNAL

OF OPERATIONAL RESEARCH

T h e o r y a n d M e t h o d o l o g y

Choosing a solid waste management system using multicriteria decision analysis

J o o n a s H o k k a n e n a, * P e k k a S a l m i n e n b

a Paavo Ristola Ltd., Consulting Engineers, V~ini~nkatu 6, FIN-40100 Jyv~skyl~t, Finland b University o f Jyvfiskylfi, P.O. Box 35, FIN-40351 Jyvfiskylfi, Finland

Received January 1995; revised October 1995

Abstract

We report on an actual application of the ELECTRE III decision-aid in the context of choosing a solid waste management system in the Oulu region, Finland, in 1993. The Electre III method proved useful, especially when dealing with environmental problems involving many decision-makers, and in cases where the outcomes of the various alternatives remain to some degree uncertain. One of the main conclusions of our study is that all the proper landfill capacity available in the planning region should be used up. In addition, the energy potential of waste should be utilized within the region. Therefore, the solution recommended for a solid waste management system was intermediate landfilling, composting and RFD-combustion. The decision-makers commented positively on the method used and were satisfied with the options recommended. The scheme will be implemented for use from the beginning of the year 1995. © 1997 Elsevier Science B.V.

Keywords: Decision-aid; Multiple criteria; Waste treatment; ELECTRE III

O. Introduction

The urgency of environmental problems has in recent years become general ly acknowledged. More and more effort is therefore being put into working out realistic solutions to such problems. Instead of money-based considerations, there appears to be a growing body of literature reporting on actual applications of mult iple criteria methods. This trend may bring about better solutions to the pressing environmental problems, as the methods employed compel decis ion-makers to take explicit ly into account a variety of other viewpoints apart from the costs involved. This also accords with the spirit o f the

* Corresponding author.

requirements set for environmental impact analyses in Finnish legislation.

A variety of multicri teria methods has been used in de a l i ng with environmental problems. Merkhofer and Keeney (1987) have employed a traditional multiattribute analysis in determining sites for the disposal of nuclear waste; Leschine, Wallenius and Verdini (1992) consider the problem of locating ocean disposal sites using Pareto Race; Briggs, Kun- sch and Mareschal (1990) have made practical use of the PROMETHEE and G A I A methods within nu- c l e a r Waste management; Stam, Kuula and Cesar (1992) have applied Wierzb ick i ' s reference point method in studying transboundary air pollution in Europe, to mention just a few examples. For solid waste management, the focal point in the present

0377-2217/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved SSDI 0377-2217(95)00325-8

20 J. Hokkanen, P. Salminen / European Journal of Operational Research 98 (1997) 19-36

Table 1 Summary of the solid waste disposal alternatives for the Oulu region

Alternative Co-operation level Treatment method Number of treatment sites

IA Decentralized Landfill 17 landfills IB Decentralized Landfill 17 landfills and 17 composting sites

Open composting IC Decentralized Landfill 17 landfills, 17 composting plants and 1 RDF

Open composting combustion site RDF-combustion

llAa a Intermediate Landfill 4 landfills llAb Intermediate Landfill 4 landfills and 4 composting sites

Open composting IIAc Intermediate Landfill 4 landfills, 4 composting plants and 1 RDF-


IIBa a Intermediate Landfill 4 landfills llBb Intermediate Landfill 4 landfills and 4 composting sites

Open composting llBc Intermediate Landfill 4 landfills, 4 composting plants and 1 RDF-


IICa Intermediate Landfill 3 landfills llCb Intermediate Landfill 3 landfills and 3 composting sites

Open composting llCc Intermediate Landfill 3 landfills, 3 composting plants and 1 RDF-


IIDa Intermediate Landfill 6 landfills IIDb Intermediate Landfill 6 landfills and 6 composting sites

Open composting IIDc Intermediate Landfill 6 landfills, 6 composting plants and 1 RDF-


IIEa a Intermediate Landfill 4 landfills llEb Intermediate Landfill 4 landfills and 4 composting sites

Open composting llEc Intermediate Landfill 4 landfills, 4 composting plants and 1 RDF-


IIIA Centralized Landfill IIIB Centralized Landfill

Decenlralized Composting IIIC Centralized Landfill, composting

and RDF-combustion IIID Centralized Landfill and RDF-combustion

Decentralized Composting

1 landfill 1 landfill and 17 composting sites

1 landfill, 1 composting site and 1 RDF-combustion site 1 landfill and 17 composting sites and 1 RDF. combustion site

a The difference between these alternatives is that each of them has a different combination of cooperating municipalities.

J. Hokkanen, P. Salminen / European Journal of Operational Research 98 (1997) 19-36 21

study, Caruso, Colorni and Paruccini (1993) have used the GFD-method, while Hokkanen et al. (1995) and Hokkanen and Salminen (1994) have applied the ELECTRE II and ELECTRE III methods, respectively.

The ELECTRE methods have proved useful decision-aids in various real applications, e.g. in water resource planning (Roy, Slowinski and Treichel, 1992), comparing energy alternatives (Siskos and Hubert, 1983), weighing different options for a high voltage route (Grassin, 1986) and assessments of nuclear power plant siting (Roy and Bouyssou, 1986; Barda, Dupuis and Lencioni, 1990).

In this paper, we will describe an actual application of the ELECTRE III decision-aid (e.g., Roy, 1991; Vincke, 1992) in choosing a municipal solid waste management system (MSWMS). ELECTRE III was selected as the decision-aid mainly because available environmental data tend to be imprecise in cases like ours. As Electre I I I has proved fairly insusceptible to variations in data and related param- eters (Vincke, 1992), an adequate amount of reliability can be expected of analyses carried out by means of it. Furthermore, according to our own experience (Hokkanen et al., 1995; Hokkanen and Salminen, 1994), the abrupt change from strict preference to indifference, characteristic of ELECTRE II, may in- volve a high degree of risk, if the environmental data on hand are unreliable. ELECTRE III was pro- grammed on a PC based on the descrip6on of the method by Skalka et al. (1986).

1. The problem

1.1. The planning region and feasible alternatives

The ELECTRE III decision-aid was applied to a MSWM problem in the Oulu district in Northern Finland. The planning region consists of 17 municipalities; the total population of the region amounts to roughly 185 000. Our objective was to find the most sensible option for MSWM, a solution that would be applicable until the year 2010. The amount of municipal solid waste in the region adds up to about 80 000 tons/annum (a), out of which 15000 tons, mainly paper and cardboard, was recyclable. The quantities

of demolition waste and industrial waste were 60 000 tons /a and 100 000 tons/a , respectively.

The starting-point for the present study was that each municipality took care of its own waste. The requirements stipulated in the Finnish Waste Act were not met, nor were the instructions given for dealing with municipal waste fully observed, either. In most of the municipalities, waste was gathered in unstaffed and unmonitored 'landfills' - without much concern for releases to the environment.

The following waste treatment methods were considered for the case in hand: sanitary landfilling, incineration and composting. Incineration proved feasible, because in the planning area there are energy producing facilities capable of incinerating the RDF-fraction (RDF = refuse-derived fuel) in an en- vironmentally acceptable way. The amount of RDF was calculated to suffice for an energy production of 72000 MWh, for which peat would otherwise be used. In addition to treatment methods, the present study involved a definition of three levels of cooperation: the decentralized (alternatives I), centralized (alternatives III) and intermediate systems (alternatives II) (see Table 1). All treatment methods, except for incineration, were capable of being used either separately in each municipality or in intermunicipal cooperation at various levels. RDF-combustion was available only in the city of Oulu.

The decentralized system required of each municipality to take care of its own waste. For the intermediate system, the region was divided into realistic 'cooperation areas'. In the centralized system, waste was to be treated at one single plant. The recycling level varied according to the treatment method employed. All the methods considered are in accor- dance with Finnish environmental legislation in effect at present (Ministry of Environment, 1992a) and in the near future (Ministry of Environment, 1992b; Commission of the European Communities, 1991). One objective was to utilize 50% of municipal waste, which included a target set for the level of recycling (30%).

1.2. Decision-makers

In Finland the final decisions on environmental affairs of this order are taken by municipal councils, after hearing proposals made by municipal boards


subordinate to them. The environment and technical committees of the municipalities are responsible for preparations. They will be called decision-makers (DM) throughout our paper. They have a highly significant role, because - as stipulated in the Finnish Waste Act - the funds for solid waste management are drawn from those by whom waste matter is produced. Thus the economic responsibility left to the municipal councils remains fairly small. The

Table 2 Preliminary classification of objectives

Economic (F, L, W, D): Capital cost (F, L, D) Operating cost (F, L, D) Revenue s (L, D) Net cost per ton (L, W, D) Net annual cost per household (D) Financing arrangements (L)

Technical (F, L, W, D): Feasibility (F, L, W, D) Operating experience (L) Adaptability to local conditions (L, D) Reliability (L) Continuous (L) Uninterrupted process (L) Potential for future development (L)

Environmental (F, L, W, D): Global (F, L, D):

Greenhouse effects (F, L) Regional (F, L):

Releases of acidificative compounds (F, L) Surface water dispersed releases (F, L, D) Releases to the air and water with health effects (F, L, W, D)

Local (F, L): Environmental hygiene (D) Surface water dispersed releases (F, L, D) Releases to air and water with health effects (F, L, D)

Political (L, D): Public acceptance (L, D)

Employment (L, D): Number of employees (L, D)

Resource recovery: Products recovered (F, L, W, D) Energy requirements; net effect on primary energy supply (L) Market potential (L) Land usage; volume reduction (L, D)

F = objectives derived from the functional elements. L = objectives found in literature. W = objectives stated in the Waste Act. D = objectives given by DMs.

preferences of the actors who are responsible for preparations define what will be suggested for implementation, therefore we use their weights of importance for the different criteria and call them 'decision-makers'.

In addition, the supervisory body is composed of all the potential interest groups involved: municipalities, municipal councils, regional planning associa- tions and districts o f water and the environment. The supervisory group oversees the project throughout its course.

1.3. Objectives and criteria

In order to outline the criteria, the objectives of the overall task were defined at first, and a preliminary classification of them was made. A preliminary set of criteria was then drafted on this basis. It was submitted to the supervisory group for approval. Thus the final decision on the family of criteria to be used was taken by the supervisory group.

In defining the objectives, we referred to the literature on the subject, including the objectives stated in the Waste Act of Finland (Ministry of Environment, 1992b). The consequences relating to the various functional elements in MSWMSs as reported by Kaila (1987) were studied and weighed in terms of objectives. After modifications carried out by analysts, these aspects were then aggregated to the objectives gathered from the literature. The pur- pose was to find a comprehensive, operational, nonredundant and minimal set of criteria that would represent the various objectives (Keeney and Raiffa 1976). Hundred and thirteen DMs participated in outlining the objectives; they also had an opportunity to add to the list objectives they felt were important. A preliminary classification of the objectives is given in Table 2.

The preliminary classification comprised objectives involving economic, environmental, political, employment and resource recovery viewpoints. The following eight criteria were selected: g l Net cost per ton. g2 Technical reliability. g3 Global effects. g4 Local and regional health effects. g5 Acidificative releases. g6 Surface water dispersed releases.


g7 Number of employees. g8 Amount of recovered waste.

These correspond to the objectives presented. Only political acceptability was excluded as a criterion, because it may in certain situations overlap considerably with some other criteria. Moreover, for an individual DM, the political acceptability of a particular solution does not necessarily depend on the system in question. A landfill plant, for example, can be generally accepted, but problems may arise when the decision concerning its location is being taken.

Certain criteria may also be strongly correlated. This may be due to the existence of some factors which affect both criteria in a parallel way. How- ever, because of the complexity of the links reflected in these factors, internal correlations can hardly be avoided when redefining the family of criteria. Fur- thermore, eliminating one criterion because of its strong correlation with another might destroy information that is not redundant (Vincke, 1992).

1.4. Evaluation o f the criteria

Net cost per ton (gl) This criterion includes all economic objectives

and is operational. It represents the total annual cost in FIM per produced waste ton in a given waste

management system, comprising all costs and rev- enues from onsite storage to final disposal.

The total amount of municipal solid waste was computed on the basis of the current amounts of waste. The calculations of waste container costs were based on the real costs and the real number of containers, as reported by the waste contractors. The costs of collection and transport were also calculated from the current costs in each municipality. Landfill costs were worked out after field and map investiga- tions. Composting and RDF-production costs were reported by the equipment producers. The income from resource recovery was also taken into account when calculating the net costs.

Technical reliabiEty (g 2 ) Estimating the technical reliability of each altema-

tive is by no means a simple task. In this case, such estimates could only be made by experts. The DMs placed particular emphasis on this criterion, however.

For criterion values referring to technical reliability, an expert questionnaire carried out in Uusimaa, Finland (Hokkanen et al., 1995) was drawn upon. In that survey a number of experts have scaled the technical reliability of alternative municipal solid waste systems from 0 to 10.

I

I=11 I

I !

RELATIVE EMISSION FACTOR

[ COMPOS]ING I I INCINERATION I i I

l i ,

II " I I tO s u r f a ~ Io air I I to ~.~tlrfa.~e 1o air

.~.A, I co2,"p ,0"411 . II "g '~ Ic%NzO ,OH,, I

i I : . . . . . . . . . . . . . - . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . t . . . . . .

I I I

RELATIVE GRFENHOUSEEFFECT

I T ~ s P o ~ I

.~,s I ~'+ct II CC, Pb 'g''+ II + % ' 0 x l ICOn' "~° C",, I ' *

I I . . . . t I

,..-.i

I I

.._.1 r

I I II I OF SO 2 NO X OF NITROGEN

Fig. 1. Environmental effects taken into account in the choice of MSWMs.

24 J. Hokkanen, P. Salminen /European Journal of Operational Research 98 (1997) 19-36

Environmental criteria (g3--g6)

Within waste treatment, the main sources of releases are leachate and airborne releases from landfills and other treatment processes (incineration, composting) as well as waste transport.

The environmental aspects were considered in four groups. All MSWMS alternatives produce different types of releases. The environmental effect value was inferred by adding up the relative impact values of each release. That is why the releases were aggregated to the total amount of a particular release or to the relative factor of a particular release (see Fig. 1).

Leachate The environmental effects of leachate depend on

the amount discharged and on the concentrations of specific substances in it. The leachate discharge was estimated on the basis of real landfill areas during the whole lifetime of each landfill and composting site:

L x = C x ~ Q i , (1) i=1

where: L x Release of contaminant (kg). Q Leachate discharge (m3). c x Concentration of contaminant x in leachate

(kg/m3).

n Duration of releases (years). In the Oulu region, the average annual rainfall is

550-650 mm and evaporation 250-300 mm (Na- tional Board of Waters and Environment, Finland, 1987). The average leachate discharge per year can thus be estimated at 300 mm.

The area needed for composting plants was derived from the amount of organic waste and mixture material. Calculations of the future need for landfill

area were based on the factual increase in such area each year. The following assumptions were taken as starting-points for the computations: - The specific gravity of waste varied from 0.8 t / m 3 to 0.450 t / m 3, depending on the machinery used. - The coverage soil was defined as 10% of the waste volume. - The landfill depth was determined on the basis of the actual situation.

The quality of leachate was taken from Ettala (1986) and Ettala et al. (1988), who have studied the quality of leachates in Finnish landfills. The average amount of total nitrogen is 66 mgN/1, cadmium 0.0054 mg/1 and lead 0.029 mg/1. The source separation of municipal waste affects the quality of waste received in landfills. Accordingly, it was assumed to affect the quality of leachate.

The production of leachate continues long after a landfill operation is closed down, as do environmental effects. Therefore post-operation leachate releases were included in the computations. Post-operation phase leachate composition was estimated according to the relationship between landfill age and leachate composition as discussed by Ettala (1986), Belevi and Baccini (1989), Ehrig (1983) and Ehrig and Stegman (1989).

Releases to the air Airborne releases are created during waste trans-

port and treatment. Landfill fires and disturbances in incinerator flue gas purification systems are not within the scope of the present study.

The total amount of gas over the time covered in the plan includes gas production during the operating time as well as post-operation gas releases.

G m = G v ~ mi6gg v, (2) i=1

Table 3 The average composition, content and density of landfill gas (Source: Ettala et al., 1988: Assmuth et al., 1990)

Gas Part of Average content in landfill gas (%) landfill gas m g / m 3

Density k g / m 3

Nitrogen 5 Carbon dioxide 40 Methane 50 Mercury

CxHx 0.0008

218

1.25 1.03 0.56

J. Hokkanen, P. Salminen ~European Journal of Operational Research 98 (1997) 19-36 25

Table 4 The typical emission factors and the range of variation using peat or RDF-waste as a fuel (Bostrt~m et al., 1990; Westas and Westerg~d, 1992; RVF 1993, real emission from J5msSnkoski paper and pulp mills)

Component Emission factor

Peal Variation range RDF Variation range

NO x 235 m g / M J 200-245 150 m g / M j 90-160 CO 2 110 g / M J 100-120 100 g / M J 80-110 N20 30 m g / M J - - SO 2 120 m g / M J 102-157 180 m g / M J 120-400 dust 16.5 m g / M J 2 0 m g / M J 10-60 As 10 /zg /MJ 2 -24 - - Cd 0.5 p~g/MJ 0.07-1.5 0.02 m g / M J 0.02-0.06 Hg 0 .01 /~g /MJ 0.001-0.03 0.1 m g / M J 0.06-0.3 Pb 15 /xg /MJ 2-51 0.1 m g / M J 0.1-0.2

where: G m amount of a gas (kg). G v rate of gas production (m3/t-waste). m yearly amount of waste (t). gv share of a particular gas out of the total volume

of gas. 6g density of the gas (kg/m3). n duration of release (years),

The rate of gas production G~ was assessed at 0.5 m3/ t -was te /a , when the waste amount remains be- low 40000 tons, and 1.0 m3/ t -was te /a , when it exceeds 40 000 tons. In this case the gas amount was 1 m 3 per waste ton only in the landfill receiving the waste of the city of Oulu, the major regional centre. The total amount of releases was based on the composition values shown in Table 3. Post-operation phase gas composition was estimated according to landfill age and gas composition as described by Lagerkvist (1986).

In order to compare the various options with and without energy recovery, releases from alternative energy production were also taken into account. Peat, the primary fuel at the power plants in the region, was selected as the alternative fuel. The amount of airborne releases was estimated on the basis of the typical emission level from RDF-combustion and peat-fired power plants in the Oulu region (Table 4).

The releases from waste transport were calculated for the transport between the cities and the treatment sites. Contaminants originating from exhaust gases were estimated from average truck release coefficients ( g / k m ) and the total transport distance re-

quired. The average truck release coefficients are 5.0 g C O / k m , 15.5 g NOx/km, 1.3 g H C / k m , 1.7 g particles/kin (L~'fikint6hallitus, 1990). The amount of greenhouse gas from waste transport was computed using the product of average truck fuel consumption, thermal value and amounts of various gases. The basic data needed for computations is presented in Table 5.

The environmental criteria chosen will be described one by one in the following.

Global effects (g3) The criterion 'global effects' represents the total

amount of greenhouse effects in each alternative. The greenhouse effect involves the following releases: carbon dioxide (CO2), methane (CH 4) and dinitrogen oxides (N20).

The greenhouse effects (Fig. 1) include releases from landfill, incineration and transport, all of them

i

Table 5 The basic data and typical emission factors for computing the greenhouse effect in waste transport (Bostr~m et al., 1990)

Component Emission factors and other basic data

Thermal value 43 M J / k g Diesel oil:

g CO 2 / M J 74 mg C H 4 / M J 2 mg NEO/MJ 32

Average fuel consumption: compactor 35 l / 100 "kin trailer lorry 50 1/100 km

26 J. Hokkanen, P. Salminen // European Journal of Operational Research 98 (1997) 19-36

airborne. Greenhouse gases absorb infrared radiation in different intensities. In our calculations, the various gases were combined according to their relative greenhouse effect per kg: carbon dioxide (CO2), 1, methane ( C H 4 ) , 70 , and dinitrogen oxide (N20), 200 (NaturvNdsverket, i991).

Relative emission factors were generated for each alternative by using emissions of lead and cadmium, which were combined in relation to their weekly rates not considered harmful to humans: cadmium 0.007 mg/weight in kg and lead (Pb) 0.05 rag/weight in kg (L~SkintShallitus, 1990).

Releases with health effects (g4 ) This criterion consists of those heavy metal re-

leases to air and water which affect health: lead (Pb) and cadmium (Cd) from leachate as well as arsenic (As) and mercury (Hg) from energy utilization. Or- ganic micropollutants, dibenzodioxine (PCDD) and furaans (PCDF), were excluded because of their extremely small amounts (Aittola et al., 1989).

Acidificative releases (gs ) This criterion stands for the total amount of acidi-

ficative emissions. In Finland, the critical emission level of gases causing acidification in nature is almost similar for each of them (Joffre et al., 1990). Therefore, the environmental effect value was calculated by adding up the emissions of sulphur dioxide (SO 2) and nitrogen oxides (NOx). The emissions of

Table 6 The criterion values of the alternatives studied

Alternative Criteria

gl (min) g2 (max) g3 (min) g4 (min) g5 (min) g6 (min) g7 (max) g8 (max) Cost Technical Global Health Acidificative Surface water Employees Resource

reliability effects effects releases dispersed recovery releases

IA 656 5 552678 100 609 1190 670 14 13 900 IB 786 4 539 113 200 575 1190 682 18 23600 IC 912 4 486565400 670 1222 594 24 39767

IIAa 589 9 559780715 411 1191 443 10 13 900 IIAb 706 7 532286214 325 1191 404 14 23 600 IIAc 834 6.5 470613514 500 1226 384 18 40667

IIBa 580 9 560987877 398 1191 430 10 13900 IIBb 682 7 532 224 858 314 1191 393 14 23 600 llBc 838 6.5 466586058 501 1229 373 22 41 747 llCa 579 9 561 555 877 373 1191 405 9 13 900 lICb 688 7 532302258 292 1191 370 13 23 600 llCc 838 6.5 465 356 158 499 1230 361 17 42467

IIDa 595 IIDb 709 llDc 849

9 560500215 500 1191 538 12 13 900 7 532974014 402 1191 489 17 23 600 6.5 474137314 648 1226 538 20 40667

IIIA 579 9 568 674 539 495 1193 558 7 13 900 IIIB 695 6 536936873 424 1195 535 18 23600 IIIC 827 7 457 184239 651 1237 513 16 45 167 IIID 982 7 457206 173 651 1239 513 16 45 167

IIEa 604 9 560500215 500 1191 538 12 13900 IIEb 736 7 532974014 402 1191 489 17 23600 IIEc 871 6.5 474 137314 648 1226 538 20 40667


hydrogen carbons (CH), hydrogen fluoride (HF) and hydrogen chlorine (HC1) were small and thus excluded.

d Demand for a particular waste (t/a). In this case the criterion values of the alternatives are given in Table 6.

Surface water dispersed releases (g6) This criterion covers the surface dispersed re-

leases from landfill and composting plants. In a vast majority of lake conditions, the most important nutri- ent factors causing the shift from a less productive state to a more productive condition are phosphorus and nitrogen (Wetzel, 1982). Nitrogen concentrations in leachates are high and the phosphorus ones low (Ettala et at., 1988). For this reason, nitrogen was chosen to represent the surface water dispersed releases.

Number of employees (g7 ) This criterion refers to the number of employees

dealing with waste treatment. The value for this criterion was defined according to data gathered from operating systems similar to the current ones. The greater number of employees being a plus for and alternative is due to the high unemployment in Finland, therefore this criterion is considered to be maximized.

Resource recovery level (g8) This criterion infers the amount of waste that can

be recovered. There are three possibilities of adding to the amount of recovered waste: 1. More effective source separation of paper and cardboard and other recyclable waste. 2. Point 1 added by source separation of organic waste. 3. Points 1 and 2 added by incineration of RDF-fuel.

The amount of recovered waste (RW) was computed as follows:

RW = 1 O0 1 O0 if d>~ m * p / 1 0 0 * e/100, (3)

d, otherwise,

where: RW Amount of a particular recovered waste (t/a). m Total amount of municipal waste (t/a). e Recovery efficiency (%). p Proportion of a particular waste component out

of the total amount of waste (%).

1.5. The process

During the year 1992 the municipalities in the planning region had acknowledged the problem caused by new Finnish Waste Act. The existing system used in dealing with waste could not be made to meet the requirements of the Waste Act without considerable costs. Therefore a supervisory group was composed of all the potential interest groups in the region for this problem.

In the first meeting, in February 1993, the supervisory group set the objectives for the analysis based on the analysts' suggestion. The analysts described how the analysis will be carried out and a general presentation of the multi-criteria method to be used was given. Also the possible criteria to be used were discussed, and the participants had a possibility to add their own criteria to the preliminary list.

In the second meeting, in June 1993, the current situation was discussed, and alternative possibilities for dealing with the waste in the region were evalu- ated with the DMs and the supervisory group. After this, the criterion values were defined for the accepted set of alternatives. These values and how they were obtained were presented to the DMs and the supervisory group in August 1993.

Starting in September 1993, the weights for the criteria were collected from technical and environmental committees of the municipalities. In this phase the participants had again a possibility to add criteria to the analysis.

In December 1993 the preliminary solution for solid waste management in the region was presented to all municipalities and the supervisory group. The solution was accepted by them and during winter/ spring 1994 all the municipal councils, one by one, accepted the solution suggested. The implementation of the solution will start from the beginning of the year 1995.

2. Description o f the ELECTRE III method

The complex problem of a multicriteria choice is usually formulated by using a set of alternatives


A = (a , b, c . . . . . n) and a set of criteria (g l , g2 . . . . . gin). In this case the criteria are real- valued functions defined on set A so that g j (a i) represents the performance or the evaluation of the alternative a ~ A on criterion gj. Depending on whether the target is to maximize or to minimize the criterion gj(ai), the higher or lower it is, the better the alternative meets the criterion in question. Conse- quently, the multicriteria evaluation of an alternative a ~ A will be represented by the vector

g(a) = (g,(a), gz(a) . . . . . gin(a)). The value gj(a) of the j-th criterion for alterna-

tive a is not fixed or known exactly. Its value is affected by three phenomena (Roy et al., 1986): - imprecision, because of the difficulty of determining it, even in the absence of random fluctuation; - indetermination, because its method of evaluation results from a relatively arbitrary choice between several possible definitions; and - uncertainty, because the value involved varies with time.

All these three phenomena are well known at various levels of solid waste management. There is a variety of solutions for modelling these phenomena. The concept of the pseudo-criterion and its two thresholds allow all three phenomena to be taken into account. Thus one is led to introducing so-called indifference and preference thresholds on the criteria used in the comparison of alternatives. Each of the g j ' s taken together with two thresholds denoted by qj and pj, respectively, constitute a pseudo-criterion (Roy and Vincke, 1984; Vincke, 1992).

When using the Electre III method, each alternative is at first compared to the other ones, with the aim of using the three aspects either to accept or to reject, or, most frequently, to assess the outranking relation: 'alternative a is at least as high in the priority order as alternative b ' , or, more briefly, a outranks b or, a S b.

The following critical information is needed for the method: - weights of the criteria; - preference and indifference thresholds; - veto thresholds.

The latter two are generally determined by analysts. However, if it is possible in a real situation involving a limited number of decision-makers, the

DMs may also have a say in fixing the veto thresholds. With a large number of DMs, as in our case, this could not be carried out; the DMs were only asked to attach weights to the criteria.

The evaluation procedures of the ELECTRE III model (Fig. 2) encompass the establishment of the threshold function, disclosure of concord index and discord index, outranking degree, and the ranking of alternatives, which are further elaborated in the following.

Let q(g) and p(g) represent the indifference threshold and preference threshold, respectively.

If g(a) >1- g(b):

(i) g ( a ) > g ( b ) + p ( g ( b ) ) ~ a P b , (4)

(ii) g(b) +q(g(b) ) <g(a) <~g(b) +p(g(b) )

a Q b, (5)

(iii) g(b)<~g(a) < ~ g ( b ) + q ( g ( b ) ) ~ a l b , (6)

where P refers to strong preference, Q weak preference, I indifference, and g(a) is the criterion value of alternative a.

The establishment of a threshold function has to satisfy the subsequent constraint equations:

(i) g(a) > g(b)

g( a) + q( g( a) ) > g( b ) + q( g( b ) ) ,

g( a) + p( g( a) ) > g ( b ) + p( g( b ) )(7)

(ii) For all criteria, p(g) > q(g).

pj(gj (a) ) and qj(gj(a)) can be calculated according to Roy ' s formula (Skalka et al., 1986):

( i ) p j ( g j ( a ) ) = ap + / 3 p g j ( a ) , (8 )

(ii) qj(gj(a)) = aq +/3qgj(a), (9)

where gj(a) is the evaluation value of alternative a on criterion j. pj(gj(a)) and qj(gj(a)) can be solved in such a way that threshold values are (Roy et al., 1986):

- either constant (/3 equals to zero and a has to be determined); or - proportional to gj(a) (/3 has to be determined and a equals to zero); or - of a form combining these two (both a and /3 have to be determined).


com0,oto o,o, j l OoterM,na,,ooo,,,,m,,y, the alternatives A of pseudo-criteria g j ]

h For each alternative a:

va ues of gj (a), pj [gj(a)], qj[gj(a)]

veto thresholds Outranking degree on ] v j ig/a)] each criterion ¢j(a,b) 1

importance indices of Levels of discordance [ ~ the criteria

Dj(a,b) ] Concordance index c(a,b)

T [ The degree of outranking S (a,b) I

I distillation

I Two complete preorders [

I T

I One final preorder j

Fig. 2. General structure of ELECTRE III.

Concordance indeX and discordance index A concordance index c(a, b) is computed for

each pair of alternatives:

1

kacj( a, b) , c(a, b) = "Kj=l

where K = k kj, (10) j=l

where kj is the weight of criterion j, and cj(a, b) is the outranking degree of alternative a and alternative b under criterion j, with:

cj(a, b )=O if g j ( b ) - g j ( a ) >pj(gj (a)) ,

cj(a, b ) = 1 if g j ( b ) - g j ( a ) ~ q j ( g j ( a ) ) ,

and

0 < c j (a , b) < 1

when

qj( gj( a) ) < gj( b) -- gj( a) ~ pj( gj( a) )

(linear interpolation).

(11)

(12)

The veto threshold vj(gj(a)) is defined for each criterion j:

vj( gj( a) ) = a v + flvgj( a). (13)

A discordance index, d(a, b), for each criterion is then defined as follows:

da(a, b) = 0 if g j ( b ) - g j ( a ) <~pj(gj(a)),

dj(a, b) = 1 if gj(b) - g j ( a ) > vj(gj(a)),

(14)

(15)

and

O<dj(a, b) < 1

when

p j ( g j ( a ) ) < g j (b) - g j (a ) ~< v j (g j ( a ) )

(linear interpolation), where pi(gj(a)) is the preference threshold value, qj(gj(a)) is the indifference threshold value and vj(gj(a)) is the veto threshold value for each criterion.

Finally, the degree of outranking is defined by S(a, b):

S( a, b) = c( a, b) if dj( a, b) ~< c( a, b)

(16) V j ~ J ,

1 - dj(a, b) S(a, b) = c(a, b) × I-I b) '

j~J(a, b) 1 -- c( a,

otherwise, (17) where J(a, b) is the set of criteria for which dj(a, b)> c(a, b).

Ranking The exploitation procedure used in ELECTRE III

is generally as follows: - Construct a complete preorder Z~. - Construct a complete preorder Z 2. - Construct the partial preorder Z = Z l A Z 2 as the final result. Z 1 and Z 2 are respectively constructed through a descending distillation procedure and an ascending distillation procedure (for details of these procedures, see, e.g. Maystre et al., 1994; Vincke, 1992). The two rankings, Z~ and Z 2, are commonly not the same. The final order could be obtained after the


downward order and upward order are averaged, that is,

Z=½(Z 1 +Z2).

However, in case the two rankings are not 'close' , one is not able to build an acceptable complete preorder. In this paper we use the procedure in which alternatives a and b are considered incomparable in case their order does not remain the same in both rankings.

3. Reasons for using ELECTRE III in this problem

On the basis of the literature, four basic ap- proaches to multiple criteria decision problem can be found: 1) multiattribute utility theory (e.g., Keeney and Raiffa, 1976); 2) analytic hierarchy process (e.g., Saaty, 1980); 3) outranking methods (e.g., Vincke, 1992); 4) interactive procedures (e.g., Steuer, 1986). Each one is different in t e r m s of collecting the preference information from the DMs, in modelling the preferences and in producing the final solution. In our case we identified the following constraints for the method to be applied:

1. The number of DMs must not limit the use of the method. In this case it was not possible to identify a DM or a small group of them. The number of DMs participating in the decision process was 113, each with differing opinions. The task was to produce an acceptable compromise solution for them.

2. The decision method must be quick and easy to use. In real applications the DMs may not give much time to the analyst to suggest a solution.

3. The method should need as little preference information as possible. The DMs pointed out that their time is limited as regards concentrating on this particular problem among all other duties they had. On the other hand, some of the criteria would appear difficult to understand for any DM. For example, the environmental criteria were measured in a way that may be difficult for nonexpert DMs. Consider, for example, tradeoff questions between criteria 3 and 4.

4. The method must have the capacity to deal with imprecise data. Most of the criterion values of the alternatives were imprecise, especially the estimates of different releases to the environment. Therefore it is not possible to adopt a method which uses strict criterion values.

Providing reliable decision-aid to this problem, while fulfilling all these requirements seemed quite demanding for any decision method. None of them is perfect. The first three constraints make it difficult to use any decision method, which would need much, and reliable, preference information from the DMs. The DMs had no time and /or did not want to participate in assessing value/utility functions, or perform any pairwise comparisons. However, they still insisted on expressing their opinions about the importance of each criterion. In this case, MAUT and AHP cannot be considered as realistic decision- aids for them. Instead, the ELECTRE methods have been developed for this kind of situations. Obvi-

Table 7 The basic data for computing the threshold values needed for the socio-economic criteria

Coefficients Criteria

g i g2 g3 g4 g5 g6 g7 g8 Cost Technical Global Health Acidifi- Surface Employees Resource

reliability effects effects releases water recovery dispersed releases

aq 30 1 4948770 - 7 6 --2730 0 3 - 2 0 3 7 ~q 0.02 0 0.086 0.42 2.39 0.21 0 0.21 ap 50 3 1647400 - 179 -- 9891 0 10 - 3588 ~p 0.05 0 0.028 0.95 8.62 1.27 0 0.41 ~v 91.4 3 4671800 278 494 621 20 7466 ~v 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5


ously, they lack the theoretical richness of, e.g., the MAUT, but they can be successfully used as decision-aids in certain problems. They are easy to dis- cuss with the DMs, and no difficult preference questions are needed. Clearly, the price for this is that a great deal of preference information (which was not possible to obtain in this case) is not taken into account; the only thing available is the importance index of each criterion.

The choice between the different ELECTRE methods, on the other hand, was easy in this context. We opted for ELECTRE III, since it can easily take the imprecise data into account. The ELECTRE IS method could also be used for Searching the best compromise in this problem. However, the DMs wanted to have the ranking of the alternatives, which it does not provide.

4. J u d g m e n t s n e e d e d b y t h e E L E C T R E I I I decision-aid

4.1. The thresholds

The preference and indifference threshold values were computed for each criterion as shown in Table 7. The relationship between the criterion outcome and possible error range was inferred with the help of a regression analysis on the criteria g3, global effects; g4, releases with health effects; gs, acidificative releases; and g8, resource recovery level. The error range was computed with the help of variation in tile basic data (Table 4). The greenhouse gases from landfills were assumed to variate between 0.2-1.5 m3/ton. The maximum values for heavy metals in leachate were used as the basis for strict preference, while the medium Values were taken as the basis for indifference (Assmuth et al., 1990). In acidificative releases, the indifference values were assumed to be about 30% of strict preference values. Resource recovery is mainly based on source separation and, on the other hand, on centralized separation in certain alternatives. More effective sorting may increase the risk of bad quality of sorted waste. Currently, about 25% of paper and cardboard is sorted incorrectly. The amount of organic waste sorted incorrectly varies between 10-50% and that of light fraction (= waste component available for incineration) ranges between 30-60% (Berg, 1993).

g l: net cost per ton. Thresholds can be inferred from the observed fact that the estimation techniques used with this criterion can lead to an error of about + / - 25% in capital costs and + / - 10% in operating costs. The total error of 10% is quite usual.

g2: technical reliability. Taking into account the standard deviations in the data (around 1) (Hokkanen et al., 1995), the threshold values were defined as follows: - the difference of one was not considered convinc- ing evidence of a preference, and - the difference of two or more was taken to imply strict preference.

g6: sur face w a t e r dispersed releases. The variation of nitrogen in leachate is usually very large, from 50 mg/ l to 150 rag/1 (Ettala, 1986; Ettala et al., 1988). To take into account this linearly dependent error range, we assumed that indifference remains up to 80 mg and strict preference starts from 150 rag/1.

gv: number of employees. The need for employees can vary especially due to changes in resource recovery and transport, because: - technical developments in waste transport and treatment may be reflected in the demand for employees, and

the market in recycled waste may fluctuate over time.

The veto values for all criteria were computed as the sum of the strict preference value and one half of the criterion outcome.

4.2. The weights

The weights, or indices of relative importance, are used in ELECTRE III to indicate the significance of a certain criterion to the DM. However, to capture the importance Of a criterion with a single number may not be easy for the DMs. For a discussion of the weights, see, e.g. Vincke (1992, p. 112) and Gold- stein (1990). In a case like the MSWMS-problem, the number of DMs is often large and they do not give equal value to the individual weights. Thus, to be able to make use of the information on the importance of the various criteria, an inquiry needs to be carried out. After that the data need to be formulated in some sensible manner, so as to obtain the overall weights of the group.


In this study the group of subjects consisted of 113 DMs in charge of environmental and technical affairs in 17 municipalities in the Oulu region. The paper-and-pencil version was used to obtain the weights for each criterion ( g l . . . . . gs)- Dubious as we were about the DMs' ability to determine weights, we decided to use two different procedures for ac- quiring the weights. This was done although in our earlier paper (Hokkanen and Salminen, 1994) the given weights were almost equal with both procedures. The procedures were as follows: 1) the DMs were asked to assign the criteria weights ranging from 1 to 7, 7 being the most important; 2) the DMs were asked to assign number 1 to the least important criterion, and then base the other importance values on how many times more important they appeared than the least important criterion. Thus, if a criterion was considered, say, 3 times more important than the least important one, 3 was the value to be assigned to it. However, again in the present inquiry, the average and median weights showed only minor differences between these two different procedures. In the analysis we will use the weights taken from type 1 questioning.

The final weights were determined on the basis of majority. The highest (lowest) weight which could obtain a majority (i.e., 57 DMs out of 113) was considered to be the weight of the group.

The values obtained were: gl: 0.27, g2: 0.26, g3: 0.016, g4: 0.096, gs: 0.047, g6: 0.09, g7:0.05 and gs: 0.14.

It is difficult to judge, how well the weights given correspond to the DMs' actual opinions. However, the sensitivity analysis is designed to capture the possible biases in the given weights.

5. Result of the analysis

The ranking of alternatives by ELECTRE III (Fig. 3), indicates the benefit of a certain degree of intermunicipal cooperation (e.g., alternatives IIBc, IIAc, IICb). Centralized systems may not be the best solutions in all cases, even if the Finnish Ministry of Environment has set them as its target (alternatives IIIA, IIIB, IIIC, II1D). In an area with not much proper landfill capacity available, all of it should be used up. The energy potential of waste should also be utilized within the region. Such emphases led us to suggest the choice of intermediate landfilling, composting and RDF-combustion (IIBc).

6. Sensitivity analysis and discussion

The ranking of alternatives remains, nevertheless, dependent on the values of the various thresholds

Table 8 The values used in the sensitivity analysis

The factor Criteria

changed gl g2 g3 g4 g5 g6 g7 g8 Cost Technical Global Health Acidificative Surface water Employees Resource

reliability effects effects releases dispersed recovery releases

Weights': min 0.25 0.2 0.02 0.01 0.005 0.11 0.04 0.06 max 0.35 0.3 0.39 0.20 0.04 0.18 0.08 0.17 No. of steps 16 11 21 22 14 11 5 12

Olp or [~p " min 0.01 2 0.1 0.5 8.62 0.3 4 0.4 max 0.1 5 1.5 2 8.62 1.3 15 2 No. of steps 10 4 15 31 1 11 12 33

O~p or ~q "

rain 0.01 1 0.1 0.05 2.22 0.1 l 0.1 max 0.03 2 0.6 0.6 2.39 0.21 5 0.4

No. of steps 3 2 l I 12 18 12 5 7


FINAL PREORDER OF ALTERNATIVES

1

[5~-] 10

Fig. 3. The final preorder of alternatives.

and indices of importance. A sensitivity analysis is therefore recommended, so as to high-light which priority order is convincingly justified by the model, despite all the elements of arbitrariness involved (Roy et al., 1986). In the present study, the analysis was carded out on the basic data. The aim was:

1. To find out how the accuracy of the basic data is reflected in the order of the various alternatives. This was tested by changing /3p and j~q values for each criterion separately by certain steps. /3 's being zero, the sensitivity analysis was done by changing ap and aq. The minimum and maximum values and number of steps are shown in Table 8.

2. To find out how changes in the weights would affect the ranking of the alternatives. In subsequent analysis, the lower and upper quartile values of the weights will be used in judging the sensitivity of the obtained solution (Table 8).

3. To find out the effect of the veto values, when their level exceeds that of credibility Cj(a, b).

The solution obtained is very stable for changes in ~p, flq, •p and aq, which all influence the threshold values. The best solution (IIBc) maintained its posi- tion all the time. The only relevant change took place when the preference threshold was considerably low- ered in criterion g6: in this case alternative IIBc was ranked highest and IIAc second.

The solution is fairly insensitive to changes in the weights. Among the originally highest ranked alternatives, the order remains the same.

Changes in the values of /3 v, which, in turn, affect discordance levels, have a considerable effect on the final ranking. When the effect of discordance was removed from the model, IIBb was ranked first, and the alternatives with regional landfills (IIAa, IIBa, IICa, IIDa, IIEa) tended to come in the second place. This change was mainly due to criterion g4, resource recovery level. The lower discordance of this criterion increased the acceptability of alternatives with a low resource recovery level. However, the Finnish Waste Act includes the goal of a higher resource recovery level. Therefore, the original solution was accepted. The transitional stage before the RDF combustion can be carried out with the alternative IIBb.

A survey was conducted to find out the DMs opinions about this decision-making process. In general, all of them had a positive attitude towards the method used, considering it a fine tool for these kinds o f problems. The DMs pointed out that the time they could spare for the process was limited. In their opinion, any other task(s) in addition to evaluat- ing the weights would have been too laborious. This implies that any method requiring more preference information might not be accepted in a decision- making environment of this kind. Also, the DMs were not willing to pay for more time-consuming procedures.

The DMs showed little interest in the applicability of the method used. Instead, their main concern was that all the criteria they found pertinent would be included in the analysis. The DMs did not take much interest in the absolute values of the criteria; rather they concentrated on the differences between the criterion values. Some of the DMs would have liked to know the exact distance between the various alternatives in the final ranking, so as to fully grasp the significance of the difference between the highest


ranked alternatives. However, distance measures could not be inferred with the method used. The DMs also brought up a number of points they found positive: the method was considered to minimize both the time needed for decision-making and the possibilities of 'politicking'.

The discordance concept was considered important. However, the DMs suggested that the analysis should first be made without discordances. After that, it should be studied how the solution changes when the veto values stipulated in environmental legislation are taken into account, This comment may reflect the DMs' interest in those stipulations; it may be indicative of a need to find out whether there is any 'sense' in the Finnish environmental legislation.

The authorities were satisfied with the decision-aid applied. One of the main reasons was that the method forced them to approach the problem from a wide variety of relevant viewpoints besides the costs involved.

7. Conclusion

From the preceding results and discussion, the following can be concluded:

1) The ELECTRE III method has proved a useful tool in the choice of a MSWMS. Using this method, imprecision in the basic data could be taken into account. This feature is of great importance, especially in environmental applications. In the case discussed here, there is no single DM whose preferences would constitute a basis for the analysis; instead there is a large number of individuals who participate in the process. As the DMs' contribution to a ELECTRE III process can be limited to their assigning weights to the various criteria, the method lends itself very well to cases involving a large number of participants. Furthermore, when using this method, the procedure fulfils the requirements set for environmental impact analyses within the Finnish environmental legislation.

2) The results show that in the choice of a municipal solid waste system, a certain degree of intermunicipal cooperation is advisable. In the Oulu region studied here, only 4 to 5 out of the 24 landfills operating in the area were up to standard. For that

reason, all of the proper landfill capacity should be used up and the energy potential of waste should be utilized within the region. These considerations led to the choice of intermediate landfilling, composting and RFD-combustion. When the solution presented in this paper is compared to the current practice, our estimate is that FIM 60 million (over $10 million) will be saved during the time covered in the plans (to the year 2010). The increase in the amount of recovered waste can be estimated at 30% (total amount 60%). Furthermore, there will be a smaller amount of releases to the environment.

3) Within municipal decision-making, the DMs' attitudes towards a decision-aid appear to correlate with the extent they are required to contribute to it. The less the better, is the general attitude.

4) The method lent itself well to presenting the various aspects of the problem to be dealt with, and it also proved useful in working out a satisfying solution. Another advantage brought up by the DMs was the modest time needed for the process. In addition, the participants appeared to prefer simulta- neous handling of all the criteria and consistent structuring of the problem. This was considered to prevent the participants from indulging in 'politics', with only a subset of criteria to be discussed.

Acknowledgements

The authors wish to thank Dr. John T. Buchanan, University of Waikato, New Zealand, and three anonymous referees for helpful comments. This research is supported, in part, by the Academy of Finland.

References

Aittola, J.-P., Viinikainen, S., and Roivainen, J. (1989), "The emissions of PCDD/PCDF's and related compounds from co-combustion of RFD with peat and wood chips", Chemo- sphere 19, 353-359.

Assmuth, T., Poutanen, H., Stranberg, T., Melanen, M., Penttil~i, S., and Kalevi, K. (1990), "Kaatopaikkojen ongelmaj~ttteiden ymp~irist5vaikutukset. Riskikaatopaikkatutkimuksen p~i~ira- portti" (Environmental impacts of hazardous wastes in landfills), Vesi- ja Ymp~iristShallitus julkaisuja, Sarja A n:o 67 (National Board of Waters and the Environment, Series A, No. 67), Helsinki.

J. Hokkanen, P. Salminen /European Journal of Operational Research 98 (1997) 19-36 35

Barda, O.H., Dupuis, J., and Lencioni, P. (1990), "Multicriteria location of thermal power plant", European Journal of Oper- ational Research 45, 332-346.

Belevi, H., and Baccini, P. (1989), "Longterm assessment of leachates from municipal solid waste landfills", in: Second International Landfill Symposium Proceedings - Volume 1, 9-13 October 1989, Porto Conte (Alghero), Italy, C.I.P.A., Milano, Italy.

Berg, P.E.O. (1993), K~llsortering, Teori, Metod och Imple- mentering (Source Separation, Theory, Methodology and Im- plementation), Chalmers Tekniska H~Sgskola, Avhandling nr.9, G~teborg, Sweden.

Bostr~Sm, S., Backman, R., and Hupa, M. (1990), "'Energiatuotan- non ja -kulutuksen kasvihuonekaasujen p'~i~istiSt Suomessa" (Greenhouse gas emissions from energy production and consumption in Finland), KTM Sarja D:186, Helsinki.

Briggs, Th., Knnsch, P.L., and Mareschal, B. (1990), "Nuclear waste management: An application of the multicriteria PROMETHEE methods", European Journal of Operational Research 44, 1-10.

Caruso, C., Colorni, A., and Paruccini, M. (1993), "The regional urban solid waste management system: A modelling approach", European Journal of Operational Research 70, 16-30.

Commission of the European Communities (1991), "Proposal for a Council Directive on the landfill of waste", COM (91) 102 final, SYN 335, Brussels.

Ehrig, H.J. (1983), "Quality and quantity of sanitary landfill leachate", Waste Management and Research 1, 53-68.

Ehrig, H.J., and Stegmann, R. (1989), "Leachate production and quality - results of landfill processes and operation", in: Second International Landfill Symposium, Proceedings - Vol- ume 1, 9-13 October 1989, Porto Conte (Alghero), Italy, C.I.P.A., Milano, Italy.

Ettala, M. (1986), "Snow cover and maximum leachate discharge of a sanitary landfill", Aqua Fennica 2, 187-202.

Ettala, M., Rahkonen, P., Kitunen, V., Valo, O., and Salkinoja- Salonen, M. (1988), "Quality of refuse, gas and water at a sanitary landfill", Aqua Fennica 1, 15-28.

Goldstein, W.M. (1990), "Judgements of relative importance in decision making: Global vs. local interpretations of subjective weight", Organizational Behavior and Human Decision Pro- cesses 47, 313-336.

Grassin, N. (1986), "Constructing "population' criteria for the comparison of different options for a high voltage line route", European Journal of Operational Research 26, 42-57.

Hokkanen, J. (1993), "J~itehuoltoj~irjestelmSn valinta monitavoit- teista menetelm~i~i hyv~iksik~iytf~ien" (The choice of a solid waste management system using the multicriteria methods ELECTRE I and II), Licenciate thesis, Dept. of Ecology and Environmental Management, University of Jyv~iskyfft, Jyv~iskyl~i, Finland.

Hokkanen, J., and Salminen, P. (1994), "The choice of a solid waste management system by using the ELECTRE III decision-aid method", in: M. Paruccini (ed.) Applying Multiple Criteria Aid for Decision to Environmental Management, Kluwer Academic Publishers, Dordrecht.

Hokkanen, J., Salminen, P., Rossi, E., and Ettala, M. (1995), "The choice of a solid waste management system using the ELECTRE II decision-aid method,', Waste Management and Research 13, 175-193.

Joffre, S.M., Laurila, T., Hakola, H, Lindfors, V., Konttinen, S., and Taalas, P. (1990), "On the effects of meteorological factors on air pollution concentrations and deposition in Fin- land, , in: P. Kauppi, P. Anttila and K. Kentt~imies (eds.), Acidification in Finland, Springer-Veflag, Berlin, 43-94.

Kaila, J. (1987), Mathematical Model for Strategy Evaluation of Municipal Solid Waste Management Systems, VTT Publishers 40, Espoo.

Keeney, R.L., and Raiffa, H. (1976), Decisions with Multiple Objectives: Preferences and Value Tradeoffs, Wiley, New York.

Lagerkvist, A. (1986). Om Nedbrytnings- och Transportprocesser i Avfitllsupplag, Licensiatsuppsats, HiSgskolan i Lule~i, Lule~.

L~i~tkintbhallitus (1990), "Yleiskirje n:o 1977", Helsinki (in Finnish) (National Board of Health 1990: General letter).

Leschine, T.M., Wallenius, H., and Verdini, W.A. (1992), "Inter- active multiobjective analysis and assimilative capacity-based ocean disposal decision", European Journal of Operational Research 56, 278-289.

Maystre, L.Y., Picted, J., and Simos, J. (1994), M~thodes Multi- crit~res ELECTRE, Presses Polytechniques et Universitaires Romandes, Lausanne, Switzerland.

Merkhofer, M.W., and Keeney, R.L. (1987), "A multiattribute utility analysis of alternative sites for the disposal of nuclear waste", Risk Analysis 7, 173-194.

Ministry of Environment (1992a), "Finnish Environmental Legis- lation", General Management Division, Ministry of Environ- ment, Helsinki, Finland.

Ministry of Environment (1992b), "Waste Act", Draft version II, Ministry of Environment, Helsinki, Finland.

National Board of Waters and Environment (1987), Hydrological Yearbook 1981-1983, Publication of the Water Research In- stitute 66, Valtion painatuskeskus, Helsinki, Finland.

Naturv~irdsverket (1991), "Olika gasers bidrag till v~ixthuseffek- ten - en jRmf'trelse" (The relative greenhouse effect of different gases), Naturv~rdsverket, rapport 3647, Solna, Sweden.

Roy, B. (1991), "The outranking approach and the foundations of ELECTRE methods", Theory and Decision 31, 49-73.

Roy, B., and Bouyssou, B. (1986), "Comparison of two decision- aid models applied to a nuclear power plant siting example", European Journal of Operational Research 25, 200-215.

Roy, B., and Vincke, Ph. (1984), "Relational systems of preference with one or more pseudo-criteria: Some new concepts and results", Management Science 11, 1323-1335.

Roy, B., Present, M., and Silhol, D. (1986), " A programming method for determining which Paris metro stations should be renovated", European Journal of Operational Research 24, 318-334.

Roy, B., Slowinski, R., and Treichel, W. (1992), "Multicriteria programming of water supply systems for rural areas", Water Resources Bulletin 28/1, 13-31.

Saaty, T. (1980), The Analytic Hierarchy Process, McGraw-Hill, New York.


Siskos, J., and Hubert, P. (1983), "Multicriteria analysis of the impacts of energy alternatives: A survey and a new compara- tive approach", European Journal of Operational Research 13, 278-299.

Skalka, J.M., Bouyssou, D., and Be mabeu, Y.A. (1986), "Electre III et IV. Aspects m6thodologiques et guide d'utilisation", Document du Lamsade No. 25, 3 e 6dition, Laboratoire d'analyse et Mod61isation de Syst~mes pour 1'Aide h la Deci- sion, Univ6rsite de Paris-Dauphine.

Stare, A., Kuula, M., and Cesar, H. (1992), "Transboundary air

pollution in Europe: An interactive multicriteria tradeoff analysis", European Journal of Operational Research 56, 263- 277.

Steuer, R. (1986), Multiple Criteria Optimization: Theory, Com- putation and Application, Wiley, New York.

Vincke, Ph. (1992), Muhicriteria Decision-aid, Wiley, New York. Westas, H., and Westerghrd, B. (1992), "Biobr'~_nsle i Avfall"

(Biofuel in waste), Sveriges Plastf'6rbund, Stockholm. Wetzel, R.G. (1983), Limnology, Saunders College Publishing,

Philadelphia, PA.

electre iii example

Documents