doi:10.1006/jema.2001.0527, available online at http://www.idealibrary.com on
Multicriteria analysis of environmental quality
in Taipei: public preferences and
improvement strategies
Gwo-Hshiung Tzeng
*†, Sheng-Hshiung Tsaur
‡, Yiou-Dong Laiw
§and
Serafim Opricovic
¶†
Institute of Management of Technology and Institute of Traffic and Transportation, National Chiao
Tung University, 1001 Ta-Hsueh Road, Hsinchu 300, Taiwan
‡
Department and Graduate School of Tourism, Chinese Culture University, Yang Ming Shan, Taipei
§Urban Planning Division, Bureau of Construction and Building, Ministry of the Interior, Taipei
¶ Faculty of Civil Engineering, University of Belgrade, Bulevar revolucije 73, 11000 Belgrade,
Yugoslavia
The public preferences for environmental quality should be a primary consideration of planners and decision-makers in environmental systems planning. In the first stage of multicriteria analysis, a multi-attribute evaluation model for determining public preferences is formulated. The environmental indices are defined for a comparison of environmental quality in different metropolitan districts. The public preferences of the environmental quality in Taipei are obtained using the weighted average rating method. The results indicate air quality and noise pollution as main public concern. In the second stage of multicriteria analysis, strategies are proposed to improve the air quality, and criteria are established. The experts evaluated all alternative strategies according to the criteria. The alternatives are ranked applying the compromise ranking method.
2002 Published by Elsevier Science Ltd
Keywords: environmental quality, public preference, improvement strategies, multicriteria analysis,
compromise solution.
Introduction
The concentration of population and industrial development in many urban areas has led to dete-rioration of the metropolitan environment and increasing public concern about environmental quality. In Taiwan the percent of PSI>100 (Pol-lutant Standards Index) was around 67% (percent of days) between 1984 and 1991, indicating poor (bad) air quality. The Taiwan Environmental Pro-tection Administration (EPA) therefore developed managerial policies based on both administra-tive controls and economic instruments (Chen and Fang, 1998). The Taiwan Air Quality Monitoring
ŁCorresponding author. Email: [email protected]
Network (TAQMN) was completed in 1996, with 66 air quality monitoring stations, including 50 new locations. To meet the public needs for a better air quality, the EPA has initiated a plan to supple-ment the current administrative controls. Public concern about environmental quality has forced the development of evaluation models and the imple-mentation of improvement strategies. The percent of PSI>100 dropped to 6% currently (from 67%).
A number of environmental evaluation models have been previously developed (Teng and Tzeng, 1994; Ribeiro, 1996; Parson, 1997), and they can be distinguished into two types. One includes quantifiable measures with data obtained through monitoring or surveys, and the other is generally based on people’s preferences. In the paper by Par-son (1997), two conventional assessment methods, 0301-4797/02/$ – see front matter 2002 Published by Elsevier Science Ltd
formal models and expert panels, are considered, noting that ‘weaknesses are in how well present models attain their representational goals’. Since in most cases only a few attributes of the environment are quantifiable, collecting data and formulating integrated models are the primary difficulties in applying formal models. As an alternative to for-mal modeling, judgments can be gathered from people who have knowledge and experience rele-vant to the particular problem. In the paper by Margai (1995), manmade environmental hazards and their impact on affected communities are con-sidered; questionnaires were administered by mail to assess the residents’ awareness and perception, and the results showing high levels of environmen-tal awareness, specifically, air pollution and waste disposal were cited as the most significant prob-lems. In Teng and Tzeng (1994), the multicriteria evaluation of strategies for improving and control-ling air quality in the metropolitan area was con-sidered; experts evaluated feasible strategies, and the ELECTRE III model was used to rank alter-natives. Subsequently, the multiattribute utility theory was considered as an applicable approach (Tzeng et al., 1996).
In this paper we develop a procedure to deter-mine people’s response to environmental quality and to find improvement strategies to satisfy the residents. So, instead of a formal model, an assess-ment method based on responses of residents is applied, with the main goal of meeting public needs. This evaluation approach starts from the viewpoint of residents, and derives evaluation val-ues and subjective estimates of relative importance (weights) of the attributes from questionnaires completed by the residents. In other words, a multi-attribute evaluation model for examining environmental quality is formulated through a questionnaire to estimate environmental quality.
These results become the integrated environmental quality indices (Naito, 1986). This public prefer-ence, expressing the people’s response to the exist-ing problem, was considered before formulatexist-ing the criteria for evaluating improvement strategies.
A two-stage model for multicriteria analysis of environmental quality is developed in this paper. In the first stage, the environmental quality anal-ysis is based on the estimation by the residents, expressing their views and feelings. Since the responses and estimates by the residents towards the environment vary due to different activity spaces, designing the investigation to cover the investigated area is critical. The results of the first stage analysis are used to establish the envi-ronmental evaluation indices for the comparative regional analysis. In the second stage, these eval-uation results are used to formulate the criteria and the improvement strategies for environmental quality in Taipei. All alternatives were evaluated by experts, and the ranking of alternatives was per-formed by a multicriteria decision making method. The main contribution of this paper is the development of an assessment method applica-ble to environmental quality analysis, with the main goal of meeting the perceived needs of resi-dents. Based on the public preferences determined, improvement strategies are established and the best compromise measures are determined using a multicriteria decision making method.
Public preference elicitation
technique
In establishing the environmental quality indices for metropolitan regions, a goal hierarchy structure of objectives is first established (Table 1), providing
Table 1. The environmental quality goal structure
Goal Objectives Attributes
Air quality Physical environment Water quality
Land quality Noise
Environmental quality Drainage facility
Public environment Solid waste
Land use cross-interference Park, green areas, open space Space of living
Domestic environment Ventilation Availability of light Quality of drinking water
the relationships between the main goal and the attributes.
Investigation of the metropolitan environmen-tal quality can be performed by asking residents to evaluate the attributes and to assign rela-tive importance (weights) to the attributes. The evaluation (subjective) values of the environmen-tal attributes in each district could be obtained from the questionnaires submitted by the respon-dents. The evaluation data (attributes values and weights) are collected by the procedure similar to the Analytical Hierarchy Process (AHP) developed by Saaty (1980). The AHP method, used in this paper, is presented in the Appendix A.
The evaluation model of the metropolitan envi-ronmental quality applied in this paper is based on the weighting method. In the paper by Ribeiro (1996) the weighting methods were considered as ‘more widely applied to multiple attribute prob-lems’, and they are applicable to problems with the attribute values and relative importance (weights) expressed as linguistic values for presenting the public preference. The model is based on the fol-lowing relations: ud.ai/D Nd X jD1 wijdrijd XNd jD1 wijd; iD1, . . . , n; 8 d .1/ where:
ud.ai/ represents the average evaluation value
of the attribute ai(criterion) for district d; wijis the relative importance of attribute i
given by resident j;
rijis the evaluation value of attribute i
given by resident j;
Ndis the total number of residents participating
in evaluation process in district d,
n is the number of attributes.
For a multiple district (D) or metropolitan area, the following relation was used:
uD.ai/D
X
d2D
.Pd/PD/ud.ai/; iD1, . . . , n .2/
where: Pd indicates the population in district d,
and PDis the population in the entire area D.
This method was applied for the individual districts of a metropolitan area in order to examine
the space distribution of public preferences of environmental quality.
It is preferable to prepare histograms of the data to check their distribution before proceeding with weighted average rating, since the shape of the distribution is best communicated visually. Mea-sures of a distribution’s deviation could be used, such as standard deviation, coefficient of variation, skewness, and kurtosis (see Table 5). Small values of these measures (close to 0) indicate consensus among the evaluators, whereas large negative kur-tosis indicates that the evaluators are far from consensus. If the data are not normally distributed, and the measures of spread are not small, the aver-age value is not representative. If the measures of shape are large (negative kurtosis) for the criteria weights, sensitivity analysis covering the range of weights should be performed within multicriteria decision making procedure.
Public preferences for
environmental quality in Taipei
Metropolitan Taipei area
Metropolitan Taipei has an area of 2325 km2 (Taipei city 273 km2, and outside Taipei city 2052 km2) and a total population of 5Ð9 million (Taipei city 2Ð6 million and outside the city 3Ð3 million) with population density of 2555 per km2 in 1996 (Taipei city 9680 and outside Taipei city 1611). Taipei city is the capital of Taiwan and the center of politics, economics, money market and culture. Urbanization has spread rapidly and dramatically to suburban areas with the growth of the economy and standard of living during the past two decades. The governmental policies and infrastructure investment cannot follow this rapid growth and the changes in residential needs (Chen and Fang, 1998). Tremendous increases in motor transportation have caused deterioration of the metropolitan air quality to an almost intolerable extent. This is exacerbated by the fact that Taipei city is located in a basin, so that emissions are not easily diffused to surrounding areas. Urban residents have realized the seriousness of pollution, especially air pollution created by motor vehicles, and they expect an improvement in environmental quality.
Before 1995 there were years with more than 50% of the days with poor air quality, but envi-ronmental quality in Taiwan has improved since 1995, with the percent of PSI > 100 dropping to
6%. In spite of this general improvement, the environmental quality in Taipei should be better. Very often air quality is moderate (PSI>50), the particulate matter (PM10), and to some extent ozone (O3), contribute to poor air quality. The case study was undertaken (beginning in 1994) to gather data about public preferences and to determine the best compromise measures for envi-ronmental quality improvement, to fulfill human requirements.
Investigation design
This investigation of public preferences about envi-ronmental quality was designed to obtain informa-tion on: (a) the relative importance (weights) of the environmental quality attributes, (b) the evalua-tion values of the environmental quality attributes from very dissatisfied to very satisfied, and (c) the basic information observed by the interviewees (checking the locality of investigation). The num-ber of samples (areas) was determined according to the population in each homogeneous district group. Each district was divided into several inves-tigation squares using city maps, so that the uniformity of random samples within each district was maintained. Furthermore, random sampling was performed in each of the investigation areas so that the samples were selected with an even spatial distribution. The total number of samples in this investigation was planned to be 2500, and a minimum sample threshold of 30 was set up for each administrative district. A total of 2739 valid samples were retrieved.
The public preferences
The questionnaire was designed with seven possi-ble answers for the environmental attributes, i.e. very dissatisfied, rather dissatisfied, mildly dissat-isfied, fair, mildly satdissat-isfied, rather satdissat-isfied, and very satisfied. These linguistic values are simply transformed into the real numbers using the lin-ear scaling as: 0, 16Ð7, 33Ð3, 50Ð0, 66Ð7, 83Ð3, 100, respectively. The evaluation data (attributes val-ues and weights) are collected by the procedure similar to the Analytical Hierarchy Process (AHP) presented in the Appendix A.
Using this environmental quality assessment, integrated evaluation indices were determined by applying equations (1) and (2). The evaluation values are presented in Table 2, but only the data for characteristic districts (those with maximum or/and minimum value). There are 40 districts in the total area. In Table 2, the last three rows contain data for Taipei city, outside Taipei city area, and the metropolitan area.
The results for the districts show that the residents are very dissatisfied with air quality in Panchiao district outside Taipei city (evaluation value of 9Ð04) and in Lungshan district (20Ð25) within Taipei city, both of which are old towns with heavy traffic congestion. Air quality is much better evaluated for Wulai (95Ð60) and Pinglin (93Ð95) districts outside of Taipei city, both are in the mountain areas. According to EPA data, both Panchiao and Lungshan, very often have moderate are quality (PSI>50). The EPA considers Wulai and Pinglin as districts with better air quality. This indicates that the response of residents may have a good correlation with measured data,
Table 2. The evaluation values of the attributes for districts
District Air Pollut. Hilly Noise Drainage Solid Land Parks, Space Ventila- Avail. Drinking quality rivers slopes facility waste use O.space living tion light water Kuting 34Ð41 35Ð61 17Ð50 36Ð58 58Ð05 48Ð11 57Ð22 51Ð49 44Ð54 61Ð30 59Ð96 49Ð48 Shuangyuan 33Ð71 27Ð05 95Ð00 27Ð94 82Ð36 46Ð12 34Ð73 56Ð69 51Ð95 70Ð47 60Ð92 71Ð21 Lungshan 20Ð25 32Ð35 90Ð51 21Ð35 57Ð10 37Ð30 30Ð70 29Ð60 56Ð55 46Ð10 50Ð50 53Ð91 Chenchung 48Ð30 42Ð25 93Ð12 24Ð22 92Ð30 37Ð30 46Ð65 45Ð00 43Ð35 65Ð29 64Ð25 82Ð89 Panchiao 9Ð04 28Ð13 94Ð31 17Ð01 38Ð96 38Ð68 35Ð82 36Ð10 72Ð87 49Ð94 47Ð63 53Ð37 Sanchung 23Ð96 30Ð52 25Ð75 23Ð71 31Ð10 32Ð30 32Ð00 25Ð65 45Ð03 44Ð83 43Ð80 37Ð90 Yingko 39Ð50 90Ð52 51Ð19 39Ð28 49Ð95 49Ð95 47Ð75 44Ð45 54Ð90 52Ð15 58Ð75 56Ð76 Ssnhsia 70Ð30 55Ð31 55Ð78 65Ð86 62Ð05 54Ð35 59Ð30 79Ð65 56Ð55 62Ð60 56Ð55 54Ð35 Tanshui 47Ð75 12Ð31 34Ð00 39Ð50 44Ð68 38Ð12 48Ð50 33Ð54 48Ð67 51Ð88 48Ð21 40Ð88 Taishan 31Ð25 17Ð50 30Ð41 31Ð16 68Ð18 37Ð30 45Ð55 30Ð59 60Ð95 74Ð70 75Ð80 19Ð15 Pinglin 93Ð95 49Ð95 94Ð50 85Ð70 79Ð65 87Ð90 93Ð95 11Ð24 52Ð15 93Ð40 89Ð19 64Ð80 Wulai 95Ð60 74Ð70 79Ð65 82Ð95 86Ð25 87Ð90 85Ð70 34Ð00 76Ð67 89Ð55 88Ð45 87Ð35 Taipei city 37Ð82 38Ð70 49Ð67 37Ð05 59Ð79 47Ð63 51Ð74 46Ð87 54Ð40 60Ð07 57Ð31 55Ð48 Out. Taipei c. 29Ð87 33Ð98 57Ð99 31Ð94 50Ð89 41Ð62 46Ð35 36Ð99 57Ð34 58Ð21 54Ð84 47Ð20 Metropolitan 33Ð68 36Ð24 54Ð00 34Ð39 55Ð15 45Ð00 48Ð94 41Ð73 55Ð93 59Ð10 56Ð03 51Ð17
and the public preferences could be a base for environmental quality analysis.
The ranking based on minimization of average evaluation values of the attributes for metropolitan area is as follows: air quality (33Ð68), noise (34Ð39), pollution of rivers and streams (36Ð24), parks, green area and open space (41Ð73), solid waste (45Ð00), land use cross-interference (conflicts) (48Ð94), the quality of drinking water (51Ð17), conservation of hilly slopes (54Ð00), drainage facility (55Ð15), space of living (55Ð93), availability of light (56Ð03), and ventilation (59Ð10). This does not mean that the residents are not interested other qualities of their living space, but they do consider air quality as the most serious problem. The results indicate that Taipei city, with a total environmental quality value of 46Ð98, is slightly better than the suburban area outside Taipei city, with total environmental quality value of 41Ð81.
The ranking list based on maximization of the average normalized weights (relative impor-tance) of attributes is as follows: air qual-ity (0Ð132), noise (0Ð106), pollution of rivers and streams (0Ð102), solid waste (0Ð087), qual-ity of drinking water (0Ð082), conservation of hilly slopes (0Ð077), space of living (0Ð077), parks, green areas and open space (0Ð073), ventila-tion (0Ð073), availability of light (0Ð070), drainage facility (0Ð062), land use conflicts (0Ð060). These weights also show that air quality problem is very important in Panchiao district (weight 0Ð231), but not so important in Wulai district (0Ð062). The peo-ple living in mountain areas like Wulai do not perceive an air quality problem, whereas residents living in a downtown area like Panchiao have neg-ative perceptions.
The data in Table 2 are informative, showing the public concern about environmental quality attributes. These results show that air qual-ity and noise pollution are the most trouble-some attributes. Both air and noise pollution are caused by the same sources, primarily motor vehi-cles, factories and construction works. The entire Taipei city and part of the densely-populated districts outside Taipei city (Taipei Hsien) are located in a basin and emission gases are not diffused to the surroundings. Most air tion is caused by motor vehicles (mobile pollu-tants), and in some of the suburban districts outside Taipei city the air pollution is addition-ally caused by the manufacturing plants (immobile pollutants). These results point out what mea-sures could be considered to improve the sat-isfaction of residents concerning environmental quality.
Multicriteria analysis of
improvement strategies
The most severe and unsatisfactory environmental attributes according to the evaluation results by residents are considered as a main goal for improvement strategies. The results of the public preference investigation indicated that air quality and noise are considered to be the most serious and dissatisfactory attributes for the metropolitan Taipei. Based on this conclusion from the first stage, and because the sources of air pollution and noise are mostly the same, only the strategies for improving air quality are investigated in the next stage. The multicriteria analysis of improvement strategies consists of four main steps, described below.
Generating improvement strategies
Generating alternative strategies requires devis-ing different scenarios, based on understanddevis-ing problems, experience, knowledge and information for improving environmental quality. Then, rele-vant strategies are defined for achieving the goal of improving the environmental quality. In the Metropolitan Taipei area, air quality improvement is the main goal according to public preference, the result of the analysis in the first stage.
The deterioration of air quality in metropolitan Taipei is caused by different pollutant sources, which can be divided into two categories: mobile pollutant sources, which are emissions from motor vehicles; and immobile pollutant sources, such as factories, construction work, etc. In this investi-gation, strategies to improve air quality are for-mulated through discussions by the experts. Two groups of measures to improve air quality were considered.
Measures to control mobile pollutant sources
(1) Increasing the traffic flow efficiency. Through strategies of transportation system manage-ment, such as improvement of the traffic signal system, and the implementation of one-way lanes; the frequency of vehicle acceleration, deceleration and stops could be decreased, and air pollutants from motor vehicles could be reduced.
(2) Reducing traffic flow during peak hours. Through the adjustment of working hours, such
as using flexible working hours, or by reduction of working days per week, the traffic flow dur-ing peak hours as well as emitted pollutants could be reduced.
(3) Controlling the increase of motor vehicles. Through methods like the increase of license tax and cost of self-owned parking spaces, the number of motor vehicles could be controlled, decreasing motor vehicle emissions.
(4) Improving and encouraging travel by mass
transportation. By improving the mass
trans-portation system and its service quality, the public would be encouraged use it, and the use of motor vehicles would be reduced.
(5) Elimination of old vehicles. Increasing the examination frequency for old vehicles, as well as establishing a unitary insurance system and a fixed-period maintenance, would reduce the air pollution.
(6) Encouraging the use of low-pollution fuel. Through the comprehensive provision of un-leaded petroleum and the use of liquefied petroleum gas, the level of emitted pollutants could be decreased.
(7) Installation of pollution control devices. The uti-lization of low-pollution engines, requirements for catalytic converters on new vehicles, and promotions to install catalytic converters on vehicles in use, would lead to reducing emis-sions.
(8) Intensifying examination of air emission by
motor vehicles. Through the increased efforts
of fixed-period and roadside interception exam-inations of motor vehicles, repair and mainte-nance of motor vehicles would be improved. (9) Establishing rigid pollution emission
stan-dards. Through the enforcement of emission
standards, the owners of motor vehicles would improve their vehicle emission, and to reduce the pollution.
Measures to control immobile pollutant sources
(1) Controlling construction work. Through penal-ties and restriction of the pollution caused by civil engineering and roadside works, the air pollution may be controlled.
(2) Regulating open air burning. By investigat-ing incidents and applyinvestigat-ing pollution penalties for open air burning, air pollution would be reduced.
(3) Rigid regulation standards. Through the en-forcement of emission control standards for
immobile pollution sources, the polluters would decrease the pollution.
(4) Encouraging manufacturers to install
anti-pollution equipment. By implementing policies
such as low-interest loans and tax exemptions, manufacturers would be encouraged to install anti-pollution equipment to reduce the emis-sion of pollutants.
(5) Pollution taxes. Based on the emission quantity according to a pollution standard, taxes could be levied upon polluters as external costs. This tax revenue could be used to compensate victims and compel polluters (manufacturers) to lower their emission of pollutants.
(6) Implementing a system of transferable permit
of emission. Through this anti-pollution
sys-tem, the government decides the quantity of permit emission according to the conceived total emission of pollutants (Chen and Fang, 1998). Under this system, a polluter has to have permits in order to emit a certain amount of pollutants. Such permits are publicly sold so that once the manufacturers producing massive pollution are able to perform more efficient pro-duction and cut down the amount of pollutants, they can reduce pollutants to below the permit level and sell these permits to other manu-facturers while the total amount of pollutants remains the same.
Establishing criteria
Establishing criteria is performed in two steps: (1) acceptance of the set of criteria and defini-tion of each criterion, and (2) establishing the way of expressing the relative importance of each criterion. In establishing the criteria, there are five principles: (i) completeness, (ii) decomposabil-ity, (iii) nonredundancy, (iv) operational feasibil-ity, and (v) minimum size (Keeney and Raiffa, 1976). The decision-maker could assign weight (as relative importance) for each criterion. The decision-makers for this study were environmental protection experts, government authorities, aca-demic research groups and local residents.
Seven criteria (fi, i D 1, . . . , 7) were established
by the decision makers, the first three from implementation aspects, and next three from social aspects.
f1– Implementation costs The cost of
implemen-tation for measures or improvement strategies is likely to be covered by government funds. Higher costs mean that more funding would be required,
and it would be more difficult to obtain the necessary money from the government for imple-mentation.
f2– Cooperation of government administration
This refers to the degree of government admin-istrative cooperation and support for the strategy implementation. Higher evaluation results mean that the government administration would be more cooperative and supportive.
f3– Existing legal acts Higher evaluation results
indicate that the supplement and amendment of legal acts would not be a problem.
f4– Social equity If the implementation of these
strategies would satisfy the public needs, then a higher degree of social equity would be achieved, and the evaluation values are higher.
f5– Acceptance by non-polluters The
evalua-tion value should be high if there is a percepevalua-tion that the polluters (the general public and non-motor vehicle users) will accept the strategy after its implementation.
f6– Acceptance by polluters The evaluation
value should be high if there is a perception that the polluters (polluting manufacturers or users of motor vehicles) will accept the strategy after its implementation.
f7– Amelioration of air quality The evaluation
value should be high if there is a perception that the air quality will substantially improved after implementation of the strategy.
Evaluating the improvement strategies
The final step in evaluation is the generation of a performance matrix. Since quantified information for the improvement strategies is difficult to obtain, the expert evaluation approach was used. Experts may judge the effects of each measure of improve-ment strategy. In this study linguistic values were used to design the evaluation questionnaire.
The group decision-making approach was em-ployed to evaluate the improvement strategies of air quality. The expert group for decision-making consisted of nine primary members invited from related environmental academic institutes, gov-ernment authorities, and assemblies of elected representatives. Six additional experts from aca-demic institutes of transportation were asked to
joint this group because mobile pollution sources involve transportation and motor vehicles. For the immobile pollution sources, the nine evaluators in the first group evaluated the air quality improve-ment strategies. All measures to control pollutant sources were evaluated according to established criteria for improving air quality. The linguistic values: very very high, very high, high, median, low, very low, and very very low, were transformed by scaling into the numbers: 100, 83Ð3, 66Ð7, 50Ð0, 33Ð3, 16Ð7, 0, respectively.
The evaluation results were obtained as the performance matrix (measures versus criteria); in this case there are two matrices, one for mobile polluters and one for immobile polluters. The data obtained are presented in Tables 3 and 4. The first criterion function (f1) has to be minimized, and all others have to be maximized. The results obtained by the Factor Analysis show that there is no dependency between the criterion functions (fi, iD1, . . . , 7).
Within the weight assessment procedure, fifteen experts gave values to the weights. The values given were analyzed in order to test the represen-tativeness of the mean values. The data in Table 5 show that the distribution of weights is right (posi-tive) skewed, with the weight values stretch toward larger values. For instance, for the weight w5there is one value of 0Ð577, but the kurtosis indicates the distribution similar to normal. For the weight
w7, the negative kurtosis shows that there is no consensus among the evaluators; the values of w7 are more evenly distributed (platykurtic shape). The measures of shape and spread are not small, although the weight stability intervals show that sensitivity analysis covering the range of weights is not necessary (see Table 8).
Multicriteria ranking and compromise
solution
Among the numerous approaches available for conflict management, the most prevalent include multicriteria decision making (MCDM). Practical problems are often characterized by several non-commensurable and competing (conflicting) crite-ria, with no solution satisfying all criteria simulta-neously. Applying MCDM, the compromise solution for problem with conflicting criteria can be deter-mined, which can help decision makers to reach a final decision.
The compromise ranking method (named VIKOR) is introduced as one applicable technique to implement within MCDM, and it is presented in
Table 3. The values of criterion functions (for mobile polluters)
Alternatives Criterion functions
f1 f2 f3 f4 f5 f6 f7
1 Increasing traffic flow efficiency 49Ð37 45Ð19 46Ð91 55Ð42 63Ð87 65Ð71 52Ð15 2 Reducing traffic during peak hours 59Ð67 40Ð07 41Ð58 54Ð93 63Ð41 46Ð60 58Ð15 3 Controlling growth of m. vehicles 56Ð27 40Ð66 45Ð96 38Ð23 70Ð49 32Ð45 69Ð98
4 Mass transportation 59Ð48 52Ð60 45Ð03 63Ð45 67Ð55 50Ð59 70Ð05
5 Elimination of old vehicles 47Ð62 48Ð36 58Ð28 50Ð84 64Ð13 40Ð35 60Ð23
6 Use of low-pollution fuel 39Ð46 40Ð86 48Ð44 56Ð00 70Ð67 65Ð32 75Ð12
7 Installation of poll. control devices 54Ð00 55Ð49 61Ð00 59Ð37 78Ð05 40Ð15 71Ð83 8 Intensifying exam. of air emission 55Ð44 51Ð35 48Ð39 57Ð42 73Ð18 44Ð71 49Ð63
9 Establishing rigid standards 36Ð42 52Ð58 58Ð85 67Ð43 79Ð28 36Ð49 65Ð24
Table 4. The values of criterion functions (for immobile polluters)
Alternatives Criterion functions
f1 f2 f3 f4 f5 f6 f7
1 Controlling construction work 48Ð02 48Ð18 52Ð09 72Ð38 80Ð76 33Ð20 58Ð99
2 Regulating open air burning 42Ð57 37Ð50 48Ð82 64Ð17 80Ð76 30Ð20 47Ð56
3 Rigid regulation standards 31Ð70 49Ð89 62Ð51 69Ð93 74Ð84 24Ð91 68Ð11
4 Install anti-pollution equipment 53Ð55 42Ð06 52Ð32 47Ð49 66Ð24 42Ð08 70Ð11
5 Pollution taxes 29Ð87 51Ð16 48Ð69 58Ð97 80Ð08 21Ð41 60Ð23
6 Transferable permit of emission 33Ð10 44Ð36 59Ð30 59Ð51 78Ð33 44Ð58 63Ð79
Table 5. The values of criteria weights
Weights w1 w2 w3 w4 w5 w6 w7 Mean 0Ð070 0Ð164 0Ð157 0Ð118 0Ð108 0Ð062 0Ð320 Median 0Ð069 0Ð077 0Ð134 0Ð107 0Ð046 0Ð035 0Ð297 Coefficient of Variation 0Ð824 1Ð019 0Ð761 0Ð721 1Ð275 1Ð083 0Ð623 Skewness 1Ð371 1Ð059 0Ð995 1Ð323 2Ð545 1Ð789 0Ð597 Kurtosis 1Ð812 0Ð181 0Ð234 1Ð036 6Ð038 2Ð021 1Ð004
the Appendix B. Compromise ranking is performed by comparing the measure of closeness to the ideal alternative (Duckstein and Opricovic, 1980). This method was applied with data (average evaluation values and weights) given by the expert group. The obtained ranking lists are presented in Table 6, for mobile polluters, and in Table 7 for immobile pollutes. There are two compromise solutions for immobile polluters obtained by VIKOR, because the top two are ‘close’.
In addition, the ranking results are obtained by applying another method, the Technique for Order Preference by Similarity to Ideal Solution (known as TOPSIS), which is also a modification of compromise programming (similar to VIKOR). This method was developed based on the concept that, using Euclidean distance, the chosen alternative should have the shortest distance from the ideal solution and the farthest from the negative-ideal solution, and on the entropy method to modify
weights (Hwang and Yoon, 1981). The ranking results (by TOPSIS) are presented in Table 6 and in Table 7. The results obtained using the two methods (VIKOR and TOPSIS) suggest that the differences are due to different weights (TOPSIS uses modified weights). With the same weights the results of the two methods are almost the same.
The VIKOR algorithm determines the weight stability intervals (Opricovic, 1994) for the obtained compromise solution with the ‘input’ weights, given by the experts. The values of weight range (given wmin, wmax) and stability interval (wmL, wmR,
for mobile polluters; and wim
L , wimR , for immobile
polluters) presented in Table 8 show the preference stability of obtained compromise solutions, with a few exceptions, such as wm
R,1, wimR,5, wimR,6, wmL,7.
As a compromise solution, the following measure could be proposed for implementation: mobile polluters should promote and implement the utilization of low-pollution engines and
Table 6. Ranking list of measures to control mobile polluters
Ranking by VIKOR Ranking by TOPSIS
Rank Index Q Measure Rank Index
1. 0Ð0 Installation of pollution control devices 3 0Ð614
2. 0Ð168 Establishing rigid pollution emission standards 4 0Ð542
3. 0Ð333 Mass transportation 2 0Ð628
4. 0Ð345 Use of low-pollution fuel 1 0Ð777
5. 0Ð537 Elimination of ruined vehicles 7 0Ð415
6. 0Ð569 Controlling the growth of motor vehicles 6 0Ð433
7. 0Ð802 Reducing traffic flow during peak hours 8 0Ð357
8. 0Ð864 Increasing the traffic flow efficiency 5 0Ð466
9. 0Ð874 Intensifying examination of air emission 9 0Ð293
Table 7. Ranking list of measures to control immobile polluters
Ranking by VIKOR Ranking by TOPSIS
Rank Index Q Measure Rank Index
1. 0Ð0 Rigid regulation standards 2 0Ð549
2. 0Ð186 Transferable permit of emission 1 0Ð810
3. 0Ð397 Controlling construction work 5 0Ð413
4. 0Ð422 Pollution taxes 4 0Ð464
5. 0Ð432 To install anti-pollution equipment 3 0Ð511
6. 1Ð000 Punishing open air burning 6 0Ð307
Table 8. Weight range and stability interval i wim L wmL wmin wi wmax wmR wimR 1 0Ð0 0Ð0 0Ð008 0Ð070 0Ð236 0Ð143 0Ð749 2 0Ð0 0Ð0 0Ð006 0Ð164 0Ð554 1Ð0 0Ð736 3 0Ð0 0Ð0 0Ð037 0Ð157 0Ð430 1Ð0 1Ð0 4 0Ð0 0Ð0 0Ð022 0Ð118 0Ð336 0Ð312 0Ð699 5 0Ð0 0Ð0 0Ð011 0Ð108 0Ð577 0Ð580 0Ð263 6 0Ð0 0Ð0 0Ð012 0Ð062 0Ð249 0Ð196 0Ð130 7 0Ð0 0Ð152 0Ð096 0Ð320 0Ð685 0Ð685 0Ð808
the installation of catalytic converters, whereas immobile polluters should establish rigid emission standards and implement a system of transferable emission permits. However, all of the top four measures in the ranking lists for mobile polluters may be considered for implementation.
Conclusions
A two-stage model for multicriteria analysis of environmental quality in the metropolitan area is developed. In the first stage, the environmental quality analysis is based on the estimation by the residents, expressing their views. In the second stage, the evaluation results are used to formulate the criteria and the improvement strategies of the environmental quality. The experts evaluated the
alternative strategies, and then the ranking of alternatives was performed by two multicriteria decision making methods.
Environmental evaluation indices are a subjec-tive manner, reflecting the actual responses and preferences of residents towards environmental quality. Their expressed preferences have to be considered by planners and decision-makers dur-ing policy formulation. Public preferences obtained in the first stage of the presented methodology indicate that air quality is the most urgent matter to deal with in order to improve environmen-tal quality in Taipei. The results of the second stage provide a compromise improvement measure which should be implemented in order to improve the air quality. This assessment method is based on data-gathering and on evaluation of alterna-tives by experts, without using a mathematical model of environmental management, and this approach could be considered as a contribution of this paper.
The results of multicriteria analysis indicate
what to do first in order to improve
environmen-tal quality. To answer the questions where and
when to implement the improvement strategies
over metropolitan districts, the research should continued by solving the allocation and scheduling problem under budgetary constraints. The combi-nation of policy measures could be analyzed as a portfolio synergy case.
Acknowledgements
This paper is a partly result of the project NSC88-2811-H009-0001, supported by the National Science Council of Taiwan. The constructive comments of the editor and the reviewers are gratefully acknowledged.
References
Chen, H. W. and Fang, S. H. (1998). Air Pollution Control Fee: The Taiwan Experience. EPA paper. Bureau of Air Quality Protection and Noise Control, Environmental Protection Administration, Taiwan.
Duckstein, L. and Opricovic, S. (1980). Multiobjec-tive optimization in river basin development. Water Resources Research 16(1), 14–20.
Hwang, C. L. and Yoon, K. (1981). Multiple Attribute Decision Making: Methods and Applications. New York: Springer-Verlag.
Keeney, R. L. and Raiffa, B. (1976). Decision with Multiple Objectives: Preferences and Value Tradeoffs. New York: John Wiley & Sons, Inc.
Margai, F. L. (1995). Evaluating the potential for environmental quality improvement in a community distressed by manmade hazards. Journal of Environ-mental Management 44, 181–190.
Naito, M. (1986). Environmental Indices: Applications and Systems. Research Report From the National Institute for Environmental Studies, No. 88. Japan. Opricovic, S. (1994). Preference Stability of
Compro-mise Solution in Multicriteria Decision Making. XI-th International Conference on Multiple Criteria Decision Making. Coimbra, Portugal.
Opricovic, S. (1998). Multicriteria Optimization in Civil Engineering. Belgrade: Faculty of Civil Engineering. Parson, E. A. (1997). Informing global environmental
policy-making: a plea for new methods of assessment and synthesis. Environmental Modeling and Assess-ment 2(4), 267–280.
Ribeiro, R. A. (1996). Fuzzy multiple attribute decision making: a review and new preference elicitation techniques. Fuzzy Sets and Systems 78, 155–181. Saaty, T. L. (1980). The Analytic Hierarchy Process. New
York: McGraw-Hill.
Saaty, T. L. and Vargas, L. G. (1982). The Logic of Pri-orities: Application in Business, Energy, Health, and Transportation. Boston: Kluwer-Nijhoff Publishing. Teng, J. Y. and Tzeng, G. H. (1994). Multicriteria
evaluation for strategies of improving and controlling air quality in the super city: a case study of taipei city. Journal of Environmental Management 40, 213–229. Tzeng, G. H., Chen, J. J. and Yen, Y. K. (1996).
The strategic multicriteria decision-making model for managing the quality of the environment in metropolitan Taipei. Asian Journal of Environment 4(1), 41–52.
Yu, P. L. (1973). A class of solutions for group decision problems. Management Science 19(8), 936–946. Zeleny, M. (1982). Multiple Criteria Decision Making.
New York: McGraw-Hill.
Appendix A
The analytical hierarchy process – AHP
The Analytical Hierarchy Process (AHP) is a systematic procedure for representing the elements of a problem, hierarchically. The AHP method was developed by Saaty in 1971 (Saaty, 1980; Saaty and Vargas, 1982). The procedures, used in this paper, may be summarized as follows:
(1) Structuring the hierarchy, for evaluation
The AHP method is used to make the decomposition (or structuring) of the problem as a hierarchy. In general, the AHP method divides the problem into three levels: (a) the goal for resolving problem; (b) the objectives for achieving the goal; (c) the evaluation criteria for each objective. An example of hierarchy is presented in Table 1.
(2) Constructing the pairwise comparison matrix
After structuring a hierarchy, the pairwise compar-ison matrix for each level is constructed. During the pairwise comparison the nominal scale is used for evaluation.
The scale of relative importance is presented in Table A.1. In the evaluation process (by experts), the AHP questionnaire sheet for environmental quality is used. The sheet for the objective level, as an example, is presented in Table A.2, and it could be used in comparing objectives, one on the left with one on the right side. In Table 1 there are three objectives.
(3) Calculating the weights and testing the consistency for each level
For each pairwise comparison matrix (A), using the theory of eigenvector, i.e. (A-lmax I) w = 0, to calculate the eigenvalue (lmax) and the eigenvector
w = (w1, w2, . . . , wn), weights can be estimated.
The consistency of the comparison matrix was tested and the opinions of the regional decision-maker group were integrated. In the consistency test, consistency index (C.I.) could be utilized to determine the degree of consistency, generally speaking, when C.I. < 0Ð1 it is considered to be acceptable.
Table A.1. Scale of relative importance
Intensity of importance Definition Explanation
1 Equal importance Two activities contribute equally to the
objective
2 Intermediate between equal and weak
3 Weak importance of one over another Experience and judgment slightly favor one activity over another
4 Intermediate between weak and strong
5 Essential or strong importance Experience and judgment strongly favor one activity over another
6 Intermediate between strong and
demonstrated
7 Demonstrated importance An activity is strongly favored and its
dominance is demonstrated in practice
8 Intermediate between demonstrated and
absolute
9 Absolute or extreme importance The evidence favoring one activity over another is of the highest possible order of affirmation
Reciprocals of above non-zero numbers
If activity i has one of the of above non-zero numbers assigned to it when compared with activity j, then j has the reciprocal value when compared with i.
Reasonable assumption
Table A.2. AHP questionnaire sheet for environmental quality Importance Physical 9:1 7:1 5:1 3:1 1:1 1:3 1:5 1:7 1:9 Public Environment 8:1 6:1 4:1 2:1 1:2 1:4 1:6 1:8 Environment Physical 9:1 7:1 5:1 3:1 1:1 1:3 1:5 1:7 1:9 Domestic Environment 8:1 6:1 4:1 2:1 1:2 1:4 1:6 1:8 Environment Public 9:1 7:1 5:1 3:1 1:1 1:3 1:5 1:7 1:9 Domestic Environment 8:1 6:1 4:1 2:1 1:2 1:4 1:6 1:8 Environment
Marking in questionnaire (Table A.2), for exam-ple, 5:1 means that physical environment is of strong importance over public environment.
Appendix B
The VIKOR method
The compromise ranking method (named VIKOR) has been introduced as one applicable tech-nique to implement within MCDM (Opricovic, 1998). Assuming that each alternative is evaluated according to each criterion function, the compro-mise ranking could be performed by comparing the measure of closeness to the ideal alternative. The multicriteria merit for compromise ranking is developed from the Lp-metric used in compromise programming method (Yu, 1973; Zeleny, 1982). The various alternatives are denoted as a1, a2, . . . , aJ.
For an alternative aj the merit of i-th aspect is
denoted by fij, i.e. fij is the value of i-th criterion
function for the alternative aj; n is the number of
criteria.
Figure B.1 illustrates ideal FŁD.fŁ
1, f2Ł/ and com-promise solution FcD.f1c, f2c) within bicriteria prob-lem; a compromise as an agreement established by mutual concessions is illustrated by fc
1f1Ł and fc 2f2Ł. f2* f1c f2c f1* F*ideal Fccompromise solution
Compromise programming method introduced
Lp-metric as an aggregated function. The
develop-ment of VIKOR method started with the following form of Lp-metric Lp,jD ( n X iD1 [wi.fiŁ fij//.fiŁ fi /] p )1/p , 1p1; jD1, 2, . . . , J
Within VIKOR method L1,j(as Sj) and L1,j(as Rj)
are used in formulating ranking merit (‘boundary solutions’). The solution obtained by min
j Sj is with
a maximum ‘group utility’ (‘majority’ rule), and the solution obtained by min
j Rj is with a minimum
individual regret of the ‘opponent’.
Weighting coefficients (weights wi) are
intro-duced to express the relative importance of the criteria. The weights have no clear economic mean-ing, but the use of weights gives the opportunity for modeling the real decision making.
The compromise ranking algorithm VIKOR has the following steps:
(a) Determination of the best fŁ
i and the worst fi
values of all criterion functions, iD1, 2, . . . , n. If the i-th function represents a benefit than:
fiŁDmax
j fij, fi Dminj fij
(b) Compute the values Sj and Rj, jD1, 2, . . . , J, by
the relations SjD n X iD1 wi.fiŁ fij//.fiŁ fi / RjDmax i [wi.f Ł i fij//.fiŁ fi /]
where wiare the weights of criteria.
(c) Compute the values Qj, j D1,2,. . .,J, by the
relation QjDv.Sj SŁ//.S SŁ/C.1 v/.Rj RŁ//.R RŁ/ where: SŁDmin j Sj, S Dmaxj Sj, RŁDmin j Rj, R Dmaxj Rj
v is introduced as weight of the strategy of
‘the majority of criteria’ (or ‘the maximum group utility’), usually vD0Ð5.
(d) Rank the alternatives, sorting by the values
S, R and Q. The results are three ranking lists.
(e) Propose as a compromise solution the alterna-tive (a0) which is the best ranked by the measure Q if the following two conditions are satisfied:
C1. ‘Acceptable Advantage’:
Q.a00/ Q.a0/½DQ
where: a00 is the alternative with second position in the ranking list by Q; DQD1/.J 1/; J is the number of alternatives. (DQD0Ð25 if J4).
C2. ‘Acceptable Stability in decision making’:
The alternative a0also has to be the best ranked by S or by R, or by both, as well. This compromise solution is stable within a decision making process, which could be: ‘voting by majority rule’ (when
v>0Ð5 is needed), or ‘by consensus’ v³0Ð5, or ‘with
veto’ (v < 0Ð5). Here, v is the weight of decision making strategy ‘the majority of criteria’ (or ‘the maximum group utility’).
If one of the conditions is not satisfied, then a set of compromise solutions is proposed, which consists of:
Alternatives a0and a00if only the conditions C2 is not satisfied, or
Alternatives a0, a00,. . ., a.k/ if the conditions C1 is not satisfied; a.k/ is determined by the relation Q(a.k// Q.a0/ ³ DQ (the positions of these alternatives are ‘in closeness’).
The best alternative, ranking by Q, is one with the minimum value of Q.
The main ranking result is the compromise ranking list of alternatives, and the compromise solution with the ‘advantage rates’.
Ranking by this algorithm may be performed with different values of criteria weights wi,
analyzing the impact of criteria weights on pro-posed compromise solution. The VIKOR algorithm determines the weight stability intervals, for the obtained compromise solution with the ‘input’ weights, indicating the preference stability of obtained compromise solution. This is a helpful tool in multicriteria decision making, particularly in the situation when the decision maker is not able to express preference at the beginning of system design. The compromise solution could be accepted by the decision makers because it provides a maxi-mum ‘group utility’ of the ‘majority’ (with measure
S, representing ‘concordance’), and a minimum of
individual regret of the ‘opponents’ (with measure
