A MULTICRITERIA-BASED LOCATION OF AN INDUSTRIAL PARK IN A DEFINED AREA IN IPATINGA, MINAS GERAIS STATE, BRAZIL Localização de um parque industrial em uma área definida em Ipatinga-MG utilizando a análise multicritério

This paper aims to identify an area of over 5 ha for setting up an industrial park in Ipatinga microregion in Minas Gerais State, Brazil. Ipatinga is both nationally and internationally regarded as a model city in terms of development because of Usiminas Company, which is one of the greatest metallurgic industrial powers of Minas Gerais State. Setting up an industrial park in this framework is essential to attract potential investors, to foster purchase of products manufactured by Usiminas, and thus to promote fiscal incentives to the municipality, which may contribute to Ipatinga socioeconomic development. A multicriteria analysis was carried out, applying fuzzy logic (IDRISE software), ArcGis and images available from Google Earth and made a mosaic. Areas suitable for the industrial park implantation and the most suitable area were found by means of a cost-benefit analysis. Since hiring qualified labor after setting the park up would eventually implicate migration of several people from other regions, a dwelling study was also carry out. The results indicate that multicriteria analysis is an important tool for decision-making throughout the process of assessing and selecting areas to set up striking enterprises.


INTRODUCTION
Increasing urbanization has caused the number of available areas for location of industrial enterprises matching entrepreneurs', environmentalists' and civil society's expectation to decrease.As a result, demands for more accurate technical analyses have enhanced.Industrial location is essentially a process of decisionmaking aiming to compare different space alternatives to set industrial units up or, in a more general way, aiming to identify areas in a territory found to be more suitable to industrial use (Soares et al., 2005).
Ipatinga Municipality is located in Steel Valley microregion.Its population increased by the end of the 50s after the setting up of Usiminas facilities, which is an important attraction point for regional labor.Usiminas Company manufactures steel-derived products (e.g., heavy plates, hot strips, cold strips, metallic and galvanized sheets) for a wide range of application.Its main customers are the following industries: automotive, household appliance, electric motor and compressor, packaging, furniture, civil construction and parts supply in general.Such potential urges proposing a metallurgic industrial park implementation aiming to rationalize costs of raw material flow to the greatest laminate consumers in Southern and Southeastern Brazil.
Building on multicriteria analysis and geographic information systems, this paper outlines the methodology applied for assessing and selecting areas to set an industrial park up in Ipatinga Municipality, Minas Gerais State, Brazil.

REVIEW OF LITERATURE
Several theories concerning location have been postulated by economists and geographers, and Alfred Weber has particularly defined the grounds for industrial location theory.Notwithstanding, this paper builds on multicriteria analysis applying fuzzy logic in order to establish industrial location.

Multicriteria analysis
The relevant features in a process of multicriteria assessment are: evaluation of criteria weights, criteria normalization, and criteria combination.For a better understanding of these topics, including detailed description of feasible methods, see Ramos (2000), Mendes (2000), Ramos & Mendes (2001), Calijuri & Lorentz (2003), and Soares et al. (2005).
In a general way, a location model comprises a group of factors and constrictions covering, on the one hand, goals, objectives and policies drafted A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo throughout planning and, on the other hand, theoretical models regarding each particular use (Ramos, 2000).Constriction is a criterion which limits alternatives reckoned during analyses, whereas a factor enhances or increases the suitability of a given alternative.

Evaluation of criteria weights
A pair comparison method was carried out for this paper.According to Ramos & Mendes (2001), despite its complexity, slowness and usual need for interaction in order to retain an acceptable consistence degree, the results and the very procedure of this method fit perfectly to the industrial location problem.

Criteria normalization
This process allows for normalizing criteria values that are not comparable to each other into the same scale, enabling thus their aggregation (Zambon et al., 2005).Normalization process is essentially identical to the process introduced by fuzzy logic (Calijuri et al., 2002).
The authors also claim that the fuzzy set is a generalization of the ordinary group.It is defined building on a continuous domain, and it degrees relevant range from 0 to 1, or from 0 to 255, after normalization.In the general theory, relevance or affirmative of a given phenomenon is relative.Such a theory affords an appropriate conceptual framework to decision-making, because fuzz logic contributes to reduce subjectivity in choice as well as to increase reasoning in the decision process.
Several functions can be applied to set the range between the minimum point and maximum point in criteria normalization.Criteria scores contribute to decision from the minimum point on, and higher scores bring no additional contribution to decision beyond the maximum point (Ramos, 2000).Some of the most applied functions are: Sigmoidal, J-Shaped, Linear and Symmetrical.

Criteria combination
After normalizing criteria scores in a 0-1 scale (or any other scale), it is possible to aggregate them as determined by the decision rule (Zambon et al., 2005).In the decision processes, the most applied procedures regarding space are Weighted Linear Combination (WLC) and Ordered Weighted Average (OWA) for criteria aggregation.
WLC technique enables total trade-off among factors by means of weighed values, also called factor weights.The risk assumed in the analysis is average, placed exactly between AND (minimum) and OR (maximum) in the Boolean Analysis.
Factors are compound in WLC by applying a weight for each of them, and results are summed to produce a final adequacy map.The most important feature of WLC is factor trade-offs, which means that low adequacy in a factor may be traded-off by a set of good adequacies in others (Calijuri et al., 2002).
According to (Ramos & Mendes, 2001), OWA applies criteria weights carried out in WLC and also includes a set of weights that are not specifically linked to any factors, but are applied to them in an order that depends on factor values after the application of the first set of weights.Calijuri et al. (2002) point out that risk level depends on ranking position of weights and magnitude of their values.The highest values in the first positions represent lower risks, and values in the last positions stand for higher risks.The authors emphasize that, by varying order weights, any combination is possible provided that its sum equals one.
OWA enables a wide range of aggregation.Displacement in order weights from the minimum point towards the maximum point controls risk level, called ANDness.Alternatively, distribution homogeneity in order weights throughout different positions controls overall trade-off level (Calijuri et al., 2002;Soares et al., 2005).
A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo As shown in Figure 1, the result is a mostly triangular strategic-decision spectrum, set, on the one hand, by risk attitude, and, on the other hand, by tradeoff (Eastman, 1998).where n is the total number of factors, i is the factor order, and O i is the order weight for the order factor.

Materials
The following materials were used for this paper: • Data available from the Graduate Program in Civil Engineering of the Federal University of Viçosa; A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo

Location and features of the studied area
Ipatinga is located in Steel Valley microregion, Eastern Minas Gerais State, at 19°20'30'' South latitude, and 45°32'30'' West longitude.It is 217 km far from Belo Horizonte (capital of Minas Gerais) and holds around 200,000 inhabitants.
Usiminas is placed in Ipatinga, and it is one of the largest Brazilian metallurgic industries.The fact that the municipality and its trade depend on the steel industry reveals the importance of this company.Since the building of Usiminas, the area is regarded as one of the most important industrial agglomerates in Brazil.Figure 2 shows the location of Ipatinga municipality in relation to Usiminas.A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo Owing to the excellent location of Southern and Southeastern regions, pre-selected districts of Ipatinga municipality were skimmed off to establish suitable industrial park areas.The objective is to set up industries which may demand Usiminas products as their main raw material and also to attract investors for products manufactured by this company.

Stages of the research
The methodology carried out in this research built on a multicriteria analysis applying fuzzy logic.The following stages were espoused in the research development: (i) criteria establishment; (ii) definition of factors and restrictions; (iii) criteria prioritization; (iv) criteria normalization, by means of fuzzy function; (v) weight attribution to criteria in a weight matrix; (vi) criteria combination; (vii) assessment and definition of suitable industrial areas; (viii) analysis and assessment of the most suitable area.
In the first stage, criteria establishment complied with industrial location theory (Ramos, 2000;Ramos & Mendes, 2001).Criteria, thus, comprised industrial, environmental, and socioeconomic activities.In the second stage, factors and constrictions were outlined building on industrial, environmental and socioeconomic concerns for Ipatinga municipality.Data available from Civil Engineering Department of UFV (Federal University of Viçosa, Minas Gerais State)1 , from the Brazilian Institute of Geography and Statistics (2000), specific dispositions of federal and municipal legislations, and researches on the issue were also taken into account.
In the next stage, criteria prioritization was drawn in function of each factor importance.Socioeconomic features relevant to the municipality and from entrepreneurs' and environmentalists' standpoints were observed at this moment.For criteria normalization, Sigmoid and Linear functions were applied to the fuzzy set.
The criteria were compound according to WLC and OWA procedures in order to run on run macro tool, which displayed the most suitable areas for the industrial park location in Ipatinga.
Satellite images have been carried for environmental, social, economic and agricultural analyzes for the last years.There is a wide range of satellites providing images nowadays, yet they are often very expensive for the Brazilian reality.The use of free images such as CBERS has thus posed a solution for Brazilian researches.However, resolution is still a deadlock for creating maps displaying scales under 1:50.000.For this reason, this study suggests the use of Google Earth images as a tool for territory analyzes, although this resource is possible only for areas available from Google Earth.
Images of the focused area at a previously determined height were stored on the very Google Earth, and Visual Stitcher software was used in order to connect images.After connecting each line, images were displayed in a mosaic.It enabled georeferencing images on Arcgis software in the last stage, and such a georeference was used for research analysis.
The generated images may be worked at a larger scale (1:10.000)provided that several control points are fixed to georeference the image precisely.Moreover, 1 pixel was verified to contain 24in.The potential of this tool could then be verified, and it can generate maps displaying scales superior to 1:10.000, if it is well georeferenced.
In the stage of analysis and selection of the most suitable area, transportation cost was assessed building on a methodology provided by National Department of Roads (DNER, 2003).The following transportation-cost formula was applied for determining transportation cost: Operative Hourly Cost (OHC) divided by Production, that is, Y=OHC/ Production.OHC can be determined either by a A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo particular method or by a method recommended DNER or other institutions.

Factors
Table 1 shows the factors associated to environmental, industrial and socioeconomic criteria, including respective code, description, types of fuzzy functions applied to normalize data, and their control points.
Increasing Fuzzy Sigmoid functions were applied to those factors in which a given area was intended to become suitable from a certain distance (control point b).Alternatively, decreasing linear function was applied for maintaining factors as close as possible to the industrial park location.In this function, the maximum distance of each factor was applied after the generation of maps by means of distance tool.Data used to find both economically active population's socioeconomic and life-quality (considering the best wages) factors were determined according to the census of the Brazilian Institute of Geography and Statistics (IBGE, 2000).
Table 2 shows the following data: Ipatinga total population; population of each district surveyed; economically active population (EAP) of each district.For this analysis, districts were assumed to hold similar EAP values, and calculation basis had a 36.4% value for Ipatinga municipality.

Structure of the assessment model
Criteria assessment was based on Figure 3, adapted from Ramos (2000).In this framework, factors were firstly normalized by means of fuzzy logic.Secondly, weights were ordered conforming to their order of relevance.Factors were compounded pair-to-pair, and then weights were computed and ordered by means of weight matrix.Once results were found, the criteria were combined in MCE module (WLC and OWA), where several scenarios were found.Table 4 shows weights calculated in the weight matrix, where weight values were ordered according to the importance of the normalized factors.
Risk and trade-off analysis was carried out by means of MCE (WLC and OWA) in order to locating the most suitable areas for setting up the industrial park.This analysis provided several scenarios, as shown in Table 5.A deep study drawing on these scenarios was carried out in order to identify the most suitable area.
A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo For comparison reasons, some scenarios were structured in such a way that different risk and tradeoff levels could be assessed, allowing for an analysis involving from the most conservative to the riskiest solutions.
In a decision analysis, choosing low risk criteria turn the analysis very conservative, and a risky option urges large knowledge of the surveyed area.Nevertheless, a little riskier analysis was chosen because of high trade-off values.In other words, the higher the compensation is and the higher the risk is, the more efficient the analysis is likely to be.Therefore, values found in WLC, medium risk and total trade-off, were applied, aiming to compare them with OWA values regarding high risk and high trade-off.Scenario S9, carrying 0.41 risk and 0.90 trade-off, was chosen in this paper.
In the face of the results, areas of over 5 ha were found by running macro tool, aiming to find the most suitable areas.Figure 4 presents the areas posing values of 210, which is the highest score found for suitability.A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo

Analysis and selection of the most suitable area for the industrial park location
Building on environmental, industrial and socioeconomics criteria previously defined and using IDRISI software tool, three most suitable areas were found for the industrial location.
The objective of this paper demanded a wide evaluation comprising economical, environmental and industrial standpoints.Furthermore, nearness to the raw material location (Usiminas) and nearness to highway and railroad for production flow were included in the analysis.
As transportation taxes increase with distance, entrance and exit costs are prevailing location force.In other words, a facility placed between a source of raw materials and a market point will find a minimum transportation cost.This can be thus considered a suitable solution.
For the situation under scrutiny, efficient solutions were found to be those posing the lowest cost and the largest benefit, and impossible solutions are those carrying high costs and low benefits.
In this context, an assessment was carried out considering the current structure of Usiminas.Possible storage costs of the generated products (heavy plates, cold strips, hot strips and coated products) and road and rail transportation costs to Southern and Southeastern Brazil were estimated for this evaluation.The objective was to compare these costs with transportation costs in the future industrial park location.Figure 5 demonstrates the assessment of these costs.The highest suitability was found for the scenario outlined in Figure 5.This one was chosen for the assessment of the most suitable area.Some criteria were applied, and transportation cost was the most relevant economic criterion.It specifically involved the observation of shortest distances from raw materials (USIMINAS) and production flow distances by highways and railroads.Table 6 shows distance values run on Lengh tool and ArcGis software.Other features observed in this study were A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo distance from urban area, existent infrastructure 2 , water stream, and green areas.Table 7 shows cost values for distances from highway, railroad, Usiminas and infrastructure.
2 Infrastructure -Existence of water and sewerage systems and electric and telecommunication network.Once some infrastructurerelated data are missing, some data refer to infrastructure in the urban area edge.Figure 5 points out that areas 2 and 3 are in the urban area edge.Table 7 shows that area 1 presented the lowest cost, the shortest distance from raw material source and the largest distance from urban area.

Area
After the selection of the area, the construction of 200m pavement, industrial wastewater treatment station and water treatment station, and enlargement of electric network (700m of the urban area) were found to be necessary.Moreover, duplication of the access roads was also observed to be necessary, owing to increase in truck traffic.
As stated by Macedo (2005), 1 kg steel production consumes 95L of water.Considering that each 1m 3 of treated water in Southeastern Brazil costs U$ 0.5, the consumption from water supply network of the municipality would not be feasible.Building a Water Treatment Station capturing underground water was then found to be a plausible solution.
As current researches recommend recycling industry-generated solid residues, the construction of an industrial landfill was not necessary.A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo The setting up of the industrial park close to raw material source ascribes for reducing costs generated by keeping raw materials in Usiminas patio and by transporting products to Southern and Southeastern Brazil.It also aims to match economic interest of the municipality and of Steel Valley microregion.

Analysis and selection of the most suitable area for residential location
After the Industrial Park implantation, hiring qualified labors will consequently implicate a great number of people migrating from other places.For this reason, a study of dwelling location was performed, aiming to provide dwellings to future workers.
The same methodology was applied to select the best area conforming to locational and environmental criteria (see data displayed in Attachment 1). Figure 7 shows the areas found for this purpose.The assessment of the best habitation area included the following factors: the shortest distances from the area where the industrial park will be set up, and the shortest distance from locations such as urban area, schools, day nurseries, and care units.The best area was the one closest to the future industrial area, and both areas are relatively close to the infrastructure, which will thus implicate no high costs.Area 1 was selected owing to low cost for employees' displacement towards the industrial park as well as to its nearness to infrastructure, which implicates no need to set up day nurseries, schools and other supporting places.Figure 8 shows variation in costs.A multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo

Costs Analysis
The most suitable areas for setting up the industrial park and for implanting dwellings are located in Bom Jardim district (Attachment 2).The table 8 presents plot-of-land costs in this district, total area of each lot and its respective value, and unitary cost of each dwelling (IBGE, 2006).A useful 80 m² area and total dwelling construction costs were also computed.

CONCLUSION
The results point out that multicriteria analysis is an important tool for decision-making in the processes of assessing and selecting areas for setting up striking enterprises.The model being structured in SIG ambient provided the observer with larger support for locational decision.
Several scenarios were modeled by means of compounding different criteria aggregated on OWA.Value weights were tested by altering risks and tradeoffs, which allowed for strategic-decision-space definition.
An optimal analysis demands a greater tradeoff among factors.As high trade-off values were found in this study, a riskier scenario carrying more suitability in relation to other possible scenarios could be chosen.
The choice of the most suitable area drew mainly on economical criteria, and transportation cost was the most significant one in the analysis.These factors alongside environmental and socioeconomic considerations allowed for identifying the best area for the industrial park setting up.

Figure 2 .
Figure 2. Location of the studied area.

Figure 3 .
Figure 3. Analysis structure according to level and criteria set.Source: adapted from Ramos (2000).

Figure 5 .
Figure 5. Assessment flow of raw material storage and transportation costs.

Figure 6 .
Figure 6.Cost-assessment scenario for selection of industrial area.

Figure 7 .
Figure 7. Map of residential district suitability.

Figure 8 .
Figure 8. Cost-assessment scenario for selecting the residential area.

Table 2 .
Establishment of EAP socioeconomic and life-quality factors.

Table 3 .
Constrictions associated to industrial, environmental and socioeconomics activities.

Table 4 .
Weight values obtained in the Comparison Matrix.

Table 5 .
Scenarios of assessment.

Table 7 .
Synthesis of transportation costs.multicriteria-based location of an Industrial Park in a defined area in Ipatinga, Minas Gerais State, Brazil Saulo Henrique de Faria Pereira , Maria Lúcia Calijuri , Sheila Cristina Martins Pereira , Nolan Ribeiro Bezerra , Maria de Nazaré Costa de Macedo A