A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection

Utilities Policy xxx (2015) 1e11

Contents lists available at ScienceDirect

Utilities Policy journal homepage: www.elsevier.com/locate/jup

A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection Kazim Baris Atici a, *, Ahmet Bahadir Simsek a, Aydin Ulucan a, Mustafa Umur Tosun b a b

Department of Business Administration, Hacettepe University, 06800 Ankara, Turkey Department of Public Finance, Hacettepe University, 06800 Ankara, Turkey

a r t i c l e i n f o

a b s t r a c t

Article history: Available online xxx

This paper deals with site selection problems for wind power plants and aims to propose a structural procedure for determining the most feasible sites. The application area is Western Turkey. The methodology is mainly composed of two stages: the first stage is pre-elimination of infeasible sites, and the second stage is evaluation of the available ones. Geographic Information Systems (GIS) are used to generate layers of data and to apply the elimination criteria and constraints. The alternative land areas are handled in terms of identical-sized grids, which are large enough to install one wind turbine each. Multiple Criteria Decision Analysis (MCDA) is then used to rank and sort the grids via the identified evaluation criteria. The problem is evaluated in 13 fields, which are a collection of several grids in order to evaluate larger scale areas to construct wind farms rather than for individual turbines. The evaluations are made both at grid (micro) and field (macro) levels and both deterministic and uncertain data are used. The results reveal a high level of consensus on the most feasible sites between the different MCDA methods applied. The proposed methodology provides a structured decision aid, which can be applied to other energy site selection problems. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Geographic Information System Multiple Criteria Decision Analysis Renewable energy

1. Introduction Turkey's geostrategic position of being located at the gateway between the East and the West makes it an important player in Eurasian energy affairs. As well as serving as an energy corridor, Turkey itself has been transformed into an energy-hungry country due to the increasing urban population, restructuring of the settlement areas and emerging industrialisation. This naturally brings out a need for new energy resources. Besides the new energy resources needs, Turkey is executing accession negotiations with the EU. The EU aims to get 20% of its energy from renewable energy sources by 2020. The energy resource shift from fossil fuels to the renewable energy sources will reduce Turkey's regulatory burdens in the accession period. More renewable energy will also enable Turkey to cut greenhouse emissions and make it less dependent on imported energy. Considering such an emerging market, which is not garnished with a large amount of hydrocarbon resources, the

* Corresponding author. E-mail addresses: [email protected] (K.B. Atici), abahadirsimsek@hacettepe. edu.tr (A.B. Simsek), [email protected] (A. Ulucan), [email protected]. tr (M.U. Tosun).

efficient use of renewables is much more important since there exists a vast potential of different types of such resources (wind, solar, geothermal, hydro and biomass) all over the land. Wind power is one of the most promising sources of energy especially in the Western part of Turkey. Recently, research efforts and support mechanisms have been growing in order to increase the share of wind-generated power in Turkish network sectors. There has been an increasing interest in installation of wind turbines in different regions by the private sector. By boosting wind power, Turkey will encourage technological innovation and employment. The General Directorate of Renewable Energy (organized under the Turkish Ministry of Energy and Resources) is responsible for engineering services for the investigation of potential resources and the licensing mechanism for the intended power stations is carried out by Turkish Energy Market Regulatory Authority. Recently, the potential of wind energy in Turkey has been investigated and publicly reported by the General Directorate of Renewable Energy as a Wind Energy Potential Atlas (2007). The information presented in the report relies on measurements all over the country and deals with the technical factors to be considered in generating wind power such as the wind speed or the capacity factor. In the current setting, identification of a new wind power site depends on the application made by an entrepreneur in an

http://dx.doi.org/10.1016/j.jup.2015.06.001 0957-1787/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

2

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

intended area, then evaluated by the authority as to if it can be licensed or not. Therefore, the process is application-specific and leans on the question of “Can we install wind turbines in this area?” With all the information available about the wind potential of different areas from the published official reports and geographical information from different resources, our motivation in this paper is to ask the question in a different way as “Which areas are more suitable to install wind turbines?” The initial task before transforming this motivation to a structured methodology is to identify the different stakeholders and their preferences. Straightforward application of decision analysis techniques without a detailed stakeholder and preference investigation may lead to arguable and irrational results. The preferences may be contradictory, therefore any methodology modelling a real world case should look at the problem from different perspectives. Our aim is to deal with the wind power site selection problem, which involves different perspectives such as technical potential, financial feasibility or policies of regulatory authorities. One land area can possess a high wind power potential in technical terms; however, it may be impossible to construct a wind turbine in the given area since it might be declared as protected land by the governmental regulations, or it can be very distant from main roads that cause a very costly construction process. It is important to realize these contradicting perspectives to come up with a robust identification of feasible land areas satisfying the preferences of different stakeholders of wind power plants. In this paper, we mainly benefit from two decision-aiding tools in order to handle this multi-dimensional problem: Geographic Information Systems (GIS) and Multiple Criteria Decision Analysis (MCDA). MCDA techniques are well known to enable the integration of the different decision maker preferences. In this manner, they can be beneficial to cope with the above mentioned contradictions in the wind site decision process. Of course, before deciding on the most feasible areas, it is important to identify the alternatives. GIS play the important role of data generation and unsuitable alternative elimination. We analyse a specific region with the highest wind energy potential in Turkey, and it consists of two districts in the Western part of the country (Balıkesir and Çanakkale). We design a step-bystep process that initially eliminates the infeasible or unsuitable areas and then, ranks or sorts the remaining sites relying on multiple criteria. Although the methodology seems too technical, it produces important policy implications for different parties, which are also discussed throughout the paper. In the literature, GIS-based approaches have been used in Renewable Energy Systems (RES) planning and decisions, which include wind farm siting. Recent examples include Lejeune and Feltz (2008), Aydin et al. (2010), Tegou et al. (2010), Haaren and Fthenakis (2011), Phuangpornpitak and Tia (2011), Sliz-Szkliniarz and Vogt (2011), Zhou et al. (2011), Al-Yahyai et al. (2012), Grassi et al. (2012), Ouammi et al. (2012) and Gass et al. (2013). The studies mostly focused on identifying the most appropriate land areas for installation of wind turbines and developed several criteria relying on different sources, which can lead to a robust selection process. It is very often that the decision-making process is supported with the use of MCDA methods. The current paper contributes to the literature in several ways. We follow a two-stage methodology in the selected region: elimination and evaluation. We integrate two approaches (GIS and MCDA) in order to propose a structural procedure for the site selection problem for wind turbines. GIS serves as an elimination and data generation tool and feeds the evaluation stage. In order to be robust about the elimination, we create layers of data and keep it as rich as possible so that the remaining land areas are totally feasible to consider. Here, we contribute to the site selection problem with a wide literature research to identify the elimination constraints. After a tight elimination, we have several land alternatives to

evaluate, and their data are generated in the GIS. In the evaluation stage (which is a multiple criteria decision problem), we handle the alternative areas with a novel gridding approach, where each grid is technically and legally large enough to construct one turbine. We provide three different perspectives via application of three contemporary MCDA techniques for ranking and sorting purposes. Our evaluations are both at micro and macro levels, which means we have results for both for the grids and collection of grids (fields). At the micro level, we work with deterministic data for each grid. We rank and sort alternative grids using Elimination and Choice Translating Reality (ELECTRE) methods (ELECTRE III for ranking, ELECTRE-TRI for sorting). At the macro level, we work with collections of grids (fields) as alternatives. The idea of fields is important because it enables us to evaluate neighbouring land areas for multiple turbines. The motivation behind this is that no investor would search for an area to install only one turbine. Therefore, a selected site's performance is not only underlies its individual performance but also its neighbour grid's performance. The data for a field are the collection of deterministic data forming the given field, so the data becomes bounded. Therefore, in order to perform analysis at the field level, we also have to deal with uncertain data. The Stochastic Multiobjective Acceptability Analysis (SMAA) method is used for analysing the field level data. The results are presented at the field level, and they reveal a consistency between different methods applied and highlight a number of fields as highly ranked or of a top category. The paper is organised as follows: Section 2 provides brief information about Geographic Information Systems. In Section 3, we present the methodological considerations of three Multiple Criteria Decision Analysis methods; ELECTRE III, ELECTRE-TRI and SMAA-TRI, respectively. Section 4 is the methodology and data section, where our step-by-step approach, data generation and analysis framework are explained in detail. Section 5 discusses the results of multiple criteria analysis and Section 6 provides the policy implications and limitations of the proposed methodology. Section 7 concludes. The graphical material (maps generated in GIS) are provided in the Appendix. 2. Geographic Information Systems (GIS) A Geographic Information System (GIS) can be defined in basic terms as ‘a computer system that can hold and use data describing places on the Earth's surface’ (Rhind, 1989). It is a geographic data acquisition and a decision aiding tool with a widespread use in different areas such as urban management, transport planning, environmental management, telecommunication, service planning, national defence, network management or marketing (Heywood, 2010; Rolf, 2001). As a result of this wide scope of use, the definition of the concept can differ depending on who is using it for what purpose. In broader terms, we can define a GIS ‘as a system of hardware, software, and procedures to facilitate the acquisition, management, manipulation, analysis, modelling, representation, and output of spatially referenced data to solve complex planning and management problems’ (Carrion et al., 2008; Tegou et al., 2010). The primary purpose of a GIS is maintaining data about geographic space (Rolf, 2001). It allows geographic data to be organised, to be integrated with other data and to be analysed resulting in creation of new data and useful information in decisionmaking. One of the fundamental features of a GIS is data layers representing different characteristics in a given area. It enables creating specific criteria for each layer and overlaps all the layers so that the optimum area satisfying all criteria can be generated. Therefore, it is widely used in site selection problems for different purposes. In general, site selection via GIS follows the steps of identifying relevant factors, collecting data, identifying the siting criteria, overlaying and

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

extracting the final potential areas for the problem (Heywood, 2010). As mentioned above, GIS is a computerised system and one of the essential elements of it is the use of software. Both open source and commercial types of software packages can be found that serve as tools for GIS analyses.1

that there is no criterion for which b is better than a. It is made up from a weighted comparison of the performances over each criterion taken individually, cj(a, b). After a cumulative concordance matrix is constructed, in a similar manner, then using the veto thresholds instead of indifference thresholds, the disconcordance matrix is constructed through the following function:

3. Multiple Criteria Decision Analysis Multiple Criteria Decision Analysis (MCDA) is one of the wellknown branches of Decision Analysis in Operational Research. It deals with conflicting decision problems under the evaluation of several criteria. A variety of MCDA methods exists in the literature that can be used for various purposes such as choosing, ranking, sorting or describing. The decision maker usually decides which method to be used by taking the nature of the problem into consideration (Atici and Ulucan, 2011). The Elimination and Choice Translating Reality (ELECTRE) method is rooted in the study by Benayoun et al. (1966) and presented by Roy (1968). The method has evolved overtime to several derivations serving different purposes. One derivation of the method is ELECTRE III, and it is used for ranking purposes. ELECTRE-TRI is a member of the ELECTRE family aiming to sort alternatives (Figueira et al., 2005). 3.1. ELECTRE III In general, multiple criteria decision problems have an alternative set as A ¼ (a, b, … n) and a criteria set as G ¼ (g1, g2, … gm). The term gj(ai) refers to the criteria value of alternative ai with respect to criteria gj. In ELECTRE III, we have three types of thresholds associated with each criteria to be used in pairwise comparisons: preference threshold pj(gj(*)), indifference threshold qj(gj(*)) and veto threshold vj(gj(*)). These thresholds may be simple numerical constants, or they may be functions of the level of performance of one of the options being compared. To use the model these thresholds must be determined by the decision makers for all criteria as well as with the importance ratings (weights) for each of the criteria. Once threshold and weight values are decided, the concordance index c(a, b) is computed with the following function for each pair (a, b) of alternatives according to each criteria.

cj ða; bÞ ¼

8 > > > > > <

1

0 > > gj ðaÞ  gj ðbÞ þ pj > > > : p q j

if

gj ðaÞ  gj ðbÞ  qj

if

gj ðaÞ  gj ðbÞ  pj otherwise

j

This is based on a general comparison of the performances of alternative a and alternative b over all the criteria. Preference and indifference thresholds decided earlier are used here for comparisons. As a result, we obtain m number of n  n size matrices. After obtaining concordance matrices, we calculate a cumulative concordance matrix, which is a kind of combination of concordance matrices according to each criterion by taking the weights of the criteria into account.

Cða; bÞ ¼

n 1 X wj cj ða; bÞ W j¼1

where W ¼

n X

wj

j¼1

The cumulative concordance matrix is an n  n size matrix with the values between 0 and 1. A value of 0 indicating that alternative a is worse than alternative b for all criteria, and a value of 1 indicates

1 http://en.wikipedia.org/wiki/List_of_geographic_information_systems_ software.

3

dj ða; bÞ ¼

8 > > > > > <

0

if

gj ðaÞ  gj ðbÞ  pj

1

if

gj ðaÞ  gj ðbÞ  vj

> > gj ðbÞ  gj ðaÞ  pj > > > : v p j

otherwise

j

The essence of discordance is that any outranking of b by a indicated by the concordance index can be overruled if there is any criterion for which alternative b outperforms alternative a by at least the veto threshold. Even if option a is better than option b generally, there may be some criteria (possibly only one) for which alternative a is so much worse than alternative b that it moderates any overall preference for alternative a. If so, then the discordance index for that criterion reflects this. It can have a value from 0 to 1. A value 0 indicates that alternative b is not better than alternative a by more than the preference threshold, and a value 1 indicates that alternative b is better than alternative a by a margin greater than the veto threshold. By using cumulative concordance and disconcordance matrices, a final matrix, the credibility matrix, is obtained using the function below:

Sða; bÞ ¼

8 > > < > > : Cða; bÞ

Cða; bÞ Q j2Jða;bÞ

1  dj ða; bÞ 1  Cða; bÞ

if

dj ða; bÞ  Cða; bÞ; cj otherwise

J(a, b) set represents the set of criteria for which dj(a, b) > C(a, b). Credibility matrix is an n  n size matrix obtained through concordance and disconcordance matrices. After constructing the credibility matrix, as a final step, by using two kinds of distillation processes (upward and downward distillations), two rankings are obtained as ascending and descending rankings. The full ranking of the alternatives are obtained through intersecting the above two rankings (Rogers, 2000). 3.2. ELECTRE-TRI ELECTRE-TRI is a multiple criteria sorting methodology, which was developed to assign alternatives into pre-defined ordered categories (Mousseau et al., 2000a, 2000b; Dias and Climaco, 2000; Figueira et al., 2005). The assignment process of an alternative is done by comparison of alternatives with the profiles defining the boundaries of the categories (See Fig. 1). The ELECTRE-TRI methodology is divided into two phases: outranking and classification. In the outranking phase, a credibility index is developed to measure the degree of credibility of the assertion aSbh. Here S is an outranking relation operator and aSbh means “a is at least as good as bh”. Three types of thresholds associated with each criteria are used in alternative/profile pairwise comparisons: preference threshold pj(gj(*)), indifference threshold qj(gj(*)) and veto threshold vj(gj(*)). These thresholds are built to deal with the imprecise values of the alternatives. In order to validate aSbh by using the thresholds, two conditions should be satisfied; concordance and discordance. While the concordance condition computes the degree of support of aSbh by most of the criteria, discordance computes whether there exist strong objections to aSbh by the minority of the criteria. In the classification phase, the final assignment of each alternative into one category is

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

4

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

Fig. 1. ELECTRE-TRI categorisation.

done by using two methods; optimistic and pessimistic approaches. While the optimistic approach classifies an alternative into the highest possible category, the pessimistic approach classifies an alternative into the lowest possible category.

potentials of 13,827.36 and 13,012.56 Mega Watt (MW) per year, respectively. Currently, 46 wind power plant projects are licensed by the Turkish Energy Market Regulatory Authority in these two districts with the total potential of 2012.55 MW, which points out a great potential remaining idle.

3.3. SMAA-TRI 4.2. Layers of GIS Stochastic Multiobjective Acceptability Analysis (SMAA) is a recently developed family of Multiple Criteria Decision Analysis (MCDA) methods. Different SMAA methods can be used to handle the three main MCDA problem statements: choosing, ranking, and sorting (Tervonen and Figueira, 2008). The SMAA methods are developed for situations where criteria values and/or weights or other model parameters are not precisely known. Uncertain or imprecise criteria values are represented by stochastic variables corresponding to the deterministic evaluations with density function. Similarly, uncertain preferences are represented by a weight distribution with joint density function (Tervonen and Figueira, 2008). SMAA utilises certain descriptive measures calculated as multidimensional integrals over stochastic parameter spaces: the acceptability index, the central weight vector, and the confidence factor (Lahdelma et al., 1998). These measures are computed through Monte Carlo simulation. SMAA-TRI is the first SMAA method for the sorting problem. SMAA-TRI generates a category acceptability index for all pairs of alternatives and categories. The category acceptability index describes the stability of uncertain parameter values that assigns an alternative ai to category ck (Tervonen et al., 2009). 4. Data and methodology We aim to apply an integrated approach of two decision tools, the Geographic Information Systems (GIS) and the Multiple Criteria Decision Analysis (MCDA) to determine the most feasible sites for constructing wind power stations. Here, GIS serves as a data extraction tool and MCDA is used as an evaluation tool. This section explains our analysis framework step-by-step and is followed by the discussion of results in Section 5. 4.1. Application area We determine the application area as two neighbour districts, Balıkesir and Çanakkale, which are located in Western Turkey. The motivation behind this selection is the considerable wind energy potential of these districts, which is formally pointed out by the legal authorities. According to the Wind Energy Potential Atlas published in 2007 by General Directorate of Renewable Energy operating under the Turkish Ministry of Energy and Natural Resources, the districts with the largest potential of wind energy production in Turkey are Balıkesir and Çanakkale with the capacity

The Geographic Information System (GIS) is our tool to incorporate different types of data and generate our data set for Multiple Criteria Decision Analysis. As mentioned in Section 2, the GIS enables us to join different layers of data together. In accordance with the literature, legislation and considering the data availability, we identify 13 layers of data, which are related to the construction of wind power stations, as shown in Table 1. We extract these layers for our application districts Balıkesir and Çanakkale using ArcGIS 10.2 software. 4.3. Gridding the area The legislation of the General Directorate of Renewable Energy (dated as 22/05/2009) about the secure land size required for wind turbines states that one wind turbine requires a maximum turbine installation location of 700 m in radius and around this land, 300 m for a safety zone. This gives us a circle of land with a diameter of 2000 m. We use this information to divide the area into smaller pieces in order to have our alternative land areas to build wind power turbines. For ease of use in the GIS software, we handle the problem with a gridding approach and obtain square land pieces of 2000 m in length, which gives us pieces of land with a size of 4 km2 each, and each piece of land is large enough to build one wind turbine. Using ArcGIS software, 10,966 grids are obtained in total for the two districts, and a new layer is created in the database. 4.4. Elimination of infeasible areas Our methodology consists of two main stages: elimination and evaluation. We have more than 10,000 alternative pieces of land (grids), with different topographical or other sorts of properties in each layer. In order to determine the most feasible sites, the first step to undertake is the elimination of infeasible areas since it is obvious that not every piece of land is suitable for constructing a wind power station relying on their different properties. There are rules and regulations as well as commonly accepted criteria. Our elimination is based on a review about a relevant research on the wind power plant sites selection in the literature, as well as the technical reports of the General Directorate of Renewable Energy (GDRE) about the construction of wind power plants, as stated in the Wind Energy Potential Atlas (2007). Table 2 summarises our elimination criteria and constraints. Table 3 presents their relation with the constraints of several academic papers and sectorial reports.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

5

Table 1 Layers of GIS.

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13

Layer

Source of data

Elevation Slopes Energy transmission lines Capacity factors Lakes and rivers Protected areas Roads Railways Airports Urban areas Fault lines Mining sites Radio and TV stations

ASTER GDEMa ASTER GDEM General Directorate of Renewable Energy General Directorate of Renewable Energy GeoData Application by Turkish Ministry of Forestry and Water Affairsb GeoData Application by Turkish Ministry of Forestry and Water Affairs GeoData Application by Turkish Ministry of Forestry and Water Affairs GeoData Application by Turkish Ministry of Forestry and Water Affairs GeoData Application by Turkish Ministry of Forestry and Water Affairs Google Mapsc Turkish General Directorate of Mineral Research and Explorationd Turkish General Directorate of Mineral Research and Exploration Turkish Radio and Television Corporation (TRT)

a The ASTER Global Digital Elevation Model (ASTER GDEM) is “a joint product developed and made available to the public by the Ministry of Economy, Trade, and Industry (METI) of Japan and the United States National Aeronautics and Space Administration (NASA). See http://gdem.ersdac.jspacesystems.or.jp/. b http://geodata.ormansu.gov.tr/index.html?lang¼en. c https://www.google.co.uk/maps/@39.4590215,28.1648589,9z. d http://www.mta.gov.tr/v2.0/eng/index.php.

The elimination is implemented in ArcGIS software via incorporating elimination criteria and constraints we identify in Table 2. While defining our constraints, we follow mostly a conservative selection approach among the constraints available in our pool of sources. This type of elimination leaves us the areas that show a high level of suitability for wind power plant construction relying on the identified criteria. After a tight elimination process of infeasible areas relying on the constraints we defined, we have 398 grids remaining, which are shown in Figure A.1 in the Appendix. Now, our problem is to decide which grids are better for constructing wind power plants. It is of a multiple criteria decision problem in nature. 4.5. Defining the evaluation criteria and model parameters In defining the evaluation criteria and model parameters, we collaborated with a consultancy firm providing consultancy and subcontractor services to the electricity production companies (see Acknowledgements). After confirming with the experts that our elimination criteria and constraints were acceptable in sector terms, we worked on the identification of the evaluation criteria. Thirteen criteria are listed in Table 2 and used for elimination. Among them, we have to identify the ones that can matter in terms of construction cost, income and performance of wind turbines. Criteria such as distances to lakes, rivers, protected areas, urban areas, railways, airports, mining sites or TV stations are critical for elimination since wind turbine sites should not be very close to such places. However, after elimination, having one site closer to a lake or closer to an airport does not make that site more advantageous in constructing a wind turbine. On the other hand, factors Table 2 Elimination criteria and constraints. ID

Criteria

Constraint

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13

Elevation (m) Slope (%) Distance to transmission lines (m) Capacity factor Distance to lakes & rivers (m) Distance to protected areas (m) Distance to roads (m) Distance to railways (m) Distance to airports (m) Distance to urban areas (m) Distance to fault lines (m) Distance to mining sites (m) Distance to radio and TV stations (m)

<1500 <10 >250 >35 >3000 >2000 >500 >500 >5000 >2000 >200 >100 >600

such as Distance to the Energy Transmission Lines and Distance to the Roads are directly related to the income and cost of the turbines, respectively. Between two sites, the one closer to the transmission line will dominate the other one since the shorter the distance, the lower the losses in the network. Similarly, being close to roads will affect the cost of construction and maintenance since if the land is closer to the main road, these costs are lower. Distance to Fault Lines is another factor that can make a difference between two sites. Both districts are located in the first-degree earthquake areas of Turkey. Having one site closer to a fault line makes it riskier than the other in case of an earthquake. Obviously, the extent of importance of this criterion is lower than the factors such as distances to transmission lines or roads, but it is still a factor to be evaluated considering the degree of risk for the evaluation area. Slope of the land is another important factor that can affect the cost of construction. Although we already eliminated the very steep areas, the cost can still differ considerably between two sites with slopes of 1% and 10%. Finally, another criterion, while evaluating a piece of land in terms of wind turbine suitability, is the Capacity Factor, which represents a forecast of electricity production in a given area assuming a standard turbine is installed in the given area. According to the experts, it is the most important factor, and it is highly correlated with the elevation and wind speed in the given area. In order to provide different perspectives to different types of potential investors, we aim to implement three MCDA techniques explained in Section 3: ELECTRE III (for ranking), ELECTRE-TRI (for sorting) and SMAA-TRI (for sorting under uncertainty). Each modelling approach requires preliminary information from the decision makers such as criteria weights, difference, indifference and veto threshold values. We benefited from the expertise of the consulting firm in wind power plant location selection in both the definition of the evaluation criteria and the model parameters summarised in Table 4. In determining the criteria weights, we follow an Analytical Hierarch Process (AHP)-based approach (see Reference), where we ask the experts to fill in pairwise comparison matrices between the listed criteria and produce the criteria weights relying on these pairwise comparisons. Criteria type represents if a larger or lower values of a given criterion is preferable. Distance to the energy transmission lines and distance to the roads criteria are “Min” criteria since lower distances are better. 4.6. The data set In summary, our data set to apply MCDA approaches consist of

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

6

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

Table 3 Elimination constraints in the literature. Reference Baban and Parry (2001) PGWCa (2006) GDRE of Turkey (2007) Lejeune and Feltz (2008) Aydin et al. (2010) Haaren and Fthenakis (2011) Phuangpornpitak and Tia (2011) Sliz-Szkliniarz and Vogt (2011) Zhou et al. (2011) Al-Yahyai et al. (2012) Grassi et al. (2012) Ouammi et al. (2012) Gass et al. (2013)

C1

C2

C3 <10000 >250

<1500

<10% <25% <20%

C4

C5 >400 >500

>35 >150

<10% >200

<2000

<30% <10% <20% <10% <15%

>2500 >3000 >200 >250 >400 >240

C6

C7

>1000 >2000 >500 >2000 >300

<10000 >500 >100

>100 >500 >500 >300 >1500

C8

C9

C10

C11

C12

C13

>500 >250 >40

>500 >300 >100

>100 >500

>240 <1500 >150

>2500 >3000 >5000

>3000

>150 >2500 >150

>500 >500 >600 >2000 >1000 >500 >500 >500 >2000 >240 >1000 >1000

>100

Note: Highlighted data indicate the elimination constraints used for our analysis. See Table 2. a Provincial Government of the Western Cape. Table 4 Evaluation criteria and model parameters. Criteria

Type (Min/Max)

Criteria weight

Indifference threshold

Preference threshold

Capacity factor Slope Distance to fault lines Distance to transmission lines Distance to roads

Max Min Max Min Min

50.1% 10.3% 3.2% 25.6% 10.7%

1 0 1000 1000 5000

2 1 2000 2000 10000

398 alternative grids (4 km2 each) and 5 decision criteria. The necessary data is extracted from the corresponding layers in ArcGIS 10.2. The descriptive statistics for the data are given in Table 5. 4.7. Ranking with ELECTRE III Using the criteria weights and model parameters we defined in Table 4, we conducted an ELECTRE III analysis to rank the decision alternatives. The analysis provided us a micro level point of view as each grid has a size available for construction of only one turbine. In addition, the number of alternatives are large in number considering that we have only five criteria, which can decrease the discrimination ability of the model to the ranks. Due to the closeness of the data to each other (especially in capacity factor and slopes), small deviations can cause big differences in ranks and this may lead neighbour grids not to follow an order. Such results can be confusing from the investor point of view since investing on only one grid (i.e., one turbine) may not be very feasible. Therefore, the ranking of alternatives is not enough information alone. We need additional information for a better assessment. We extend the analysis into two directions, one of which is the sorting of alternatives into categories, and the second direction is the investigation of the best clusters of grids that are neighbours to each other. 4.8. Sorting with ELECTRE-TRI To provide a broader point of view on the decision alternatives, we conducted an ELECTRE-TRI analysis to assign the alternatives into 3 categories where category 1 represented the worst and category 3 represented the best alternatives. The same criteria weights and the thresholds as in ELECTRE III were used in order to sustain the consistency between analyses. In ELECTRE-TRI analysis, profile values that reveal the cutting points (established as percentages) between the categories play an important role. We consider three equal sized categories, where the profile values are taken as 33.3%. Category 1 represents the alternatives in the worst 33.3%, whereas category 3 consists of the best 33.3% considering all five criteria, their weights and threshold values.

4.9. Sorting with SMAA-TRI In both ELECTRE III and ELECTRE-TRI analyses, we looked at the problem from the level of the grids and determined to what rank or what category each grid belongs. Since, for an investor, it would not be feasible to construct only one turbine, what can also matter is to know how the neighbour grids behave. In order to answer this question, we clustered the individual grids into fields with the neighbour grids. This provided pieces of land area consisting of several grids that are next to each other at least along one edge. We treated the fields of grids as decision alternatives and tried to decide which fields were better. Two main issues arose at this point. First, there were single grids or fields with only 2, 3 or 4 grids, which were small in size and might not be feasible as well (see Figure A.1 in Appendix). Because the focus was to investigate the larger areas that could be more reasonable for investment, treating them as individual fields would distort the analysis. On the other hand, excluding those fields changed the sample, and there arose the issue of comparability with the prior analyses. Therefore, instead of excluding the single or relatively small grids, we presumed them to be as a single field. In the case of obtaining that field of singles as a favourable alternative, we had the ELECTRE III and ELECTRE-TRI results for each single grid in that field so that we could evaluate them separately. As a result, we had 13 fields, one of which was the ‘field of singles’. The fields are shown in Figure A.2 in Appendix and the number of grids in each field are given in Table 6. The second issue to deal with was regarding the data set. The data required revisiting since we had a grid-based data. Now that we had fields consisting of several grids as alternatives, the data were in the form of intervals with a mean and standard deviation for each criteria. Therefore, we needed a decision approach that enabled the handling of uncertain data. SMAA-TRI as described in Section 3 could serve for this purpose and sort the alternatives with uncertain criteria values. Using ArcGIS 10.2 once again, we extracted the data set for each field in the form of mean and standard deviation and applied the SMAA-TRI method using the open source JSMAA software.2 All the model parameters (criteria weights and

2

http://smaa.fi/jsmaa/.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

7

Table 5 Descriptive statistics for the data set.

Mean Stdev Min Max

Capacity factor

Slope (%)

Distance to fault lines (m)

Distance to transmission lines (m)

Distance to roads (m)

50.7 5.2 35.0 56.0

2.1 3.2 0.0 10.0

23080.9 17740.8 201.8 73265.7

13248.4 10905.7 279.0 64382.5

6580.2 4107.8 512.5 20742.6

Table 6 Number of grids in each field. # of grids Field Field Field Field Field Field Field

1 2 3 4 5 6 7

43 42 39 36 16 16 15

# of grids Field Field Field Field Field Field

8 9 10 11 12 13

14 13 10 9 9 136

thresholds) were taken as the values shown in Table 3. As in ELECTRE-TRI, we considered three equal sized categories, where the profile values were taken as 33.3%. 5. Results In this section, we present the results of our evaluations of suitable areas through MCDA methods. From ELECTRE III, ELECTRETRI and SMAA-TRI analyses, designed as described in Section 4, we have answers to the following questions, respectively:  What is the rank of each grid?  What category does each grid belong to?  How do neighbour grids behave? In order to better present all the information above, we benefit from the idea of clustering the neighbour grids into fields rather than presenting the rank or category of individual grids, which may not be very meaningful. ELECTRE III provides us with 298 ranks. Some of the 398 grids attain the same ranks. Table 7 shows how ranks are distributed between fields. It is observed that the majority of the ranks between 1 and 50 are attained by the grids located in fields 5, 8 and 13. For instance, 21% of the ranks between 1 and 50 are attained by the grids in Field 5. The majority of the ranks between 250 and 298, which represents the bottom, are occupied by the grids in fields 6,10, and 13. Note that field 13 is a special case. It is a collection of grids with a very small number of neighbours. They are located in very different areas as seen in Figure A.2 in the Appendix. The high percentages given for this field in all ranks top to bottom is due to the high number of grids in this field. Although some of these grids are highly ranked, it is not very acceptable to recommend these areas for wind turbines because turbines that can be installed in those areas will be limited in number. Also, the high percentage in the bottom ranks (33%) point out that there exists many grids that not very highly ranked in this field. Considering that those are very small land areas, we can say that the many of the grids that do not belong to large groups are not very highly ranked anyway. Leaving field 13 out of discussion, in a search for the most suitable areas, the ELECTRE III analysis emphasises the fields 5 and 8. Let us examine what is the situation for ELECTRE-TRI with keeping the ELECTRE III results in mind. Table 8 summarises the results in terms of fields putting ELECTRE III and ELECTRE-TRI findings together. Recall that in ELECTRE-TRI, we did the sorting for three equal sized groups where profile values of 33.3% and category 3 (C3)

represented the top 33.3% alternatives, whereas Category 1 (C1) represented the bottom. For a viable comparison, we undertake the ELECTRE III ranks in three equal sized groups as well: ranks from 1 to 99, 100 to 198 and 199 to 298. The Left panel of Table 8 presents the percentage of grids for each field in each of these categories. For instance, 55% of the grids in field 2 are ranked among the top 99, 31% in between 100 and 199 and 14% among the bottom 99. As discussed in Section 3, ELECTRE-TRI method proposes two approaches: optimistic and pessimistic. Results for both are provided on the right panel of Table 8. Since the optimistic model approaches the problem in a way that provides comparative advantage to the alternatives (Siskos et al., 2007), the results indicate very favourable results for the majority of alternatives where there exists vast category 3 assignments. All the grids in 9 out of 13 fields are assigned to a top category. This makes the results for the optimistic approach results difficult to interpret. On the other hand, the pessimistic ELECTRE-TRI approach, which is usually used when it is required to apply a conservative policy (Siskos et al., 2007), highlights fields 3, 5, 8, and 12. These fields have the majority of their grids assigned in the top category. When we examine the position of the fields pointed out by ELECTRE III (5, 8, and 12), we observe that they also perform well in ELECTRE-TRI, since all of the alternatives in those fields are assigned to the top category. Another noticeable field is field 3, which has 95% of its grids in the top category. It also performs relatively well in ELECTRE-TRI. Finally, we have results for SMAA-TRI analysis conducted with uncertain data at the field level. We aim to sort the fields (rather than grids themselves) into three equal sized categories. The analysis provides us the probabilities for the fields belonging to a given category. Since SMAA-TRI is based on ELECTRE-TRI, it also propose the optimistic and pessimistic points of view. However, note that there are fundamental differences in our approach to the problem in handling those methods. First of all, in SMAA-TRI, we deal with uncertain data. Secondly, in SMAA-TRI analysis, we have the alternatives as fields whereas in ELECTRE-TRI, the alternatives are grids. Therefore, we intend to present SMAA-TRI results as a different perspective for our site selection problem rather than comparing the results with the ELECTRE methods. However, it is interesting that there is a high level of consistency. SMAA-TRI results are given in Table 9. Once again, the optimistic approach does not provide much help for our problem. Seven out of 13 fields are 100% in category 3 (see left panel of Table 9) and for the others the probabilities of Category 3 assignments are higher than other probabilities with one exception. Field 10 is assigned to Category 2 with 61%. The pessimistic point of view provides more conservative results where only 3 fields (5, 8 and 12) have a 100% probability of being in category 3. They are followed by field 3, which is 98.4% in category 3. To sum up, the results exhibit consistency between the methods. Fields 3, 5, 8, and 12 are emphasised by all MCDA methods used. The location of the fields on the map can be seen in Figure A.2 in the Appendix. 6. Policy implications In this section, we discuss how the proposed methodology can

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

8

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

Table 7 ELECTRE III rankings distribution. Ranks

Fields

1 to 50 51 to 100 101 to 150 151 to 200 201 to 298 250 to 298

1

2

3

4

5

6

7

8

9

10

11

12

13

0% 0% 10% 19% 23% 10%

13% 23% 10% 10% 6% 3%

13% 21% 16% 6% 3% 1%

0% 6% 13% 13% 13% 9%

21% 5% 0% 0% 0% 0%

0% 0% 0% 0% 0% 24%

0% 0% 10% 13% 0% 0%

18% 5% 0% 0% 0% 0%

0% 0% 2% 6% 12% 0%

0% 0% 0% 0% 0% 15%

3% 3% 2% 3% 3% 0%

6% 8% 0% 0% 0% 0%

26% 30% 39% 32% 41% 37%

Table 8 ELECTRE III Ranks and ELECTRE-TRI Categories for the Grids in each Field. ELECTRE III ranks 1 to 99

Field Field Field Field Field Field Field Field Field Field Field Field Field

1 2 3 4 5 6 7 8 9 10 11 12 13

ELECTRE-TRI categories 100 to 198

0% 55% 56% 11% 100% 0% 0% 100% 0% 0% 44% 100% 26%

47% 31% 36% 44% 0% 0% 100% 0% 38% 0% 33% 0% 35%

199 to 298

Optimistic

53% 14% 8% 44% 0% 100% 0% 0% 62% 100% 22% 0% 39%

Pessimistic

C3

C2

C1

C3

C2

C1

95% 100% 100% 100% 100% 88% 100% 100% 100% 50% 100% 100% 96%

5% 0% 0% 0% 0% 13% 0% 0% 0% 50% 0% 0% 3%

0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1%

74% 90% 95% 75% 100% 0% 0% 100% 31% 0% 56% 100% 59%

23% 5% 5% 17% 0% 81% 73% 0% 69% 100% 22% 0% 23%

2% 5% 0% 8% 0% 19% 27% 0% 0% 0% 22% 0% 18%

Note: Highlighted data corresponds to the most feasible sites.

foster the wind energy policymaking process. After looking at the stakeholder and financial feasibility dimensions, we emphasize why the idea of fields is important. Finally, potential limitations of the current methodology are explained. 6.1. Stakeholders While handling real world problems with multiple stakeholders, it is critical to consider the interests of different parties. The wind power plant site selection problem is such a multi-dimensional problem. We can put the stakeholders into three main groups. One is the regulatory authority concerning the efficient use of the potential as well as the eligibility regarding the rules, with regulations and technical restrictions identified. The second group can be placed as investors mostly concerned about the financial feasibility and aiming to attain a high level of electricity production capacity at a Table 9 SMAA-TRI results. SMAA-TRI results Optimistic

Field Field Field Field Field Field Field Field Field Field Field Field Field

1 2 3 4 5 6 7 8 9 10 11 12 13

Pessimistic

C3

C2

C1

C3

C2

C1

96.4% 100.0% 100.0% 99.5% 100.0% 57.6% 62.2% 100.0% 100.0% 39.1% 100.0% 100.0% 78.1%

3.6% 0.0% 0.0% 0.5% 0.0% 42.4% 36.6% 0.0% 0.0% 60.9% 0.0% 0.0% 21.5%

0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.2% 0.0% 0.0% 0.0% 0.0% 0.0% 0.4%

82.6% 86.6% 98.4% 66.1% 100.0% 9.9% 4.7% 100.0% 47.2% 31.8% 55.0% 100.0% 50.3%

17.3% 13.0% 1.6% 30.8% 0.0% 72.3% 64.0% 0.0% 52.8% 66.8% 36.0% 0.0% 43.2%

0.1% 0.4% 0.0% 3.1% 0.0% 17.8% 31.3% 0.0% 0.0% 1.5% 9.0% 0.0% 6.5%

minimum installation cost. Finally, the society populated around the investment area is a stakeholder since not only does it embody the last users of the services but also their physical or ecological environment can be affected. It is important to note that any methodology aiding the wind power site problem should pay attention to those interests, and they can contradict each other. As pointed out in the introduction, one land area can have a high wind power potential in technical terms, which may seem very promising for investors; however it may be too close to an urban area and, therefore, can be conflicting with the regulatory authority or society's interests. On the other hand, there can be many alternative land areas, which seem to be eligible for the authority and have no opposition from the community. Then, it is important to decide the most feasible areas with the highest wind potential and the least installment cost. The above mentioned characteristics of the problem make it suitable for application of Multiple Criteria Decision Analysis (MCDA) approaches that enable incorporation of different perspectives into the modelling process. Together with the use of Geographic Information Systems (GIS), the proposed methodology allows merging multiple stakeholder preferences in order to come up with a satisfying decision frame for all parties. In this way, the decision-making process evolves to a group decision-making process where the results can reflect the preferences of different interest groups. We can say that in our application conducted over a promising region of Turkey for wind power generation, the elimination stage mostly reflects the interests of the regulatory authority and society. The evaluation stage, on the other hand, mostly serves the interests of investors via ranking or sorting of identified alternatives. Of course, the importance of the elimination stage for the investors cannot be underestimated since it provides information to reveal the areas that are not suitable for investing. Although, the application is conducted in a specific region, the methodology is adaptable to any area and flexible to handle changing preferences.

Note: Highlighted data corresponds to the most feasible sites.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

9

6.2. Financial feasibility

7. Conclusion

Both regulatory authorities and investors can benefit from the different aspects of the proposed methodology. From the regulatory authority point of view, the information obtained is valuable for policy-making objectives and licensing of the wind power plants. As well as deciding to permit an investor to construct turbines in a given area, the regulatory authority also serves as a competent party for directing the investor to feasible investments. Use of a structured decision-making process can produce more rational decisions. From the investor point of view, financial feasibility is one of the important factors to be considered in every investment. Although the analysis seems to be oriented to technical criteria, the financial feasibility is the main concern, especially, in the evaluation stage. Table 10 summarizes how our evaluation criteria can be associated with different monetary variables, mainly as revenues, initial investment costs and operational costs. Of the criteria used, topographical properties such as slope and distance to roads are directly related with the construction and maintenance costs assuming identical turbines are going to be installed. Another evaluation criteria, distance to transmission lines, is closely related with operational costs, specifically with the cost of distribution of the power since the more distance there is, the more likely it is to have energy loss during the transmission. The capacity factor is important for power generation, therefore, for the revenues. Thus, taking the given technical criteria into account will have a direct effect in the financials of the investments.

Considering the growing demand for energy in Turkey and the conspicuous attention on renewables worldwide, careful selection of energy investment projects and efficient use of resources is gaining importance. This paper aims to provide an insight for the site selection problem of renewable energy investments in specific wind power plant site locations. The wind power site selection problem, in general, involves different stakeholders such as the regulation authority, investors and society, where each party can pursue different types of priorities and preferences. Together with this, the different topographical or other sort of properties, rules and regulations as well as commonly accepted criteria makes the site selection procedure multidimensional. In this paper, we deal with this multi-dimensionality and propose a decision-aiding approach on the site selection problem for wind turbines through an integration of Geographic Information Systems and Multiple Criteria Decision Analysis together. Several policy implications can be mentioned for different stakeholders. We advocate that a structured decision-making process can yield more rational decisions on investments. We illustrate our proposed methodology through an application in Turkey. Although it focuses on a specific region in Turkey, the proposed methodology is adaptable to any area. Our application area is the Balıkesir and Çanakkale areas in Western Turkey, which possess the largest potential for wind power generation from all over the country. The proposed methodology consists of preelimination of infeasible sites, ranking and sorting of available sites. It involves both handling of deterministic and uncertain data. In such a manner, it provides a boutique approach to the initial problem and provides different types of results. Besides, it suggests a structural procedure and is applicable to other types of site selection problems with the related criteria depending on the problem. The selection problem is induced to 13 large-scale land areas (fields) consisting of several 4 km2 sized neighbouring grids obtained after an elimination of infeasible areas in the Geographic Information System and the results are presented in terms of fields. Three MCDA methods (ELECTRE III, ELECTRE-TRI and SMAA- TRI) are applied for different purposes and with different considerations and the result reveals high levels of consistency between methods. The consistency of the results reveal that both the regulators and the investors can benefit from a structured multiple criteria decision analysis approach integrated with Geographic Information Systems. The proposed approach can address different stakeholders and has a flexible design that can reflect different preferences of the dealing parties and is applicable to any potential area for wind power sites.

6.3. Idea of fields As noted, the proposed methodology can produce results both at macro and micro levels such that we can deal with individual grids of the same size or collection of neighbour grids (fields). The idea of fields is important to benefit from economies of scale, since it enables us to evaluate neighbouring land areas for multiple turbines. The motivation behind this is that neither the legal authorities nor the investors would be interested in installation of only one turbine. We approach the problem from the ranking of the single grids, which are land areas large enough to build one turbine and progress to the selection of multiple neighbour grids in order to come up with larger scale alternative land areas. 6.4. Limitations The proposed methodology and results are not without their limitations. In some cases (as in ours) some information about the land areas evaluated is not available. The first that comes to mind is the ownership status of the land areas that are found to be highly feasible. Whether it is the regulatory authority or the investor, a further examination is naturally required for the exact location. Actually, in any case, a further investigation in the area is necessary. However, this does not devalue the obtained information since we have ranks or groups of alternative grids or fields, which always leave room for an alternative plan. It is very important to narrow down the alternatives and decide the most promising area. Nevertheless, the method is flexible for integrating additional criteria and preferences if necessary data is available.

Acknowledgements anay Yola from the “enerGY The authors are grateful to Dog Consultancy & Project” company in Ankara, Turkey and Mustafa Caliskan from the Turkish General Directorate of Renewable Energy for their support in the data collection process.

Table 10 Relationship of the evaluation criteria with monetary variables. Monetary Variable Revenue Initial investment costs Operational costs

Related criteria Generated power Turbine costs Construction costs Maintenance costs Distribution costs

Capacity factor Assumed identical Slope, distance to roads Slope, distance to roads Distance to transmission lines

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

10

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

Appendix

Fig. A.1. Alternative areas for wind power sites in Balıkesir and Çanakkale.

Fig. A.2. Alternative fields for wind sites.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001

K.B. Atici et al. / Utilities Policy xxx (2015) 1e11

References Al-Yahyai, S., Charabi, Y., Gastli, A., Al-Badi, A., 2012. Wind farm land suitability indexing using multi-criteria analysis. Renew. Energy 44, 80e87. Atici, K.B., Ulucan, A., 2011. A multiple criteria energy decision support system. Technol. Econ. Dev. Econ. 17 (2), 219e245. Aydin, N.Y., Kentel, E., Duzgun, S., 2010. GIS-based environmental assessment of wind energy systems for spatial planning: a case study from Western Turkey. Renew. Sustain. Energy Rev. 14 (1), 364e373. Baban, S.M., Parry, T., 2001. Developing and applying a GIS-assisted approach to locating wind farms in the UK. Renew. Energy 24 (1), 59e71. Benayoun, R., Roy, B., Sussman, B., 1966. ELECTRE: Une methode pour guider le choix en presence de points de vue multiples. Note de travail 49. SEMA-METRA International, Direction Scientifique. Carrion, J.A., Estella, A.E., Dols, F.A., Toro, M.Z., Rodriguez, M., Ridao, A.R., 2008. Environmental decision-support systems for evaluating the carrying capacity of land areas: optimal site selection for grid-connected photovoltaic power plants. Renew. Sustain. Energy Rev. 12, 2358e2380. Dias, L., Clímaco, J., 2000. ELECTRE-TRI for groups with imprecise information on parameter values. Group Decis. Negot. 9 (5), 355e377. Figueira, J., Mousseau, V., Roy, B., 2005. Electre methods. In: Figueria, J., Greco, S., Ehrgott, M. (Eds.), Multiple Criteria Decision Analysis: State of the Art Surveys. Springer Verlag, Boston, Dordrecht, London, pp. 133e162. Gass, V., Schmidt, J., Strauss, F., Schmid, E., 2013. Assessing the economic wind power potential in Austria. Energy Policy 53, 323e330. General Directorate of Renewable Energy, 2007. Turkish Wind Energy Potential Atlas. Ankara, Turkey. Grassi, S., Chokani, N., Abhari, R.S., 2012. Large scale technical and economical assessment of wind energy potential with a GIS tool: case study Iowa. Energy Policy 45, 73e85. Haaren, R. v., Fthenakis, V., 2011. GIS-based wind farm site selection using spatial multi-criteria analysis (SMCA): evaluating the case for New York State. Renew. Sustain. Energy Rev. 15 (7), 3332e3340. Heywood, I., 2010. An Introduction to Geographical Information Systems. Pearson Education India. Lahdelma, R., Hokkanen, J., Salminen, P., 1998. SMAA - stochastic multiobjective acceptability analysis. Eur. J. Oper. Res. 106 (1), 137e143. Lejeune, P., Feltz, C., 2008. Development of a decision support system for setting up a wind energy policy across the Walloon region (southern Belgium). Renew. Energy 33 (11), 2416e2422.

11

Mousseau, V., Figueira, J., Naux, J.Ph, 2000a. Using assignment examples to infer weights for ELECTRE-TRI method: some experimental results. Eur. J. Oper. Res. 130 (2), 263e275. Mousseau, V., Slowinski, R., Zielniewicz, P., 2000b. A user-oriented implementation of the ELECTRE-TRI method integrating preference elicitation support. Comput. Oper. Res. 27 (7e8), 757e777. Ouammi, A., Ghigliotti, V., Robba, M., Mimet, A., Sacile, R., 2012. A decision support system for the optimal exploitation of wind energy on regional scale. Renew. Energy 37 (1), 299e309. Phuangpornpitak, N., Tia, S., 2011. Feasibility study of wind farms under the Thai very small scale renewable energy power producer (VSPP) program. Energy Proc. 9, 159e170. Provincial Government of the Western Cape, 2006. Strategic Initiative to Introduce Commercial Land Based Wind Energy Development to the Western Cape. CNdV africa planning & design, Cape Town. Rhind, D.W., 1989. Why GIS? ARC News (summer) 28e29. Rogers, M., 2000. Using Electre III to aid the choice of housing construction process within structural engineering. Constr. Manag. Econ. 18, 333e342. Rolf, A., 2001. Principles of Geographic Information Systems: an Introductory Textbook. Enschede: The International Institute for Aerospace Survey and Earth Sciences. sence de points de vue multiples (la Roy, B., 1968. Classement et choix en pre thode ELECTRE). RIRO 8, 57e75. me Siskos, Y., Grigoroudis, E., Krassadaki, E., Matsatsinis, N., 2007. A multicriteria accreditation system for information technology skills and qualifications. Eur. J. Oper. Res. 182 (2), 867e885. Sliz-Szkliniarz, B., Vogt, J., 2011. GIS-based approach for the evaluation of wind energy potential: a case study for the KujawskoePomorskie Voivodeship. Renew. Sustain. Energy Rev. 15 (3), 1696e1707. Tegou, L.-I., Polatidis, H., Haralambopoulos, D.A., 2010. Environmental management framework for wind farm siting: methodology and case study. J. Environ. Manag. 91 (11), 2134e2147. Tervonen, T., Figueira, J.R., Lahdelma, R., Dias, J.A., Salminen, P., 2009. A stochastic method for robustness analysis in sorting problems. Eur. J. Oper. Res. 192 (1), 236e242. Tervonen, T., Figueira, J.R., 2008. A survey on stochastic multicriteria acceptability analysis methods. J. Multi-Criteria Decis. Anal. 15 (1e2), 1e14. Zhou, Y., Wub, W.X., Liua, G.X., 2011. Assessment of Onshore wind energy resource and wind-generated electricity potential in Jiangsu. China Energy Proc. 5, 418e422.

Please cite this article in press as: Atici, K.B., et al., A GIS-based Multiple Criteria Decision Analysis approach for wind power plant site selection, Utilities Policy (2015), http://dx.doi.org/10.1016/j.jup.2015.06.001