Introduction

The population growth and rapid urbanization that have resulted from technological development and industrialization are increasing human activities all over the world (Ripa et al. 2017). In this process, the waste generated with the increasing consumption is threatening the environment and human health due to the amount and harmful content. According to the World Bank, as a result of the population growth, generation of municipal waste is expected to reach 2.2 billion tons worldwide by 2025. Managing waste is essential to create sustainable and liveable cities in the world (Özkan et al. 2019).

The amount of municipal solid waste (MSW), which typically includes household, garden/park, and commercial/industrial waste, has been increasing rapidly in urban centres of developing countries during the last few decades (Ripa et al. 2017; Rupani et al. 2019); however, municipal solid waste management (MSWM) remains inadequate (Henry et al. 2006; Zhang et al. 2010; Rupani et al. 2019). Landfill remains the main approach in many countries, for example, Chinese MSWM is 60.16% (Mian et al. 2017). In Iran, which is one of the countries where an effective waste management approach is lacking, the main method for the final waste disposal is the use of landfill, which is mostly carried out not in accordance with the environmental regulations (Rupani et al. 2019). Indian MSWM focuses on collection services rather than scientific treatment and disposal (Priti and Mandal 2019). MSWM is a significant item in the municipal budget, and approximately 65% to 80% of cost is spent on the collection and transportation of solid waste (Rızvanoğlu et al. 2020). The first condition for economic recovery from solid waste is separate collection, with the primary approach using individual collection boxes (Terazono et al. 2015; Pham Phu et al. 2019). Collection points can be created with collection boxes that ensure the separation of waste at source, and waste can be controlled within the framework of sustainable development principles. In Turkey, where the average amount of municipal waste per capita is 1.16 (kg/day), and the total amount of MSW collected by municipalities was 32.2 million tons in 2018 (TSI 2019), collection boxes for waste oil, especially domestic vegetable waste oil, and collection boxes of waste batteries are widely used within the MSWM system. In this context, in Turkey, the ‘Regulation on the Control of Waste Vegetable Oil’ was published in April 2005, and the ‘Regulation on the Control of Waste Batteries and Accumulators’ was published in August 2004 (TCA 2007).

For the first time in Turkey, a collection box was designed by the İlkadım District Municipality in Samsun in 2018 as part of a zero waste project, to collect domestic waste vegetable oil and waste batteries simultaneously. This waste vegetable oil and waste battery collection box was called an ‘AYPIKUT’, consisted of two parts (Fig. 1). Waste vegetable oil was collected in two ways, either poured directly into a vessel within the collection box (made to look like a vegetable oil bottle to attract attention) or left in a container, such as a glass or plastic bottles, that was deposited in the collection box. Waste batteries were collected in a part of the collection box that was designed to look like a battery, also in order to attract attention. The study was aimed to determine suitable locations for AYPIKUT in Atakum. Atakum is a district of Samsun located on the Black Sea coast of Turkey, with a population of 215,633 in 2019. Atakum has the highest population growth rate in Samsun city, with an annual growth of 6.4% for the last 5 years; 10,607 dwellings were sold in Atakum district in 2019. It was listed as the 23rd district in the most residential sales ranking among 923 districts of Turkey in 2019 (TSI 2020; GDLRC 2020). There has been a commensurate increase in municipal waste. This paper develops an analytical tool for use by local government that combines geographic information systems (GIS), multi-criteria hesitant fuzzy linguistic term set (multi-criteria HFLTS), and the full multiplicative form of multi-objective optimization by ratio analysis (MULTIMOORA), to determine suitable locations for AYPIKUT in Atakum. GIS can be considered an effective method to analyse spatial data in many areas of application, including various aspects of waste management (Rızvanoğlu et al. 2020). Multi-criteria decision-making (MCDM) approaches provide robustness and flexibility to address comparison problems involving multiple and varied units of measurement (Çalış Boyacı 2020). A combined GIS-MCDM method affords a practical approach that can manage time and costs whilst reducing errors and increasing the efficiency of the decision-making process (Eghtesadifard et al. 2020). Moreover combining MCDM methods and fuzzy sets provides a more accurate and systematic assessment. Decision-makers may have problems in identifying linguistic terms and may need flexibility in their evaluations. This challenge is overcome by the multi-criteria HFLTS method, ensuring that linguistic evaluations obtained from experts are effectively preserved without any loss of knowledge.

Fig. 1
figure 1

The frontal and back views of the waste collection box (AYPIKUT)

Full size image

In the literature, most studies using GIS for MSW consider the landfill site selection (Sumathi et al. 2008; Guiqin et al. 2009; Şener et al. 2010; Nas et al. 2010; Gorsevski et al. 2012; Eskandari et al. 2015; Bahrani et al. 2016; Jamshidi-Zanjani and Rezaei 2017; Khodaparast et al. 2018; Kapilan and Elangovan 2018; Kamdar et al. 2019; Barzehkar et al. 2019; Aksoy and San 2019; Karimi et al. 2019; Eghtesadifard et al. 2020; Rezaeisabzevar et al. 2020; Rahimi et al. 2020). Some studies consider the optimization of the collection and transportation of MSW. For example, Kanchanabhan et al. (2010) developed a design for MSW collection using GIS including a vehicle tracking system, Kallel et al. (2016) developed optimized scenarios using GIS in order to improve the efficiency of waste collection and transportation, and Nguyen-Trong et al. (2017) proposed a model for optimizing MSW collection. Lella et al. (2017) presented methods for optimal collection and transportation of MSW using GIS techniques through network analysis. Amal et al. (2018) proposed spatial GIS-based genetic algorithm to optimize the route of solid waste collection. Rızvanoğlu et al. (2020) used linear programming and GIS to develop an optimal routing schedule for MSW collection and transportation. However, there are few studies on the selection of the sites for MSW collection boxes. Vijay et al. (2008) presented a GIS-based analysis of the location of collection bins in MSWM systems using p-median constrained model. Chalkias and Lasaridi (2009) developed a methodology for the reallocation of waste collection bins, making use of the GIS spatial analysis functions. Khan and Samadder (2016) presented a suitable solid waste collection bin allocation method based on GIS technology.

The HFLTS method has been applied in several areas, such as supplier selection (Liu and Rodríguez 2014; Fahmi et al. 2016), performance evaluation (Tüysüz and Şimşek 2017), service provider selection (Wei et al. 2015; Ghadikolaei et al. 2018), web tool development (Montes et al. 2015), technology selection (Wei and Liao 2016; Khishtandar et al. 2017), healthcare risk analysis (Liu et al. 2016), facility location selection (Feng et al. 2018), eco-friendly city selection (Çalış Boyacı 2020), and website evaluation (Özkan et al. 2020). The MULTIMOORA method also has wide applicability, such as personnel selection (Baležentis et al. 2012; Baležentis and Zeng 2013), industrial robot selection (Datta et al. 2013), evaluation of excavator technologies (Altuntaş et al. 2015), materials selection (Hafezalkotob and Hafezalkotob 2015), evaluation of bike-share stations (Kabak et al. 2018), and evaluation of smart bike-sharing programs (Tian et al. 2018), technological forecasting method selection (Dahooie et al. 2019), and landfill site selection (Rahimi et al. 2020). A summary of GIS-based MCDM studies is presented in Table 1. This shows that no GIS-based MCDM studies combine GIS, the multi-criteria HFLTS, and MULTIMOORA methods. This study fills this gap.

Table 1 Review of GIS-based MCDM studies in the literature
Full size table

Methodology

This study proposes a methodology that combines GIS, multi-criteria HFLTS, and MULTIMOORA for evaluating the locations suitable for AYPIKUT. These three methods are explained, and the flow chart of the proposed method is given in this section.

Geographic information systems

Spatial data is one of the most important data types to provide understanding of the world. The GIS allows query, analysis, and display of spatial data. GIS has been defined in several ways. An early definition of a GIS was a tool using collecting, querying, analysing, transferring, storing, and displaying the data of the earth for a specific purpose (Burrough and McDonnell 1998). One of the more common uses of GIS is as an aid in decision problems involving multi-attribute or multi-objective land use allocation (Cromley and Hanink 2003).

Multi-criteria HFLTS method

Fuzzy logic and fuzzy set theory successfully deal with uncertain knowledge. However, when two or more sources of uncertainty occur simultaneously, the fuzzy sets remain limited (Rodríguez et al. 2012). Hesitant fuzzy sets, which can solve the difficulties in determining the membership degree of an element, are a generalization of fuzzy sets that permits us to represent the situation in which different membership functions are considered possible (Torra 2010). In the classical fuzzy linguistic approaches, a single expression which limits the experts should be selected. However, experts may hesitate to select the appropriate linguistic expression (Onar et al. 2016). Experts can use HFLTS when they hesitate between several linguistic expressions (Rodríguez et al. 2013).

A hesitant linguistic group decision-making model with a single criterion was suggested by Rodríguez et al. (2013). On the basis of this model, experts assess the alternatives based on a single criterion. Yavuz et al. (2015) extended this algorithm to take into account a multi-criteria decision-making problem, and the steps of the suggested algorithm are as follows:

  1. Step 1.

    Define the semantics and syntax of the linguistic term set S. A set of seven terms, S, could be given by Eq. (1) (Rodríguez et al. 2013):

$$ S=\left\{\begin{array}{l}\mathrm{no}\ \mathrm{importance}(n),\mathrm{very}\ \mathrm{low}\ \mathrm{importance}(vl),\\ {}\mathrm{low}\ \mathrm{importance}(l),\mathrm{medium}\ \mathrm{importance}(m),\\ {}\mathrm{high}\ \mathrm{importance}(h),\\ {}\mathrm{very}\ \mathrm{high}\ \mathrm{importance}(vh),\\ {}\mathrm{absolute}\ \mathrm{importance}(a)\end{array}\right\} $$
(1)
  1. Step 2.

    Define the context-free grammar GH, where GH = {VN, VT, I, P}. VN is the set of nonterminal symbols, VT is the set of terminals’ symbols, I is the starting symbol, and P is the production rules that are defined in an extended Backus–Naur form (Rodríguez et al. 2012).

  2. Step 3.

    Gather the preference relations pk given by experts k ∈ {1, 2, .…, m} for both criteria and alternatives.

  3. Step 4.

    Transform the preference relations into HFLTS using \( {E}_{G_H} \) function.

  4. Step 5.

    Obtain envelopes \( \left[{p}_{ij}^{k-},{p}_{ij}^{k+}\right] \) for each HFLTS.

  5. Step 6.

    Select a linguistic aggregation operator φ and obtain the pessimistic and optimistic collective preference relations (\( {P}_C^{-} \), \( {P}_C^{+} \)). The arithmetic mean given in Eq. (2) is used for φ:

$$ \overline{x}=\Delta \left(\frac{1}{n}\sum \limits_{i=1}^n{\Delta}^{-1}\left({s}_i,{\alpha}_i\right)\right)=\Delta \left(\frac{1}{n}\sum \limits_{i=1}^n{\beta}_i\right) $$
(2)

The S related 2-tuple set is characterized as S = [0.5, 0.5). The Δ : [0, g] → S function is provided by Eq. (3).

$$ \Delta \left(\beta \right)=\left({s}_i,{\alpha}_i\right) with\left\{\begin{array}{l}i= round\left(\beta \right)\\ {}\alpha =\beta -i\end{array}\right\} $$
(3)

where round assigns to β the integer number i ∈ {0, 1, …, g} nearest to β and Δ−1 : 〈S〉 → [0, g] is specified by Eq. (4).

$$ {\Delta}^{-1}\left({s}_i,{\alpha}_i\right)=i+\alpha $$
(4)
  1. Step 7.

    Compute the pessimistic and optimistic collective preferences for the alternatives by φ.

  2. Step 8.

    Build the vector of intervals \( \left({V}^R=\left({p}_1^R,{p}_2^R,\dots, {p}_n^R\right)\right) \) for the collective preferences \( \left({p}_i^R=\left[{p}_i^{-},{p}_i^{+}\right]\right) \).

  3. Step 9.

    Normalize the obtained interval utilities.

  4. Step 10.

    Calculate the weighted scores.

The MULTIMOORA method

The multi-objective optimization by ratio analysis (MOORA) method was introduced by Brauers and Zavadskas (2006). Brauers and Zavadskas (2010) extended this method, and it became more robust as MULTIMOORA (MOORA plus full multiplicative form). MOORA method consists of a ratio system and a reference point approach. In addition, where the priorities of the objectives of the model differ, the significance coefficient is used. MULTIMOORA is produced by adding the full multiplicative form to MOORA. This is not an independent MCDM method; it is based on the evaluation of other MOORA techniques and the ranking of the results by dominance.

The ratio system of MOORA

The ratio system starts by showing the values of different alternatives according to different objectives or criteria in the decision matrix given in Eq. (5).

$$ X=\left[\begin{array}{ccc}{x}_{11}& \cdots & {x}_{1n}\\ {}\vdots & \ddots & \vdots \\ {}{x}_{m1}& \cdots & {x}_{mn}\end{array}\right] $$
(5)

where i = 1, 2, …, n; n is the number of criteria, j = 1, 2, …, m; m is the number of alternatives, xij is the response of alternative j on criterion i, \( {x}_{ij}^{\ast } \) is a dimensionless number representing the response of alternative j on criterion i, and this number usually belongs to the interval [0;1]. The matrix is normalized using Eq. (6).

$$ {x}_{ij}^{\ast }=\frac{x_{ij}}{\sqrt{\sum \limits_{j=1}^m{x}_{ij}^2}} $$
(6)

If the desirable value of the indicator \( {x}_{ij}^{\ast } \) is maximum, it is added, and if the desirable value is minimum, it is subtracted. The total assessment of each alternative is calculated using Eq. (7) in which \( {y}_j^{\ast } \) is the total assessment of alternative j with respect to all criteria. i = 1, 2, …, g is the number of criteria that are maximized, and i = g + 1, g + 2, .., n is the number of criteria that are minimized.

$$ {y}_j^{\ast }=\sum \limits_{i=1}^g{x}_{ij}^{\ast }-\sum \limits_{i=g+1}^n{x}_{ij}^{\ast } $$
(7)

In principle, the decision-makers could give more importance to some criteria. Thus they could multiply the dimensionless number with a significance coefficient (si) as shown in Eq. (8).

$$ {y}_j^{\ast }=\sum \limits_{i=1}^g{s}_i{x}_{ij}^{\ast }-\sum \limits_{i=g+1}^n{s}_i{x}_{ij}^{\ast } $$
(8)

The final preference is determined by an ordinal ranking in descending order of the \( {y}_j^{\ast } \) (Brauers and Zavadskas 2013).

The reference point theory of MOORA

In the reference point theory, in addition to the ratio system, reference point (ri) is determined for each criterion as the maximum points if the objective is maximization and the minimum points if the objective is minimization. The distance of these determined points with each \( {x}_{ij}^{\ast } \) is calculated by Eq. (9).

$$ {r}_i-{x}_{ij}^{\ast } $$
(9)

The matrix created is calculated using the min-max metric of Tchebycheff given in Eq. (10). The results are ranked in ascending order. If the decision-maker assigns the relative importance to a criterion’s response on an alternative, Eq. (11) is used.

$$ \underset{(j)}{\min}\left\{\underset{(i)}{\max}\left|{r}_i-{x}_{ij}^{\ast}\right|\right\} $$
(10)
$$ \left|{s}_i{r}_i-{s}_i{x}_{ij}^{\ast}\right| $$
(11)

The full multiplicative form and MULTIMOORA

In the full multiplicative form, the introduction of weights is meaningless unlike the ratio method and the reference point theory of MOORA. This approach is expressed by Eqs. (12), (13), and (14):

$$ {U}_j^{\prime }=\frac{A_j}{B_j} $$
(12)
$$ {A}_j=\prod \limits_{i=1}^g{x}_{ij} $$
(13)
$$ {B}_j=\prod \limits_{i=g+1}^n{x}_{ij} $$
(14)

where i = 1, 2, …, n; n is the number of criteria, j = 1, 2, …, m; m is the number of alternatives, g is the number of criteria to be maximized, (n – g) is the number of criteria to be minimized, and \( {U}_j^{\prime } \) is the utility of alternative j with criteria to be maximized and criteria to be minimized.

The MULTIMOORA is not a stand-alone method. The MULTIMOORA performs a final evaluation based on the dominance of the rankings determined according to the ratio system, the reference point theory, and the full multiplicative form. For details on the theory of ordinal dominance, readers may refer to Brauers and Zavadskas (2013).

Proposed approach

The proposed approach is explained in detail in this subsection. The flow chart of the proposed method adopted for the selection of AYPIKUT locations is given in Fig. 2. The proposed method consists of two phases including problem definition and data collection and GIS-based multi-criteria HFLTS and MULTIMOORA model (Fig. 2).

Fig. 2
figure 2

The flow chart of the proposed method

Full size image

A case study of Atakum district

Problem definition and data collection

The AYPIKUT has been designed for collection of both domestic waste vegetable oil and waste batteries. This study focuses on determining suitable locations for the AYPIKUT in Atakum using GIS-based multi-criteria HFLTS and MULTIMOORA methods. Atakum, as a district of Samsun, covers a total of 355 km2. The eastern coastal part of the district is the most densely populated. The study area is shown in Fig. 3.

Fig. 3
figure 3

The study area (Google Earth image ©)

Full size image

In this study, experts determined the seven most important criteria affecting the choice of location of the AYPIKUT as Cj, where C1 is to be maximized and C2–C7 are to be minimized.

  • C1: Population density

  • C2: Proximity to public institutions

  • C3: Proximity to housing complexes with more than 150 dwellings

  • C4: Proximity to shopping malls

  • C5: Proximity to tram stops

  • C6: Proximity to schools

  • C7: Proximity to parks

A total of 88 items have been identified for consideration by these criteria: 13 public institutions and organizations, 9 housing complexes, 3 shopping malls, 20 tram stops, 32 schools, and 11 parks or green areas. The location of the 88 items was obtained using a GNSS (Global Navigation Satellite System) handheld receiver and satellite images.

Results and discussion

Production of maps

In the study, a few types of spatial analysis were done, and several types of maps were produced based on the population of 57 neighbourhoods and the location of 88 criteria. ArcGIS software was used to build the geodatabase and perform the spatial analysis for the criteria. The study used two types of spatial analysis: density analysis and distance analysis. Density analysis calculates and shows where features are concentrated, such as the population distribution in the city. Distance analysis determines the distance from the items; in this study, Euclidean distance analysis was used. The density analysis of the neighbourhoods and the Euclidean distance analysis of each criterion are presented in Fig. 4.

Fig. 4
figure 4

GIS layer of each criterion: (a) population density (C1), (b) proximity to public institutions (C2), (c) proximity to housing complexes (C3), (d) proximity to shopping malls (C4), (e) proximity to tram stops (C5), (f) proximity to schools (C6), and (g) proximity to parks (C7)

Full size image

In Fig. 4a–g, dark zones have higher pixel value according to the results of Euclidean distance analysis and density analysis and indicate unsuitable locations depending on the purpose of the study.

Determination of the priorities of criteria

The weights of criteria were determined using the multi-criteria HFLTS method. Table 2 presents linguistic evaluations of the experts for the criteria. The comparative linguistic expressions generated by GH were converted into HFLTS using \( {E}_{G_H} \) function. For instance, expert 1’s preference of C1 with respect to C2 is ‘greater than high importance’ in linguistic terms (Table 2). This preference relation is transformed into HFLTS using \( {E}_{G_H} \) function, where \( {E}_{G_H} \) (greater than high importance) = {very high importance, absalute importance}, and it can be represented as a discrete set {vh, a} and then as the interval [vh, a]. The envelopes obtained for the evaluations of three experts are presented in Table 7. In the next step, pessimistic and optimistic collective preference relations were obtained using the scale for the linguistic terms given in Table 3. In the scale, the value 0 means ‘no importance’, whilst the value 6 means ‘absolute importance’. The pessimistic and optimistic collective preference values are presented in Tables 8 and 9, respectively. The arithmetic mean was used for the linguistic aggregation operator to obtain these collective preference relations. For example, the pessimistic collective preference value for C1 in relation to C2 is calculated as follows:

Table 2 Expert evaluations for the criteria
Full size table
Table 3 The scale for HFLTS
Full size table
$$ {\displaystyle \begin{array}{l}{P}_{C_{12}}^{-}=\Delta \left(\frac{1}{3}\left({\Delta}^{-1}\left( vh,5\right)+{\Delta}^{-1}\Big(h,4\left)+{\Delta}^{-1}\right(h,4\Big)\right)\right)\\ {}\kern1.5em =\Delta \left(\frac{1}{3}\left(5+4+4\right)\right)=\Delta (4.33)=\left(h,+0.33\right)\end{array}} $$

Similarly, the following process is performed to calculate the optimistic collective preference value for C1 with respect to C2:

$$ {\displaystyle \begin{array}{l}{P}_{C_{12}}^{+}=\Delta \left(\frac{1}{3}\left({\Delta}^{-1}\left(a,6\right)+{\Delta}^{-1}\Big(a,6\left)+{\Delta}^{-1}\right(a,6\Big)\right)\right)\\ {}\kern1.5em =\Delta \left(\frac{1}{3}\left(6+6+6\right)\right)=\Delta (6)=\left(a,+0\right)\end{array}} $$

The weights of criteria given in Table 4 were obtained by using the values of the pessimistic and optimistic collective preferences. For example, the linguistic intervals, interval utilities, midpoint, and weight are calculated for the first row in Table 4 as follows.

Table 4 Weights of the criteria
Full size table

The pessimistic and optimistic collective preferences are (h,+ 0.06) and (vh,− 0.11) for C1. These preferences are expressed as the linguistic intervals [(h,+ 0.06),(vh,− 0.11)]. Next, the linguistic intervals are transformed into interval utilities. As h corresponds to 4, (h,+ 0.06) is expressed as 4.06. Similarly, vh corresponds to 5, and (vh,− 0.11) is expressed as 4.89. The midpoint refers to the point equidistant to these two points and is calculated as the arithmetic mean of the two points. This value is calculated as 4.47. Finally, the weight of C1 is obtained as 0.213 by normalizing this midpoint.

As shown in Table 4, C1 (population density) is the most significant criterion affecting the selection process with a weight of 0.213; C3 (proximity to housing complexes with more than 150 dwellings) is the least important with 0.081. Population density is a crucial criterion for locational analysis of MSW collection boxes. Studies done by Chalkias and Lasaridi (2009) and Khan and Samadder (2016) support our findings. Vijay et al. (2008), Chalkias and Lasaridi (2009), and Khan and Samadder (2016) considered only the road network of the study area to identify the optimized placement of bins. We considered the tram network of the study area in addition to the road network, and the results indicated that C5 (proximity to tram stops) was the second most important criterion after the population density.

Identification of locations using GIS

A density-based criterion and distance-based six criteria were combined into a single normalized map using weights determined by the multi-criteria HFLTS method. The weighted map (Fig. 5a) was classified into 9 classes according to the range of value as shown in Fig. 5b. Areas with a value between 0 and 0.20 were identified as best for the locations of the AYPIKUT (Kabak et al. 2018).

Fig. 5
figure 5

(a) Weighted map and (b) classified weighted map

Full size image

The areas with a value between 0.01 and 0.20 shown in Fig 5b were further classified into 9 sub-classes using the weighted value (Fig. 6a, b). Suitable locations for the AYPIKUT were identified using the following three criteria:

  • The AYPIKUT must be in a location with a weighted value less than 0.2.

  • Each AYPIKUT must be at least 1000 m from every other AYPIKUT.

  • The location of an AYPIKUT must be within 100 m of one criteria point and within 500 m of three criteria points.

Fig. 6
figure 6

(a, b) The suitable locations for AYPIKUT

Full size image

Fifteen (15) AYPIKUT locations (A1–A15) were identified using these criteria and taking into consideration the easily accessible places, as shown in Fig. 6a, b. Table 5 presents the normalized values for the 15 selected locations and confirms that all AYPIKUT locations had a value less than 0.2. A8 was found to have the best value based on C1, which was deemed the most important criterion for selection process, with a value of 0.016.

Table 5 The values for AYPIKUT locations based on the criteria
Full size table

Ranking the alternatives using the MULTIMOORA method

The alternative AYPIKUT locations (A1–A15) were ranked with the MULTIMOORA method. The normalized values for AYPIKUT locations given in Table 5 were used to form the initial decision matrix, and calculations were made according to Eqs. (6), (7), (8), (9), (10), (11), (12), (13), and (14). Table 6 shows the rankings of the alternative AYPIKUT locations according to the ratio system (RS), the reference point theory (RPT), the full multiplicative form (FMF), and the final ranking of MULTIMOORA.

Table 6 The rankings of alternatives
Full size table

From Table 6, A5 is identified as the most suitable location, and A6 as the least suitable location for an AYPIKUT according to the MULTIMOORA method. Moreover, A1, A2, and A3 have the same rank. Unlike many MCDM approaches, the MULTIMOORA generates an integrative outcome by combining the results of three ranking methods. It is implemented effectively in site selection problems (Kabak et al. 2018; Rahimi et al. 2020; Lin et al. 2020).

The results indicated the ability of the proposed model to select suitable locations for the waste collection box. The model can be used in similar studies for the economic recovery of solid wastes.

Conclusion

MSWM is a significant item in the municipal budget, and most of the costs are spent on the collection and transportation of solid waste. The first condition for economic recovery from solid waste is separate collection, with the primary approach using collection boxes. This paper provides a scientific framework for determining AYPIKUT locations in the Atakum, which has the highest population growth rate in Samsun city. For the solution, first, a total of 88 items have been identified for consideration by seven criteria elicited from the insights of experts, and spatial analyses were performed. Then, multi-criteria HFLTS was applied to determine the weights of the criteria, and the possible suitable locations were determined using GIS. Finally, MULTIMOORA was used to evaluate the 15 suitable locations. The use of GIS-based multi-criteria HFLTS and MULTIMOORA affords a practical approach that increases the accuracy of site selection for AYPIKUT and the efficiency of the decision-making process whilst reducing the complexity of the research. Another advantage of the proposed approach is the use of linguistic term sets, as decision-makers often prefer linguistic assessment to form the decision matrix. Decision-makers may also have problems in identifying linguistic terms and may need flexibility in their evaluation. This challenge is overcome by context-free grammar, ensuring that the linguistic evaluation elicited from experts is effectively preserved without any loss of knowledge.

As future work, this study may be extended to other districts of Samsun and the other crowded cities in Turkey. Criteria may be amended and the number of criteria may be changed according to the characteristics of the study area. The number of required AYPIKUT may be determined according to the amount of waste produced daily and weekly, or by considering the per capita MSW generation rate and extent of the service area of an AYPIKUT.