Municipal solid waste management with cost minimization and emission control objectives: A case study of Ankara

Sustainable Cities and Society 52 (2020) 101807

Contents lists available at ScienceDirect

Sustainable Cities and Society journal homepage: www.elsevier.com/locate/scs

Municipal solid waste management with cost minimization and emission control objectives: A case study of Ankara

T

Melika Mohsenizadeha, , Mustafa Kemal Turala, Elçin Kentelb ⁎

a b

Department of Industrial Engineering, Middle East Technical University, Ankara, Turkey Department of Civil Engineering, Middle East Technical University, Ankara, Turkey

ARTICLE INFO

ABSTRACT

Keywords: Municipal solid waste management Sustainable transportation CO2 emission Facility location Bi-objective optimization

Proper management of municipal solid waste (MSW) has been a crucial aspect of every society due to its social, environmental, and economic impacts. Operations research techniques have frequently focused on cost minimization objectives in locational planning of municipal solid waste management (MSWM) systems. However, transportation constitutes an integral part of this system producing a considerable amount of greenhouse gas (GHG) emissions. Therefore, sustainable management of this system with GHG emissions minimization considerations is necessary to preserve the resources and protect the environment. In this study, a bi-objective optimization model is proposed to minimize system cost and carbon dioxide (CO2) emission resulting from transportation activities in locational planning of MSWM systems. The proposed model is applied to MSWM system of Ankara to introduce transfer stations (TSs). Two extensions of the current system are examined, namely, the extended and hybrid systems, where MSW is only transported through TSs in the former, while direct shipments are also allowed in the latter. For both extensions, it is observed that with no or little increase in cost, considerable savings in emission can be achieved. Simulation analyses show that CO2 emission and cost are not subject to a considerable change due to speed variations of vehicles.

1. Introduction Increasing population, urbanization, and economic growth result in generation of higher levels of solid waste throughout cities raising serious issues regarding public health, environmental degradation, and natural resource exhaustion (Mohammadi, Jämsä-Jounela, & Harjunkoski, 2019). Efficient and timely management of solid waste throughout cities is imperative due to the inherent pollutants which can seriously risk human health, ecosystems, soil, water, and atmosphere (Edalatpour, hashem, Karimi, & Bahli, 2018). Moreover, treatment and processing operations and transportation related activities within solid waste management (SWM) systems have adverse environmental, atmospheric, and social impacts. Such issues call for a sustainable design and management of solid waste systems considering interconnections of various SWM functions, i.e., collection, transportation, treatment, recycling, and landfilling. Municipal solid waste (MSW) includes waste from residential, commercial, and institutional sources and excludes industrial, construction, and hazardous waste (USEPA, 2015). Sustainable management of MSW systems requires the incorporation of economic, environmental, and social aspects, termed as three dimensions of

⁎

sustainability (WCED, 1987). Sustainability considerations may entail introducing new technologies and facilities within the municipal solid waste management (MSWM) systems. In recent years, remote regional treatment plants and sanitary landfills have emerged in MSWM systems due to land shortage near urban centers and increased public concerns regarding potential human health risks of these facilities. As a result, the conventional structure of MSWM systems has changed due to the emergence of longer transportation routes between existing and introduced facilities (Yadav, Bhurjee, Karmakar, & Dikshit, 2017). Accordingly, transportation, which is an integral element of MSWM systems, gains more importance than before due to the increase in frequency and duration of vehicle trips. Transportation has numerous negative implications on the environment and societies especially in urban areas having large populations such as natural resources depletion, air pollution and greenhouse gas (GHG) effect, noise, land wear, acidification, human and ecosystem toxicity (Knörr, Seum, Schmied, Kutzner, & Anthes, 2011). Among these impacts, the resulting GHG emissions, mainly CO2, methane (CH4), and nitrous oxide (N2O) (United States Environmental Protection Agency, 2016), have significant adverse effects on social and physical health of public directly and indirectly via atmospheric

Corresponding author. E-mail address: [email protected] (M. Mohsenizadeh).

https://doi.org/10.1016/j.scs.2019.101807 Received 5 March 2019; Received in revised form 30 May 2019; Accepted 26 August 2019 Available online 29 August 2019 2210-6707/ © 2019 Elsevier Ltd. All rights reserved.

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

pollution, ozone layer depletion, and global warming potentials just to name a few. According to the United States (US) Environmental Protection Agency (United States Environmental Protection Agency, 2016), the transportation sector is the largest contributor to GHG emissions in the US accounting for 28% of total emissions. Among all economic sectors, between the years 1990 and 2016, the increase in GHG emissions is the largest in the transportation sector. This can be attributed to increasing transportation activities by cars and heavy and light-duty vehicles. More than 95% of GHG emissions in the US transportation sector amounts to CO2 emission in 2016 corresponding to 34% of total US CO2 emission. In recent years, there has been a considerable increase in the number of studies that aim to minimize GHG emissions in logistics networks. For example, Bektaş and Laporte (2011), in a pioneering study, introduce the pollution-routing problem (PRP) which is an extension of the vehicle routing problem with time-windows. The PRP incorporates the harmful effects of transportation in routing of vehicles by assessment of CO2 emission with respect to vehicles’ load, speed, and technical characteristics. A cost function composed of fuel, emission, and labor costs is minimized in the objective of the proposed model. Afterwards, there have been several studies focusing on extensions or modifications of the PRP. For example, Franceschetti, Honhon, Woensel, Bektaş, and Laporte (2013) consider time-dependent PRP in which the effect of traffic congestion through different time periods is included in order to assess optimal speed of vehicles along with their departure time. Also, Demir, Bektaş, and Laporte (2014) introduce the bi-objective PRP where two conflicting objectives, namely, fuel consumption and driving time of vehicles, are minimized. For a detailed review of studies regarding green location and routing problems, the reader is referred to the recent survey by Dukkanci et al. (2019, chap. 7). Transportation constitutes an integral part of MSWM systems producing a considerable amount of GHG emissions (Tavares, Zsigraiova, Semiao, & Carvalho, 2009) and should be incorporated within different decision making levels of this system not only considering its economic impacts but also environmental and social ones. MSWM involves strategic, tactical, and operational level decisions where operations research techniques can be utilized to tackle integrated decision making problems in this context (Ghiani, Laganà, Manni, Musmanno, & Vigo, 2014). These techniques widely employ mixed integer linear programming (MILP) and are commonly centered around economic aspects. However, incorporating various aspects in different decision making levels within a multi-objective optimization framework is imperative to have a comprehensive management of MSW systems. A MSWM system can be decomposed into two sub-systems, namely, regional MSWM system and collection system. The network design, a strategic level decision, is related to the regional planning of MSW, while fleet deployment, scheduling, and routing within the regional MSWM or the collection systems are among the tactical and operational level decisions (Ghiani et al., 2014). Modeling an MSWM system requires studying it as a multi-echelon supply chain in which activities such as MSW generation, collection, transportation, recovery, recycling, treatment, and landfilling take place (Ghiani et al., 2014). During the last few decades, an emerging attention has been paid to the sustainable management of these activities (USEPA, 2015). However, societal and environmental side effects of transportation activities within the logistics networks are considerable enough necessitating their incorporation into the planning of MSWM systems. In this study, we aim to investigate the impact of CO2 emission from transportation activities on locational planning of regional MSWM systems. A bi-objective facility location-allocation problem is considered and an MILP model is proposed to determine the optimal locations of transfer stations (TSs), processing and treatment (P&T) facilities, and sanitary landfills in an MSWM system; allocate MSW flows to such facilities; quantify the required number of vehicle trips; and

determine the travel speeds of vehicles by minimizing the total daily system cost as well as CO2 emission from transportation activities. We implement the Comprehensive Modal Emission Model (CMEM) (Barth, Younglove, & Scora, 2005) to calculate the fuel consumption and resulting emissions of vehicles in the system with respect to their speed, load, travel distance, and technical characteristics. In order to generate the efficient solutions of the proposed model, the ϵ-constraint method is implemented. The proposed model is applied to the MSWM system of Ankara by utilizing a geographical information systems (GIS) tool to investigate the sustainability benefits obtained by introducing TSs into the current system. Two extensions of the current system are considered; the extended one, where MSW is only transported through TSs and the hybrid one, where direct shipments are allowed in addition to indirect shipments through TSs. Even though the total daily system cost and CO2 emission from transportation activities are conflicting objectives, we have shown that for the Ankara case, CO2 emission can be decreased by a large amount with little or no increase in system cost. Moreover, the impact of speed variations of vehicles on the resulting CO2 emission and system cost is investigated by performing simulation analyses. The remainder of this paper is structured as follows. The next section provides a literature review with a main focus on location problems within waste management systems. Section 3 describes the network of the considered MSWM system, introduces the notation used, and describes the employed emission model. The mathematical model of the bi-objective MSWM problem is derived in Section 4. In Section 5, we consider the MSWM system operating in Ankara, Turkey as a case study and present the features of the considered transportation network. Computational experiments related to our case study including simulation analyses where the speeds of vehicles are considered as random variables are presented in Section 6. Finally, Section 7 covers the conclusion and future research directions. 2. Literature review In this section, we review the literature with a main focus on location problems within SWM systems. For a survey of studies on strategic and tactical issues within SWM systems from an operations research point of view, the reader is referred to Ghiani et al. (2014). In addition, Farahani, Fallah, Ruiz, Hosseini, and Asgari (2019) present a recent review regarding facility location problems within urban service systems including waste management systems. According to Li and Huang (2010), the first study on economic optimization of SWM systems is conducted by Anderson and Nigam (1968). Subsequently, the objective of cost minimization in SWM problems become popular among researchers. Fiorucci, Minciardi, Robba, and Sacile (2003) develop a decision support system for integrated planning of MSWM by solving a constrained non-linear optimization problem. They consider the composition of MSW and aim to determine the optimal MSW flows that should be sent to disposal, recycling, and treatment plants, along with the optimal selection of such facilities in the system. The number and locations of TSs, however, are fixed a priori in their model. A cost minimization objective including transportation, maintenance, and recycling costs as well as potential economic benefits of the system are considered in their mathematical optimization model. A relatively similar general purpose model concerning management of MSW is studied by Costi, Minciardi, Robba, Rovatti, and Sacile (2004) with a particular focus on environmental impacts of the overall MSWM system. The important feature of their study is the detailed analyses of chemical composition of MSW to control resulting emissions in incinerators as well as noxious chemicals appearing in refused derived fuel and stabilized organic material. However, in both of these studies, transportation related emissions are not considered. The problem of locating TSs in MSWM systems has been studied by different researchers because of their particular importance as 2

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

transshipment nodes. TSs serve as consolidation points in MSWM systems by linking the collection system to the treatment and disposal facilities where collection vehicles transfer collected MSW, typically in a compacted form, to large-volume vehicles available in TSs to be shipped to treatment and disposal facilities. This procedure results in flexibility in locating treatment and disposal facilities, increases the efficiency of collection and regional MSWM systems, and reduces the social and environmental damages resulting from transportation activities (USEPA, 2002). Kulcar (1996) suggests a two-phase optimization problem to determine the required number of TSs in SWM system of Brussels. In the first phase, the terminal site (a TS or an incinerator) for every collection route is distinguished and the TSs are selected among the candidate ones. In the second phase, the impact of the collection procedure on the required number of depots is investigated to reduce the number of necessary depots in the system where waste collectors stay overnight. Furthermore, the objective of both phases is to minimize overall costs. Antunes (1999) proposes an MILP model to study the MSWM system in central Portugal at regional level by combining the elements of the pmedian problem and capacitated facility location problem with transshipments. TSs and sanitary landfills are located by considering a minimum cost objective which comprises of annual costs of TSs and transportation costs. Ferri, de Lorena Diniz Chaves, and Ribeiro (2015) investigate inclusion of material recovery facilities functioning as intermediate nodes, similar to TSs, in a reverse logistics network. The incoming MSW from the sources undergoes sorting and consolidation operations through these facilities to be shipped to treatment and disposal facilities in a more economic way. In another study, Yadav, Karmakar, Dikshit, and Vanjari (2016) propose a non-linear mixed integer programming (NLMIP) model to site TSs and determine their capacities in an economically efficient manner. The proposed model is applied on MSWM system of Nashik, India by implementing a GIS analysis framework which enables the authors to employ precise geo-spatial parameters in their mathematical model. An interval-valued facility location model is developed in a relatively similar study by Yadav, Karmakar, Dikshit, and Bhurjee (2018) to investigate the impact of uncertainty on economically best locations of TSs. Rathore and Sarmah (2019) propose an MILP model for locating TSs in an economically optimal way considering waste segregation and unsegregation scenarios and apply it on the MSWM system of Bilaspur, India. Asefi, Lim, Maghrebi, and Shahparvari (2019) investigate the problem of locating integrated SWM system components, including TSs, considering multiple waste types sorted and separated at generation sources. The compatibility of hazardous waste types and treatment technologies is addressed in their study as well. There are various other studies in the literature that investigate the feasibility of locating TSs from different perspectives with cost control objectives. For example, Badran and El-Haggar (2006) study the problem of locating TSs in the MSWM system of Port Said, Egypt; Kirca & Erkip, 1988 study a multistage decision problem to evaluate the optimal configuration, i.e., location, technology, and capacity, of TSs in a solid waste system; Eiselt (2007) considers the problem of locating TSs and landfills where a discount factor is assigned for transferring waste through TSs. Increased environmental awareness, public concerns, and legislation call for a sustainable management of supply chains including MSW systems. Therefore, within the scope of MSWM, decision making methodologies should address issues related to human welfare and environment in addition to economic considerations (Bing et al., 2016). Multi-objective decision making approaches can be utilized to perform a comprehensive optimization within waste management context by taking various criteria and objectives into consideration and putting in a balance between them. In a study performed by Erkut, Karagiannidis, Perkoulidis, and Tjandra (2008), a multi-objective facility location-allocation model is proposed at the regional level to locate TSs, material recovery plants, incinerators, and sanitary landfills together with selecting their appropriate technologies. The environmental and

economic dimensions are included into the objective function and a fair solution is obtained by implementation of lexicographic minimax approach. The considered environmental criterion comprises of generated GHG emissions by facilities, MSW amount that is landfilled, and energy and material recovery rates. The applicability of their proposed model is tested on the MSWM system of central Macedonia in North Greece. A reasonable balance and adjustment between economic feasibility, environmental sustainability, and social justice is required in designing and operating an integrated sustainable MSWM system. Some researchers have focused holistically on all three dimensions of sustainability in MSWM systems from different perspectives. Yu and Solvang (2017) propose a comprehensive multi-objective location-allocation model in order to minimize system cost, GHG emissions from transportation activities, and environmental impact of treatment facilities. The environmental impact is included to balance the adverse effects of the implemented technologies among communities and minimize health hazards to neighboring residences. The weighted sum method is implemented in their study to aggregate the objectives and obtain an efficient solution. Mirdar Harijani, Mansour, and Karimi (2017a) propose a multiobjective MILP model to introduce recycling and disposal facilities into an MSWM system. They distinguish the location, typology, and capacity of treatment facilities as well as MSW streams among facilities by maximizing the profit of the MSWM system, minimizing resulting CO2 emission from treatment facilities and transportation activities, and maximizing the social impacts of the considered MSWM network. Social impacts of the treatment facilities are assessed by implementing the fuzzy analytic hierarchy process which assigns social scores to involved facilities. In a relatively similar study, Asefi and Lim (2017) devise an optimization model by combining the elements of multi-objective programming and multi-criteria decision analysis for designing a sustainable integrated SWM system. The developed model incorporates three objectives, namely, minimization of facilities’ establishment costs, minimization of transportation cost, and maximization of the suitability of the system's components. To mention important features of this study, we can point out consideration of MSW composition, selection of processing and treatment technologies that are compatible with the typology of MSW and residues, and inclusion of stakeholders’ preferences into their mathematical model. Mirdar Harijani, Mansour, and Karimi (2017b) propose an MILP model to introduce new treatment facilities in an MSWM system in a multi-period setting. Their model maximizes the profit of the MSWM system with a budget restriction. The environmental dimension of sustainability is embedded within the objective considering the environmental costs arising from processing of MSW and emission resulting from transportation activities. The social scores of treatment facilities are distinguished by using the social life cycle assessment method and a lower bound is assigned to overall social score of the network in a constraint. A tri-objective MILP problem is formulated by Habibi, Asadi, Sadjadi, and Barzinpour (2017) to locate facilities, allocate MSW flows, select the optimal capacities and technologies of facilities, and quantify the number of vehicles needed to transport MSW among system components which is in practice equal to the number of required trips. The objective functions aim to maximize the net profit of the MSWM system, minimize the GHG emissions resulting from transportation and processing of MSW in facilities, and minimize the side effects of facilities on residents of population centers. Heidari, Yazdanparast, and Jabbarzadeh (2019) assess economic and environmental criteria considering similar factors, while incorporating the social criterion by maximizing the job opportunities brought by the MSWM system as a result of establishment of new facilities. Santibañez-Aguilar, PonceOrtega, González-Campos, Serna-González, and El-Halwagi (2013) optimize the supply chain network of an MSWM system by proposing a biobjective MILP model considering maximization of the net profit of the system and the reused waste. The quantity of reused waste, which is 3

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Table 1 Summary of the reviewed literature on facility location problems within waste management systems. Study

Method

Multi-obj.

a

WSb

Sustainabilityc Ec.

Kirca & Erkip, 1988 Kulcar (1996) Antunes (1999) Fiorucci et al. (2003) Costi et al. (2004) Badran and El-Haggar (2006) Eiselt (2007) Erkut et al. (2008) Santibañez-Aguilar et al. (2013) Ferri et al. (2015) Yadav et al. (2016) Yu and Solvang (2017) Mirdar Harijani et al. (2017a) Asefi and Lim (2017) Mirdar Harijani et al. (2017b) Habibi et al. (2017) Yadav et al. (2018) Rathore and Sarmah (2019) Asefi et al. (2019) Heidari et al. (2019) This study a b c d e f

MILP MILP MILP NLMIP NLMIP MILP MILP MILP MILP MILP NLMIP MILP MILP MILP MILP MILP MILP MILP MILP NLMIP MILP

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

✓ ✓

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

DVd

En.

Facl.

So.

✓ ✓ ✓

✓

✓ ✓ ✓ ✓ ✓

✓ ✓ ✓ ✓ ✓

✓ ✓

✓ ✓

TSs LA LA LA LA LA LA LA LA LA, LA, LA, LA LA, LA LA, LA, LA LA LA LA, LA,

✓ ✓ ✓

Tech. Cap. Cap. Cap., Tech. Cap., Tech. Cap., Tech., Tri.

Cap., Tech., Lab. Tri.

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

e

VEf P&T

✓ ✓ ✓ ✓ ✓

GIS

Case Study

✓

Turkey Belguim Portugal Italy Italy Egypt Canada North Greece Mexico Brazil India — Iran Iran Iran Iran India India Iran Iran Turkey

SL

✓ ✓ ✓ ✓ ✓

✓ ✓ ✓ ✓ ✓

✓ ✓ ✓

✓ ✓ ✓

✓

✓

✓

EC EC EC EC

EC EM

✓ ✓ ✓ ✓ ✓ ✓

Multi-obj.: multi-objective. WS: waste segregation. Ec.: economic, En.: environmental, So.: social. DV: decision variables, LA: location-allocation, Cap.: capacity, Tech.: technology, Lab.: number of required labor, Tri.: number of required trips. Facl.: facilities to be located, SL: sanitary landfill. VE: vehicular emissions, EC: emission coefficient, EM: emission model.

inversely related to the amount of MSW landfilled, is used as an indicator for social and environmental criteria. The important features of the reviewed studies related to facility location problems within waste management systems along with the current study are highlighted in Table 1. In the studies conducted by Habibi et al. (2017), Mirdar Harijani et al. (2017a, 2017, and Heidari et al. (2019), GHG emissions of vehicles are taken into account and are calculated with respect to transported MSW flows and an emission coefficient which depends on the distance traveled. Whereas, the effects of vehicle speed and technical characteristics, speed limits and general congestion patterns on road segments on GHG emissions are left out. Studies that incorporate the social aspect of sustainability consider various indicators and methods, e.g., social life cycle assessment, creation of job opportunities, visual pollution, and the amount of reused waste. Measurement of all influencing aspects in social issues is impractical because various disciplines and stake-holders should be taken into account. In the present study, the social issues are not quantified explicitly, but they can be incorporated into the preliminary selection of the paths and potential locations of facilities. Moreover, minimization of CO2 emission from transportation activities has indirect social benefits such as lower negative atmospheric impacts on societies and improved quality of life and human welfare. There are also some studies in the literature that consider sustainability in tactical and operational planning of MSWM systems. Yu, Solvang, Yuan, and Yang (2015) formulate a non-linear mathematical programming model at operational level considering minimization of the overall cost, risk, and waste disposal value as distinct objectives in a multi-period setting. Tan, Lee, Hashim, Ho, and Lim (2014) and Mohammadi et al. (2019) develop MILP optimization models for sustainable management of MSW systems of Iskandar Malaysia, Malaysia and a region of Mexico, respectively, by incorporating the economic benefits of end-products and energy generated from treatment and processing of MSW in a multi-period setting. The main focus of these studies is centered around treatment and processing of MSW as well as

the allocation of MSW to related facilities where locational planning and transportation cost assessment are not respected. Considering the key features and gaps of the studied literature, our study aims to focus on the problem of locating TSs, P&T facilities, and sanitary landfills in an MSWM system based on two perspectives, namely, economic feasibility and environmental sustainability. The environmental sustainability criterion minimizes the resulting CO2 emission from transportation activities in this system which is calculated in a detailed manner. To the best of our knowledge, such detailed analyses of GHG emissions from transportation services in locational planning of MSWM systems at regional level have not yet been carried out in the literature. We assume that different types of vehicles operate between successive levels of the considered MSWM network and technical characteristics of a vehicle also impact CO2 emission in addition to its speed, load, and travel distance. Accordingly, the proposed model can also be used in making decisions regarding fleet composition by looking at the trade-off between costs of different vehicle types and the resulting CO2 emission. By simulation analyses, we show that, in the case study of Ankara, once the locations of the TSs are determined, the two considered objectives are not subject to considerable changes with respect to speed variations of vehicles. Note that, social sustainability aspects can also be integrated into the proposed model by considering them in the selection of candidate facility locations as well as the determination of the paths between facilities in successive levels of the network. 3. MSWM network description and notation In this study, a bi-objective facility location problem is developed in order to determine the locations of facilities selected among candidate ones so as to minimize total daily MSWM system cost and daily CO2 emission of transportation activities. The total daily MSWM system cost takes into account fixed and variable costs of facilities, variable transportation cost, fuel cost resulting from transportation activities, and the revenue generated from MSW in processing and treatment facilities. 4

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Transportation constitutes an essential part of MSWM systems as MSW flows among different levels of the system for shipment, transshipment, processing and treatment, and final disposal. In order to incorporate the impact of transportation services into strategic decision making in locational planning of MSWM systems at regional level, we should first have a general overview concerning the configuration of MSWM systems. A general framework of an MSWM network respected in this study has four levels with the following organizations. First level of the considered MSWM network corresponds to MSW generation sources, denoted by set I, where source i ∈ I has a deterministic generation rate of Gi (tonne/day). In this study, MSW collection activities are not taken into account explicitly and it is assumed that the relative activities are performed by collection system of each municipality operating separate from regional sector who is responsible for overall treatment and disposal of collected MSW. Second level accounts for TSs functioning as intermediate facilities to link MSW collection system to processing, treatment, and disposal facilities. Presence of TSs as consolidation and transshipment points would result in transportation efficiency, reduced maintenance costs and driver wages, lower fuel consumption and atmospheric pollutants, and less traffic congestions. In our study, the set of TSs is denoted by J. The index set J can be partitioned into two subsets; J1 corresponding to candidate TSs and J2 corresponding to existing TSs. We assume that transfer station j ∈ J has a daily capacity of CTj tonnes, daily fixed cost of FCTj €, and daily variable cost of VCTj € per tonne of MSW transferred. Third level composes of a set K of MSW P&T facilities. MSW is transferred from TSs by large-volume vehicles to these facilities. Separators, recycling plants, refused derived fuel facilities, composting plants, incinerators, and waste to energy plants are some examples of third level facilities where MSW is processed and treated while producing residues. In this study it has been assumed that various types of facilities operate together in a single P&T mega-facility. As pointed out in USEPA (2002), existence of such mega-facilities is beneficial as considerable construction and maintenance costs would be compensated by handling a high volume of MSW. Besides, associated processing and tipping expenses per unit of MSW would be cut off. The index set K is divided into two subsets; K1 corresponding to candidate P&T facilities and K2 corresponding to existing P&T facilities. We assume that P&T facility k ∈ K has a daily capacity of CPk tonnes, daily fixed cost of FCPk €, and daily variable cost of VCPk € per tonne of MSW processed and treated. Lastly, the average monetary benefits obtained by processing and treatment of MSW in P&T facility k ∈ K is considered as Ck € per tonne of MSW entering facility k. The constant Ck depends on the type of P&T mega-facility and implemented technologies there. Fourth and final level corresponds to a set L of sanitary landfills which are the final destination of MSW that could not be processed or treated. The residues from third-level facilities are sent to sanitary landfills for final disposal. We assume that a fraction σk of MSW entering P&T facility k ∈ K is sent to sanitary landfills. The fraction σk called the residual coefficient of P&T facility k depends on the implemented technologies and therefore may change from facility to facility. In a similar manner, the set L can be partitioned into L1 and L2 which are the sets of candidate and existing sanitary landfills in the system, respectively. Also, it is assumed that sanitary landfill l ∈ L has a daily capacity of CLl tonnes, daily fixed cost of FCLl €, and daily variable cost of VCLl € per tonne of MSW landfilled. MSW sources and facilities represent nodes of the considered MSWM network. A single path is defined to connect each pair of nodes in successive levels of the network. Preliminary selection of the paths and potential locations of facilities should incorporate variety of factors relative to environmental and social impacts in addition to economic suitability factors such as land price, reachability, and vicinity to urban centers. Reasonable distance from residential, recreational, and institutional centers; risk of accident and pollution; natural resources contamination; and population exposure are examples of such factors

which can be taken into account by utilizing GIS analytic tools or life cycle assessment methods accurately. However, such preliminary analyses are not considered in this study and we assume that vehicles of type 1, 2, and 3 perform transportation activities on the shortest paths connecting MSW sources to TSs, TSs to P&T facilities, and P&T facilities to sanitary landfills, respectively. Moreover, each shortest path on the network is divided into a number of road segments with respect to different speed limits imposed by local or governmental authorities directly and by traffic congestion indirectly. Note that, Temporal issues arising from traffic congestions in different time periods and the subsequent impacts on scheduling of vehicles are not addressed in this study and the average traffic congestion on each road segment is taken into account in determining the travel speeds of vehicles. We denote the sets of road segments on the shortest paths connecting MSW source i to TS j, TS j to P&T facility k, and P&T facility k to sanitary landfill l by Tij, Pjk, and Lkl, respectively. The travel speed of a vehicle on a road segment is evaluated explicitly considering its fuel consumption (FC), quantified by CMEM model explained in Section 3.1, and segment-wise speed limits. We present details of the implemented emission model in Appendix A. 3.1. Calculation of FC and CO2 emission Transportation is an essential and prevalent element in every MSWM network producing a considerable amount of GHG emissions with serious contributions to the greenhouse effect and acid rains (Tavares et al., 2009). The adverse atmospheric impact of transportation services in locational planning of an MSWM system is addressed in this study by controlling the resulting CO2 emission. As the amount of emitted CO2 from a vehicle is directly proportional to its FC (Franceschetti et al., 2013), we utilize the CMEM developed in Barth et al. (2005) for heavy-duty vehicles to calculate the FC rates of vehicles. According to CMEM, the FC of a heavy-duty vehicle can be calculated as

F=

1

z + v

2z

+

3z

+

4v

2z

(1)

where F is the amount of fuel consumed in liter (L) by a heavy-duty vehicle traversing a distance of z meters (m) with a constant speed of v meters/second (m/s) carrying a load of ℓ (kilograms) (Demir et al., 2014). ω1, ω2, ω3, and ω4 are vehicle-specific parameters whose descriptions are provided in Appendix A. 4. Mathematical model The MSWM system with cost minimization and emission control objectives is formulated as an MILP problem in this section. This biobjective facility location problem locates facilities in each level at candidate locations, quantifies MSW flows between the nodes of the network in different levels, and determines the required number of vehicle trips at the same time. Additionally, by selecting facilities among sets of candidate locations and determining the MSW flows, we explicitly determine travel speeds of vehicles. In order to determine where to locate the facilities, a binary variable is introduced for each candidate facility location. We denote binary variables using notations XTj, XPk, and XLl which take the value 1 if and only if TS j ∈ J1, P&T facility k ∈ K1, and sanitary landfill l ∈ L1 are opened, respectively. For the already existing facilities, the same notation is used and it is assumed that XTj = 1, XPk = 1, and XLl = 1 for each j ∈ J2, k ∈ K2, and l ∈ L2. The first set of constraints restricts the number of facilities that are going to be opened at candidate locations. If no such restrictions exist, then these constraints can be discarded:

XTj

j J1

5

TT

(2)

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

XPk

k K1

XLl

l L1

TP

amounts to FC of the vehicles coming back with no load through the same path i − j. ij1, ij2 , and ij3 are calculated using Eq. (1) as follows 3 ij

(3)

TL

(4)

1 ij

The second set of constraints are mass balance constraints presented as follows: j J

QMTij = Gi,

k K

l L

QTPjk =

i I

QPLkl =

k

i

I j

J

QTPjk,

k

QMTij

CTj XTj ,

QTPjk

CPk XPk ,

k

K

QPLkl

CLl XLl ,

l

L

j J

k K

j

2 ij

NTPjk

NPLkl

QMTij cap1

+

3 ij

QTPjk cap2

j

J

,

j

J,

k

K

QPLkl , cap3

k

K,

l

1 ij

+

2 ij

+

3 ij

(NMTij

1)cap1) DMTij +

1 3 DMTij

(VTijr )2LTijr )

LTijr

1 1(

= NMTij (

VTijr

1 3 DMTij

)+

1 4

+

(VTijr ) 2LTijr )

(17)

r Tij

where and are vehicle-specific parameters in Eq. (1) for vehicles of type 1. To obtain Eqs. (15)–(17), the first and last terms of Eq. (1) are computed for each road segment of path i − j and summed up as the travel speed of vehicles may not be constant over the segments of the path. In these equations, LTijr is the length of the r (r ∈ Tij) segment of path i − j and VTrij is feasible travel speed (i.e., satisfying the speed limits) of vehicle of type 1 corresponding to the lowest FC amount on the r segment of path i − j. If path i − j is not used by the vehicles, i.e., NMTij = 0 and QMTij = 0, then one can see that ij1, ij2 , and ij3 add up to zero. Note that, VTijr , which DMTij accounts for the distance of the shortest path connecting source i to TS j, also ij1, ij2 , and 3 ij are introduced to make the exposition simpler and they are indeed not part of the MILP model. The total FC of vehicles of type 2 transferring MSW between TS j and P&T facility k and total FC of vehicles of type 3 transferring MSW between P&T facility k and sanitary landfill l are denoted by fcTPjk and fcPLkl, respectively. The calculations of fcTPjk and fcPLkl are similar to that of fcMTij and omitted for brevity. The final set of constraints characterizes the non-negativity, integrality, and binary characteristics of decision variables, along with specifying the presence of the existing facilities: QMTij

1 2,

1 3,

0, QTPjk

1 4

0, QPLkl

0,

i

I,

j

J,

k

K,

l

L

(12)

(18) fcMTij

(13)

0, fcTPjk

0, fcPLkl

0,

i

I,

j

J,

k

K,

l

L

(19)

The number of required trips between MSW source i and TS j, TS j and P &T facility k, and P&T facility k and sanitary landfill l are denoted by NMTij, NTPjk, and NPLkl, respectively. Eqs. (11)–(13) assess the number of required trips considering the capacity of the vehicles performing the transportation services on the shortest paths connecting these facilities and the MSW quantity transferred. The objective function of the MILP model will make sure that NMTij, NTPjk, and NPLkl are equal to the ceiling of the right hand sides of Eqs. (11), (12), and (13), respectively. Next, we present constraints related to the calculation of FC of vehicles traversing a path joining two nodes of the MSWM network. First, we consider vehicles of type 1 which transfer MSW on the shortest paths connecting MSW sources to TSs. We assume that in all but one trip, vehicles travel with full capacity and all vehicles come back to source i from TS j following the same path with no load. Therefore, the total amount of FC by vehicles of type 1 over path i − j can be calculated as

fcMTij =

1 2 (QMTij

)+

r Tij

(11)

L

1 4

1 1,

(10)

I,

VTijr

(16)

(9)

i

(15)

r Tij

(8)

,

LTijr r Tij

Eqs. (8), (9), and (10) correspond to capacity constraints of TSs, P&T facilities, and sanitary landfills, respectively. The fourth set of constraints identifies the required number of trips to transport MSW on the shortest paths from MSW sources to TSs, TSs to P&T facilities, and P&T facilities to sanitary landfills, respectively, as follows:

NMTij

1 3 DMTij

+

(VTijr ) 2LTijr )

1 1(

=(

(7)

J

1 2 cap1DMTij

)+

r Tij

Eq. (5) implies that generated MSW at every source should be entirely served and transported to TSs. The MSW flows from source i to TS j, TS j to P&T facility k, and P&T facility k to sanitary landfill l are denoted by QMTij, QTPjk, and QPLkl, respectively. Eq. (6) balances the input and output MSW at every TS, and Eq. (7) specifies that the sum of the input MSW quantities of a P&T facility multiplied by its residual coefficient is equal to the sum of the output MSW quantities from that facility. The third set of constraints makes sure that MSW can only be transported to open facilities and also restricts the amount of entering MSW to each open facility by its capacity: i I

VTijr

r Tij

(6)

K

LTijr

1 1(

1)(

1 4

+

(5)

QMTij,

j J

= (NMTij

NMTij , NTPjk , NPLkl

+,

i

I,

j

J,

k

K,

l

L (20)

XTj , XPk , XLl

{0, 1},

XTj , XPk , XLl = 1,

j

j

J1,

J2 ,

k

k

K1,

K2,

l

l

(21)

L1

(22)

L2

Eqs. (18) and (19) specify that MSW flows and FC values are nonnegative. Eq. (20) indicates that the variables identifying the number of required trips have to be non-negative integers. Eq. (21) assigns 0 or 1 values to decision variables corresponding to selection of candidate locations for facilities. Finally, Eq. (22) specifies the presence of existing facilities in the system. Lastly, the two objective functions of the considered MILP model are presented in Eqs. (23) and (24), corresponding to total daily system cost and resulting CO2 emission from transportation activities, respectively:

Zcost =

(14)

XTj FCTj + j J

where ij1 corresponds to FC of vehicles traversing path i − j with full 2 capacity, accounts for FC of a vehicle transferring ij QMTij − (NMTij − 1)cap1 amount of remaining MSW on path i − j, and

VCTj j J

6

XPk FCPk + k K

QMTij + i I

VCPk k K

XLl FCLl +

(23a)

l L

QTPjk + j J

VCLl l L

QPLkl+ k K

(23b)

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

f(

fcMTij + i I j J

fcTPjk + j J k K

2(w1

fcPLkl )+

DMTij NMTij + w2 i I j J

DTPjk NTPjk j J k K

+ w3

DPLkl NPLkl )

(23d)

k K l L

Ck k K

share of 2.2 million tonnes. Furthermore, 26 municipal districts in Ankara received waste services producing an average of 1.14 kg municipal waste per capita per day in 2016. In order to apply our proposed model on the MSWM system of Ankara, we conduct an extensive investigation towards the system and its related components. ITC (Invest, Trading, & Consulting AG) is the company that is in charge of providing disposal and treatment services in Ankara and a number of cities in Turkey (ITC). We had an interview with a consultant of this company to get information regarding detailed services and operations of MSWM system of Ankara and its relative parameters. The employed parameters in this case study partially attribute to the actual operating system such as capacities of treatment and landfill facilities and MSW generation rates per capita per day, and are partially selected with respect to the literature such as fixed and variable costs of facilities (Tsilemou & Panagiotakopoulos, 2006) and vehicle related parameters. Ankara's MSWM system composes of two plants located in Sincan and Mamak municipal districts where they serve collected MSW from Ankara's municipalities on a daily basis. There are no hard rules regarding the assignment of collected MSW of certain municipalities to one of these facilities. A significant feature of this case study is that TSs do not operate in this system and the system comprises of MSW generation sources as well as Mamak and Sincan facilities. As a result, MSW is directly transferred from the generation sources to these two facilities for handling. It should be noted that various technologies such as separators, anaerobic digestors, compost plants, along with sanitary landfills operate in a single integrated mega-facility both in Mamak and Sincan. Thus, the third and fourth levels of the considered MSWM network should be combined to formulate the underlying MILP model of Ankara's MSWM system. In this case study, our aim is to locate the TSs to investigate the economic feasibility along with environmental benefits of the resulting system in comparison with the current one. Subsequently, the problem under study would have a network structured as in Fig. 2, with the paths represented by solid lines, which locates TSs to minimize total daily system cost and daily CO2 emission from transportation activities.

(23c)

k K l L

QTPjk

(23e)

j J

Zemission = c (

fcMTij + i I j J

fcTPjk + j J k K

fcPLkl ) k K l L

(24)

Eq. (23) refers to the total cost of the system on a daily basis. Daily fixed costs of open facilities are calculated by Eq. (23a). Eq. (23b) corresponds to variable costs of facilities calculated with respect to allocated MSW quantities to TSs, P&T facilities, and sanitary landfills, respectively. Eq. (23c) assesses fuel cost which is calculated by multiplying total FC of vehicles of types 1, 2, and 3 in L by unit fuel cost per liter, f. Eq. (23d) amounts to variable transportation cost calculated by multiplying the total travel distance of vehicles of type t by the unit transportation cost wt . Eq. (23e) calculates possible monetary benefits gained by processing and treatment of MSW in P&T facilities. Eq. (24) refers to the total amount of emitted CO2 from transportation activities which is calculated by multiplying the amount of FC in L by the CO2 emissions coefficient, c, in kg/L (Franceschetti et al., 2013). In conclusion, the objective function of the proposed MILP model is as follows: (25)

Minimize(Zcost, Zemission )

The conceptual framework of the MSWM network under study is shown in Fig. 1 where different levels of the network along with existing and candidate facilities with the distances and road segments connecting them are depicted in addition to the MSW flows among the facilities. The decision variables and the sets and parameters of the proposed MILP model are provided in Tables Table B1 and Table B2, respectively, in Appendix B. 5. Municipal solid waste management system at Ankara

5.1. Network construction

We apply the proposed model on the MSWM system operating in Ankara to investigate resulting sustainability benefits. Ankara is the capital and the second most populated city of Turkey with a total population of 5,346,518 in 2016 (TURKSTAT). According to municipal waste statistics released by the TURKSTAT in November 2017; of the 31.6 million tonnes of MSW collected in Turkey in 2016, Ankara has a

Geographical Information Systems have enabled decision makers to handle and manipulate spatial data in a more exact manner. In this study, the physical distances between MSW sources, existing facilities, and candidate ones are measured by utilizing the Network Analyst extension of ESRI's ArcGIS 10 software by introducing the road network of Ankara as a shapefile to ArcGIS. The daily MSW generation rates are

Fig. 1. The conceptual framework of the considered MSWM network. 7

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Fig. 2. The schematic diagram of the MSWM system under study.

calculated with respect to the population of the municipal districts of Ankara in 2016, released by (TURKSTAT), multiplied by respective 2016's MSW generation rate per capita per day. Additionally, for each municipal district having a daily MSW generation rate less than 250 tonnes, the centroid of the district obtained by ArcGIS is used as the MSW generation source. Districts with more than 250 tonnes of daily MSW generation rates are divided into a number of sub-districts possessing equal MSW generation rates that are less than the specified threshold. For each sub-district, its centroid is again taken as the MSW generation source. There are usually more than one MSW generation source in highly populated municipal districts and by dividing such districts into a number of sub-districts, we aim to obtain a relatively realistic MSWM network for Ankara. Sixty candidate locations for TSs are introduced as a shapefile in ArcGIS for the MSWM system of Ankara. We generate the daily fixed and variable costs (€) of each candidate TS uniformly at random from [250, 400] and [1, 5] intervals, respectively. Figs. 3 and 4 represent the considered MSW transportation network including the spatial distribution of the nodes of the network along with the shortest paths connecting them. Two vehicle types are considered, namely, a refuse compactor with 18 m3 capacity and a dump truck with 40 tonne capacity functioning as collection vehicles and high-volume transfer vehicles, respectively (HLTQ, 2018). There are a number of parameters with common values for both vehicle types (Demir, Bektaş, & Laporte, 2011; Koç, Bektaş, Jabali, & Laporte, 2016), see Table 2, and a few parameters with different values for each vehicle type taken from (HLTQ, 2018), see Table 3. Typically, the FC rate per unit distance as a function of speed as estimated by CMEM is a convex U-shaped curve (Bektaş & Laporte, 2011). The optimal speed Vopt of

a vehicle minimizing the FC rate is independent of the load carried and is equal to (ω1/2ω4)1/3 (Khoei, Süral, & Tural, 2017). Fig. 5 displays the FC rate as a function of speed plotted for each vehicle type carrying no load with respect to the parameters specified in Tables 2 and 3 . Increasing the load of the vehicles shifts the FC rate curve upwards. We assume that angel of every road segment is equal to zero (Φ = 0). As can be noticed from Fig. 5, transporting a specific amount of MSW through the same distance with the same speed using refuse compactors results in a slightly higher FC than dump trucks. The difference in FC of refuse compactors and dump trucks would be more noticeable when a single dump truck carries a certain amount of MSW that can alternatively be carried by multiple refuse compactors. For instance, in order to transport 24 tonnes of MSW for 100 km with a speed of 60 km/h, 3 refuse compactors are needed consuming around 177 L of diesel fuel in total. However, a single dump truck can transport the same amount of MSW over the same distance with the same speed by consuming 79 L of diesel fuel. Hence, utilizing large volume dump trucks could be beneficial for MSWM systems in terms of cost and emissions savings. An input of our model is the speed limits for each road segment. These speed limits should be specified not only based on legal speed limits imposed by legislative bodies, but also considering the impact of average traffic congestion over each road segment. For instance, in a highly congested road segment a legal upper speed limit of 100 km/h would never be achievable. We assume in our model that the speed limits, given as input, are determined with respect to the average congestion in addition to the legal speed limits and any speed value between the given speed limits is achievable by the vehicles. In our case study, as we do not have any information regarding congestion over the road network of Ankara, we define

Fig. 3. Spatial distribution of MSW sources and candidate TSs along with the shortest paths connecting them. 8

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Fig. 4. Spatial distribution of candidate TSs and treatment and landfill facilities along with the shortest paths connecting them.

6. Computational experiments and analyses

Table 2 Parameters common to all vehicle types (from Demir et al. (2011) and Koç et al. (2016)). Notation

Description

Value

ξ κ ψ ϵ π ρ e g Φ Cr Cd

Fuel-to-air mass ratio Heating value of a typical diesel fuel (kJ/g) Conversion factor (g/s to L/s) Vehicle drive train efficiency Efficiency parameter for diesel engines Air density (kg/m3) Engine friction factor (kJ/rev/L) Gravitational constant (m/s2) Road angel Coefficient of rolling resistance Coefficient of aerodynamic drag

1 44 737 0.45 0.45 1.2041 0.2 9.81 0 0.01 0.7

In the current MSWM system of Ankara, MSW collection is performed by municipalities using a specific type of refuse compactors and MSW is then sent directly to the treatment and landfill facilities by these vehicles. Presence of TSs as consolidation points proved to be cost-effective and environmentally beneficial for MSWM systems, as multiple MSW collection vehicles with limited capacities replace with large-volume transfer vehicles to transport MSW to distant facilities (USEPA, 2002). Therefore, we investigate the impact of introducing TSs into Ankara's MSWM system by applying the proposed bi-objective MILP model. We further investigate the influence of speed variations of vehicles on the resulting CO2 emission by conducting simulation analyses. The MILP models in this section are solved using IBM ILOG CPLEX Optimization Studio (12.8.0) (IBM) on a server with Intel(R) Core(TM) i7-4770S CPU 3.10 GHz (8 CPUs) and 16,384 MB RAM with operating system Windows 10. The simulation analyses are carried out using Matlab 2018RA on the same server. The FC of vehicles and the emitted CO2 are calculated by considering the best travel speed configuration of the vehicles resulting in the lowest emission on each road segment. For the refuse compactors and dump trucks considered in this study, the optimal speed values 1 2 44 km/h and Vopt 46 km/h, minimizing the FC rate are found as Vopt respectively. Travel speeds of vehicles are determined by considering the U-shaped FC function along with the lower and upper speed limits of each zone. If the optimal speed of a vehicle Vopt satisfies the speed limits, then the travel speed of a vehicle would be Vopt. Otherwise, the travel speed of a vehicle is taken as the feasible speed (i.e., speed value satisfying the speed limits) that is closest to Vopt. Thus, the travel speeds of refuse compactors and dump trucks are chosen according to Table 4. In our computational experiments, emission values are calculated considering CO2 emission coefficient c as 2.67 kg/L of diesel consumed by vehicles (USEPA, 2005). The unit fuel cost f is set to 1.01 €/L according to (GPP, 2018) and unit variable transportation cost coefficient wt is assumed to be 2 € per kilometer for each vehicle type t. The existing MSWM system in Ankara (operating without TSs) has been formulated with respect to the proposed MILP model, and optimized in terms of CO2 emission minimization objective which is solved within 0.08 s. The minimum amount of CO2 emitted by refuse compactors in this system is

Table 3 Vehicle-specific parameters (from Hubei Liwei Automobile Co. Ltd (2018)). Notation

Description

Refuse compactor

Dump truck

N V A μ

Engine Speed (rev/s) Engine displacement (L) Frontal surface area (m2) Curb weight (kg)

41.6 6.7 7.35 15,880

36.6 8.9 7.53 12,005

the speed limits as a function of population density of each municipal district. For this purpose, we divide Ankara into three zones with respect to the population density of its municipal districts defined as sparsely, medium, and densely populated zones. Correspondingly, lower and upper speed limits are assigned to each zone where they increase from densely to sparsely populated districts. In Fig. 6, categorization of the speed zones is provided. Lower and upper speed limits are assumed to be 20, 30, 40 km/h and 30, 55, 70 km/h for densely, medium, and sparsely populated zones, respectively. As locating TSs is a strategic level decision, we ignore the impact of traffic congestion over different time periods on the speed limits. Once the locations of TSs are determined by implementing the proposed biobjective MILP model, one can consider the scheduling problem in the MSWM system by taking the congestion in different time periods into account as a post hoc analysis. 9

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Fig. 5. Fuel consumption per 100 km as a function of vehicle speed.

Fig. 6. Speed zones.

calculated as 65,048 kg/day which is the result of direct shipment of MSW from sources to treatment and landfill facilities. Corresponding to this solution, the total daily cost of the current system is 63,942 €. This cost comprises of fuel cost which is 24,606 €, fixed and variables costs of the facilities which is 37,029 €, and variable transportation cost which is 2,307 €. Optimizing the current system in terms of the cost minimization

objective results in almost the same amounts of total daily system cost, 63,937 €, and CO2 emission, 65,052 kg/day, as no decision is made regarding establishing new facilities. First, we aim to extend the current system by introducing TSs to minimize the total daily system cost. We use the model developed in Section 4 for this purpose. In this case, the model locates 9 TSs out of 60

Table 4 Travel speeds of vehicles on each road segment. Road segment

Lower–upper speed limits (km/h)

Travel speed (km/h) 1 Refuse compactor (Vopt

In densely populated zones In medium populated zones In sparsely populated zones

20–30 30–55 40–70

30 44 44

10

44 km/h)

2 Dump truck (Vopt

30 46 46

46 km/h)

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

candidate locations (see Fig. 9a) with a total daily system cost of 65,131 €. 17,445 € of this value amounts to fuel cost, 46,192 € to fixed and variable costs of the facilities, and 1,494 € to variable transportation cost. Introducing these 9 TSs increases the total daily cost of the current system by 1.86%, and results in a 29.10% reduction in the amount of emitted CO2 in the extended system. Thus, a relatively small increase in the total daily system cost results in a considerable reduction in the daily CO2 emission from transportation activities. Note that, in this case, the MILP model is solved within 4.59 s. The two objectives, namely, total daily MSWM system cost and daily CO2 emission from transportation of MSW, of the proposed MILP model are in conflict with each other as increasing the system cost results in locating more TSs which brings more flexibility in transportation services and less CO2 emission. As a matter of fact, it is not possible to find a solution which optimizes these two objectives simultaneously and Pareto efficient solutions are needed to be generated instead. The ϵconstraint method is employed to generate the Pareto front (Haimes, Lasdon, & Wismer, 1971). The method solves a sequence of single-objective optimization problems by transforming the other objective into a constraint and changing its upper bound progressively. For more details on the ϵ-constraint method in bi-objective optimization problems, the reader is referred to Bérubé, Gendreau, and Potvin (2009). We transform cost minimization objective into a constraint and change its upper bound, considered as a budget, progressively to achieve the Pareto front. Fig. 7 represents the set of Pareto efficient solutions as points on the Pareto front which are obtained by implementing the ϵconstraint method for the extended system operating with TSs. As can be seen from Fig. 7, increasing the budget (total daily system cost), upon the request from the responsible organization, from the budget of the minimum cost solution by a relatively small amount results in a considerable reduction in the daily CO2 emission. As a result, we investigate the impact of increasing the budget from its minimum value step by step to investigate the gained environmental benefits in terms of CO2 emission reduction. Table 5 compares different budget values and the resulting CO2 emission amounts, along with respective percent changes from two settings, namely, the current system operating without TSs (1st setting) and the extended system optimized with respect to cost minimization objective (2nd setting). If the budget of the current system is increased by 4.78%, then with this budget 14 TSs are located out of 60 candidate locations in the extended system when the daily CO2 emission objective is minimized. This results in a 38.39% decrease in daily CO2 emission from transportation activities. In the extended system, when we compare the solution obtained by the minimization of the total daily system cost objective and the solution corresponding to the 67,000 € budget, a 2.87% increase in the budget leads to a 13.10% decrease in CO2 emission from

Table 5 Comparison of different settings in terms of the considered objectives. Setting

1st settinga 2nd settingb 3rd settingc 4th settingd 5th settinge 6th settingf a b c d e f

Objective

% change from the 1st setting

% change from the 2nd setting

Zcost (€)

Zemission (kg)

ΔZcost (€)

ΔZemission (kg)

ΔZcost (€)

ΔZemission (kg)

63,942 65,131 66,000 67,000 68,000 69,000

65,048 46,117 41,666 40,075 38,850 38,056

– 1.86 3.22 4.78 6.35 7.91

– −29.10 −35.95 −38.39 −40.27 −41.49

– – 1.33 2.87 4.41 5.94

– – −9.65 −13.10 −15.76 −17.48

Current system. Extended system with minimum Zcost. Extended system with minimum Zemission and budget ≤ 66,000. Extended system with minimum Zemission and budget ≤ 67,000. Extended system with minimum Zemission and budget ≤ 68,000. Extended system with minimum Zemission and budget ≤ 69,000.

transportation activities. This reduction in CO2 emission is achieved by opening 5 more TSs (see Fig. 9b). As mentioned before, in the current MSWM system, there are no TSs and the MSW is directly sent from MSW sources to treatment and landfill facilities. We have so far investigated the effect of using TSs as transshipment facilities. We now consider a hybrid system in which there are candidate TSs to be located while MSW is also allowed to be sent directly from MSW sources to treatment and landfill facilities via a new set of shortest paths introduced between them. The schematic diagram of the hybrid MSWM network is depicted with both solid and dashed lines in Fig. 2, with dashed lines representing the newly introduced paths. It is expected that, inclusion of these new set of paths would result in lower CO2 emission from transportation of MSW because it brings more flexibility in path selection for refuse compactors. As a result, the generated MSW at some sources that are comparatively closer to treatment and landfill facilities can be partially or totally transferred directly to the treatment and landfill facilities via the newly introduced shortest paths. In order to formulate the underlying MILP model of this hybrid MSWM system, we need to introduce new sets of parameters and decision variables, perform some modifications in the objective functions, add new sets of constraints, and modify the MSW generation balance constraints in the MILP formulation of Section 4. The details regarding new MILP formulation is omitted here, as it is analogous to the one given in Section 4. In order to compare the results for the extended and hybrid systems, Pareto front of the hybrid model is generated. The set of Pareto efficient solutions on the Pareto front of the hybrid model is shown in Fig. 8

Fig. 7. Pareto front of the extended system operating with TSs. 11

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Fig. 8. The extended vs hybrid models’ Pareto fronts.

along with the Pareto front of the extended model. As can be seen, the hybrid model's Pareto front falls below the extended one which supports the claim that for every fixed budget value, the hybrid system results in lower daily CO2 emission from transportation activities. Fixing the budget to 67,000 €, the resulting daily CO2 emission is 40,075 and 36,132 kg in the extended and hybrid models, respectively. This 9.84% decrease in CO2 emission is due to locating 22 TSs in the hybrid model in comparison to locating 14 TSs in the extended one. In the extended model, 15,160 € is allocated to fuel cost, 50,570 € to fixed and variable costs of the facilities, and 1,270 € to variable transportation cost. On the other hand, in the hybrid system, the fuel cost accounts for 13,668 €, fixed and variable costs of the facilities account for 52,205 €, and variable transportation cost amounts to 1,127 €. In the hybrid model, 17.56% of the MSW is transported directly from sources to treatment and landfill facilities. In Fig. 9, one can see all candidate TSs and compare the locations of the selected ones in the following settings:

travel speed of a vehicle depends on numerous factors such as drivers’ behavior, road conditions, and weather conditions. In this section, we want to investigate the sensitivity of the resulting CO2 emission from transportation activities and the total daily system cost in MSWM system of Ankara to speed variations. For this purpose, we consider the current system along with four network configurations designed by the optimization model proposed in this paper, where the locations of the MSWM facilities including the TSs along with the MSW flows and the number of vehicle trips between facilities are fixed. Monte Carlo simulation experiments are carried out considering vehicle speed on each road segment as a random variable following a particular probability distribution. Based on the realized vehicle speed values, we assess the total daily system cost and CO2 emission. The four network configurations considered in addition to the current system (1st setting) are the extended system with minimum CO2 total daily system cost (2nd setting), the extended system with minimum emission and 67,000 € budget (4th setting), the hybrid system with minimum total daily system cost (7th setting), and the hybrid system with minimum CO2 emission and 67,000 € budget (8th setting). We consider two types of probability distributions, namely, the uniform and triangular distributions, for vehicle speeds on each road segment. In the first set of simulation experiments, we assume that the travel speed of each vehicle traversing through densely, medium, and sparsely populated zones follow U(20, 30), U(30, 55), and U (40, 70), respectively, where U(a, b) represents the continuous uniform distribution with parameters a and b. For the second set of experiments, it is assumed that the travel speed of each vehicle traversing through densely, medium, and sparsely populated zones follow T(20, 25, 30), T(30, 40, 55), and T(40, 55, 70), respectively, where T(a, b, c) represents the triangular distribution with parameters a, b, and c. For each of the network configurations and for each of the probability distributions, 1000 replications are done and the realized minimum, average, and the maximum total daily system cost and CO2 emission values along with the margin of error for a 95% confidence interval for the mean total daily system cost and the mean CO2 emission are reported in Table 7. Moreover, for each setting, the respective percent changes from the values reported in Table 6 (i.e., the values obtained by the best speed patterns) are displayed in Table 8. Looking at Table 7, we can see that the mean CO2 emission amounts for the considered four network configurations (2nd, 4th, 7th, and 8th settings) with random speeds are significantly smaller than the CO2 emission amount in the current system under the best speed patterns (65,048 kg) at the 5% significance level. Moreover, the mean total daily system cost for the 7th setting with random speeds is significantly smaller than the total daily system cost of the current system under the best speed patterns (63,942 €) at the 5% significance level. Therefore, it can be concluded that the 7th setting improves the current system even when the vehicles do not travel according to the best speed patterns. As can be seen from Table 8, if the vehicles do not follow the best

• 2 setting: Extended system with minimum total daily system cost (Fig. 9a). • 4 setting: Extended system with minimum daily CO and 67,000 € budget (Fig. 9b). • 7 setting: Hybrid system with minimum total daily system cost (Fig. 9c). • 8 setting: Hybrid system with minimum daily CO and 67,000 € nd th

2

th th

budget (Fig. 9d).

2

In summary, if we consider the current system (1st setting) along with the mentioned settings above, the respective allocation of total daily system cost into fuel cost (Zfuel), fixed and variable costs of the facilities (Zfac), and variable transportation cost (Ztrans) are reported in c Table 6. Note that Zemission = f Zfuel . Additionally, the percent changes of the reported cost values from the values of the current system are included in Table 6. It is worth mentioning that the hybrid system with minimum total daily system cost (7th setting) improves the current system (1st setting) in terms of both total daily system cost and CO2 emission objective s. In this setting, the cost saving in transportation activities, resulted by inclusion of TSs, is large enough to be allocated for opening TSs without imposing any additional costs into the system. 6.1. Simulation analyses As was mentioned before, the vehicle speed is one of the factors affecting the FC of a particular vehicle. The CO2 emission in previous analyses was calculated with respect to optimal speed of vehicles in terms of FC as well as lower and upper speed limits of the zones according to Table 4. The 12

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Fig. 9. Selected TSs.

speed patterns, but instead drive with speeds following the considered distributions, there will be less than 2.4% increase in total daily system cost and less than 6.1% increase in the emitted CO2 amount for all five settings considered. Even though the CO2 emission amounts increase when vehicles travel with random speeds, the resulting mean CO2 emission amounts in all settings except the current one are still significantly less than the amount emitted in the current system under both best and random speed patterns.

7. Conclusion and future research directions The objective of this paper was to investigate the impact of CO2 emission from transportation activities in locational planning of MSWM systems. A bi-objective MILP optimization model is developed to manage the system economically and environmentally by minimizing the total daily system cost and CO2 emission of vehicles, respectively. The ϵ-constraint method was implemented to generate the efficient

Table 6 Comparison of different settings in terms of cost categories. Setting

st

1 setting 2nd setting 4th setting 7th settingb 8th settingc a b c

% changes from the 1st setting (€)

Daily system cost categories (€) Zemission

Zcost

Zfuel

Zfac

Ztrans

ΔZcost

ΔZfuela

ΔZfac

ΔZtrans

65,048 46,117 40,075 51,052 36,132

63,942 65,131 67,000 60,442 67,000

24,606 17,445 15,160 19,312 13,668

37,029 46,192 50,570 39,400 52,205

2,307 1,494 1,270 1,730 1,127

– 1.86 4.78 −5.47 4.78

– −29.10 −38.39 −21.52 −44.45

– 24.75 36.57 6.40 40.98

– −35.25 −44.94 −24.99 −51.13

ΔZfuel = ΔZemission. Hybrid system with minimum Zcost. Hybrid system with minimum Zemission and budget ≤ 67,000. 13

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Table 7 The realized minimum, average, and maximum total daily system cost and CO2 emission values in 1000 replications along with the margin of errors. Setting

Distribution

1st setting

Triangular Uniform Triangular Uniform Triangular Uniform Triangular Uniform Triangular Uniform

2nd setting 4th setting 7th setting 8th setting a

Total daily system cost (€)

CO2 emission (kg)

Minimum

Average

Maximum

MOE

65,175 65,235 65,938 65,985 67,700 67,732 61,362 61,408 67,623 67,666

65,239 65,335 65,977 66,042 67,736 67,792 61,415a 61,488a 67,660 67,709

65,312 65,423 66,024 66,113 67,780 67,841 61,476 61,553 67,700 67,762

1.25 1.83 0.82 1.21 0.76 1.09 0.90 1.33 0.74 1.03

b

Minimum

Average

Maximum

MOEc

68,308 68,466 48,251 48,376 41,927 42,012 53,484 53,606 37,779 37,893

68,476 68,731 48,353a 48,526a 42,021a 42,170a 53,624a 53,815a 37,876a 38,007a

68,670 68,964 48,478 48,714 42,138 42,298 53,784 53,990 37,983 38,147

3.31 4.83 2.18 3.19 2.01 2.89 2.38 3.51 1.95 2.71

Mean values are significantly smaller than the corresponding values in the current system under the best speed pattern at the 5% significance level. MOE: margin of error for a 95% confidence interval for the mean total daily system cost. MOE: margin of error for a 95% confidence interval for the mean CO2 emission.

b c

Table 8 Percent changes of minimum, maximum, and average total daily system cost and CO2 emission values obtained in Monte Carlo simulation from total daily system cost and CO2 emission of best-speed schemes. Setting

Distribution

st

1 setting nd

2

% change of total daily system cost (€)

Triangular Uniform Triangular Uniform Triangular Uniform Triangular Uniform Triangular Uniform

setting

4th setting 7th setting 8th setting

% change of CO2 emission (kg)

Minimum

Average

Maximum

Minimum

Average

Maximum

1.93 2.02 1.24 1.31 1.05 1.09 1.52 1.60 0.93 0.99

2.03 2.18 1.30 1.40 1.10 1.18 1.61 1.73 0.98 1.06

2.14 2.32 1.37 1.51 1.16 1.26 1.71 1.84 1.05 1.14

5.01 5.26 4.63 4.90 4.62 4.83 4.76 5.00 4.56 4.88

5.27 5.66 4.85 5.22 4.85 5.23 5.04 5.41 4.83 5.19

5.57 6.02 5.12 5.63 5.15 5.55 5.35 5.75 5.12 5.58

solutions of the proposed model. This model was applied on MSWM system of Ankara to investigate the economic and environmental benefits obtained by introducing TSs into the current system. Two extensions of the current MSWM system have been considered, namely, the extended system in which no direct shipment from MSW sources to treatment and landfill facilities is allowed and the hybrid system where direct shipments are allowed in addition to indirect shipments through TSs. As the hybrid system brings more flexibility into the current system in terms of transportation services, the solutions found in the hybrid system dominate those found in the extended one. However, even for the extended system, little increase in the budget results in a considerable reduction in the CO2 emission. The results of the simulation analyses indicate that for both extensions of the current system, once the locations of the facilities and the allocations of MSW flows are fixed, the resulting CO2 emission from transportation activities and the total daily system cost values are not subject to a considerable change due to speed variations. Moreover, it is observed that the 7th setting improves the current one in terms of both

objective functions even when the vehicles do not follow the best speed patterns. The proposed MSWM model can be utilized as a decision support tool by regional authorities and city logistics planners to have effective and efficient management of MSW services. This model enables decision makers to investigate MSWM system thoroughly by selecting the locations of MSW facilities, assessing the performances of different types of vehicles in the system, managing travel speeds of vehicles, and investigating the impact of imposed speed limits throughout the transportation network. As a possible future research direction, one can consider stochastic MSW generation rates, their variations through time, and their subsequent impact on capacity allocation decisions of MSWM facilities. Moreover, it is assumed that the refuse compactors start their trips from the centroids of the considered municipal districts and sub-districts. Consideration of MSW collection operations, associated costs and emissions, and the integration of the collection system with the regional planning of MSWM system could be an interesting research direction for future.

Appendix A. CMEM According to CMEM, the FC of a heavy-duty vehicle can be calculated as

F = (eNV

z + v

z +

µz +

v 2z )

(A1)

where F is the amount of fuel consumed in liter (L) by a heavy-duty vehicle traversing a distance of z meters (m) with a constant speed of v meters/ second (m/s) carrying a load of ℓ kilograms (kg). Remaining parameters are vehicle and road specific parameters with descriptions provided in Table A1. λ = ξ/κψ, γ = 1/1000ϵπ, β = 0.5CdAρ are vehicle-specific parameters while α = g sin(Φ) + gCr cos(Φ) is vehicle-road-specific parameter which depends the on road angel. For a detailed study of the parameters used in CMEM and comparison of different vehicular emission models, the reader is referred to Demir et al. (2011). Defining ω1 = λeNV, ω2 = λγα, ω3 = λγαμ, and ω4 = λβγ, Eq. (A1) can be rewritten as

14

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

Table A1 FC function parameters.

F=

1

z + v

2z

+

3z

+

4v

Notation

Description

ξ κ ψ ϵ π ρ e g Φ Cr Cd N V A μ

Fuel-to-air mass ratio Heating value of a typical diesel fuel (kJ/g) Conversion factor (g/s to L/s) Vehicle drive train efficiency Efficiency parameter for diesel engines Air density (kg/m3) Engine friction factor (kJ/rev/L) Gravitational constant (m/s2) Road angel Coefficient of rolling resistance Coefficient of aerodynamic drag Engine Speed (rev/s) Engine displacement (L) Frontal surface area (m2) Curb weight (kg)

2z

(A2)

Appendix B. Sets, parameters, and decision variables

Table B1 Decision variables. Notation

Description

QMTij, QTPjk, QPLkl NMTij, NTPjk, NPLkl XTj XPk XLl fcMTij, fcTPjk, fcPLkl

Quantity of MSW to be transfered from i to j, j to k, and k to l (tonne/day) Number of required trips to transfer quantified daily MSW from i to j, j to k, and k to l Binary variable determining presence/absence of a TS at candidate location j Binary variable determining presence/absence of a P&T facility at candidate location k Binary variable determining presence/absence of a sanitary landfill at candidate location l FC on the shortest paths from i to j, j to k, and k to l (L)

Table B2 Sets and parameters. Notation

Description

i, j, k, l I J1, K1, L1 J2, K2, L2 J, K, L t Tij, Pjk, Lkl r Gi CTj, CPk, CLl FCTj, FCPk, FCLl VCTj, VCPk, VCLl TT, TP, TL σk Ck DMTij, DTPjk, DPLkl LTijr , LP rjk , LLklr

Indices of MSW generation sources, TSs, P&T facilities, and sanitary landfills Set of MSW generation sources Sets of candidate TSs, P&T facilities, and sanitary landfills Sets of existing TSs, P&T facilities, and sanitary landfills Sets of TSs, P&T facilities, and sanitary landfills, i.e., J = J1 ∪ J2, K = K1 ∪ K2, and L = L1 ∪ L2 Index of vehicle types, i.e., t ∈ {1, 2, 3} Sets of road segments on the shortest paths connecting i to j, j to k, and k to l Index of road segments MSW generation rate of source i (tonne/day) Capacities of facilities i, j, and k (tonne/day) Fixed costs of facilities j, k, and l (€/day) Unit variable costs of facilities j, k, and l (€/tonne/day) Maximum number of TSs, P&T facilities, and sanitary landfills to be built in the MSWM system Residual coefficient of P&T facility k Unit monetary benefit coefficient of P&T facility k (€/tonne) Distances of the shortest paths between i and j, j and k, k and l (km) Lengths of the rth segment of the shortest path from i to j, j to k, and k to l (km)

c f wt capt LSLTijr , LSLP rjk , LSLLklr

CO2 emission coefficient: the amount of CO2 emission in kg per liter of fuel consumed (kg/L) Unit fuel (diesel) cost (€/L) Unit variable transportation cost for a vehicle of type t (€/km) Capacity of vehicle of type t (tonne) Lower speed limits of the r (r ∈ Tij) segment of the shortest paths connecting i to j, j to k, and k to l (km/h)

r , USLL r USLTijr , USLP jk kl

Upper speed limits of the r (r ∈ Tij) segment of the shortest paths connecting i to j, j to k, and k to l (km/h)

t Vopt

Travel speed of vehicle of type 2 on the r (r ∈ Pij) segment of the shortest path from j to k (km/h)

VP rjk ,

Travel speed of vehicle of type 3 on the r (r ∈ Lij) segment of the shortest path from k to l (km/h)

VLklr , t 1

t 2,

Optimal speed of vehicle of type t (km/h) Travel speed of vehicle of type 1 on the r (r ∈ Tij) segment of the shortest path from i to j (km/h)

VTijr ,

t 3,

t 4

Vehicle-specific parameters for vehicles of type t

15

Sustainable Cities and Society 52 (2020) 101807

M. Mohsenizadeh, et al.

References

Hubei Liwei Automobile Co. Ltd. http://www.hltq.com/ (Online) Accessed 09.10.18. IBM ILOG Cplex Optimization Studio (12.6) (2019). https://www.ibm.com/analytics/ cplex-optimizer (Online) Accessed 01.01.19. ITC Integrated Solid Waste Management Systems (2018). http://www.itcturkiye.com/en (Online) Accessed 24.10.18. Khoei, A. A., Süral, H., & Tural, M. K. (2017). Time-dependent green Weber problem. Computers & Operations Research, 88, 316–323. https://doi.org/10.1016/j.cor.2017.04. 010. Kirca, Ö., & Erkip, N. (1988). Selecting transfer station locations for large solid waste systems. European Journal of Operational Research, 35, 339–349. https://doi.org/10. 1016/0377-2217(88)90224-X. Knörr, W., Seum, S., Schmied, M., Kutzner, F., & Anthes, R. (2011). Ecological transport information tool for worldwide transports: Methodology and data update. Technical Report. Koç, Ç., Bektaş, T., Jabali, O., & Laporte, G. (2016). The impact of depot location, fleet composition and routing on emissions in city logistics. Transportation Research Part B: Methodological, 84, 81–102. https://doi.org/10.1016/j.trb.2015.12.010. Kulcar, T. (1996). Optimizing solid waste collection in Brussels. European Journal of Operational Research, 90, 71–77. https://doi.org/10.1016/0377-2217(94)00311-4. Li, Y., & Huang, G. (2010). An interval-based possibilistic programming method for waste management with cost minimization and environmental-impact abatement under uncertainty. Science of The Total Environment, 408, 4296–4308. https://doi.org/10. 1016/j.scitotenv.2010.05.038. Mirdar Harijani, A., Mansour, S., & Karimi, B. (2017a]). A multi-objective model for sustainable recycling of municipal solid waste. Waste Management & Research, 35, 387–399. https://doi.org/10.1177/0734242X17693685. Mirdar Harijani, A., Mansour, S., & Karimi, B. (2017b]). Multi-period sustainable and integrated recycling network for municipal solid waste – A case study in Tehran. Journal of Cleaner Production, 151, 96–108. https://doi.org/10.1016/j.jclepro.2017.03.030. Mohammadi, M., Jämsä-Jounela, S. L., & Harjunkoski, I. (2019). Optimal planning of municipal solid waste management systems in an integrated supply chain network. Computers & Chemical Engineering, 123, 155–169. https://doi.org/10.1016/j. compchemeng.2018.12.022. Rathore, P., & Sarmah, S. (2019). Modeling transfer station locations considering source separation of solid waste in urban centers: A case study of Bilaspur city, India. Journal of Cleaner Production, 211, 44–60. https://doi.org/10.1016/j.jclepro.2018.11.100. Santibañez-Aguilar, J. E., Ponce-Ortega, J. M., González-Campos, J. B., Serna-González, M., & El-Halwagi, M. M. (2013). Optimal planning for the sustainable utilization of municipal solid waste. Waste Management, 33, 2607–2622. https://doi.org/10.1016/ j.wasman.2013.08.010. Tan, S. T., Lee, C. T., Hashim, H., Ho, W. S., & Lim, J. S. (2014). Optimal process network for municipal solid waste management in Iskandar Malaysia. Journal of Cleaner Production, 71, 48–58. https://doi.org/10.1016/j.jclepro.2013.12.005 Special Volume: PSE Asia for Cleaner Production. Tavares, G., Zsigraiova, Z., Semiao, V., & Carvalho, M. (2009). Optimisation of MSW collection routes for minimum fuel consumption using 3D GIS modelling. Waste Management, 29, 1176–1185. https://doi.org/10.1016/j.wasman.2008.07.013. Tsilemou, K., & Panagiotakopoulos, D. (2006). Approximate cost functions for solid waste treatment facilities. Waste Management & Research, 24, 310–322. https://doi.org/10. 1177/0734242X06066343 PMID:16941990. Turkish Statistical Institute (2019). http://www.turkstat.gov.tr/Start.do (Online) Accessed 22.02.19. United States Environmental Protection Agency (2002). Waste transfer stations: A manual for decision-making. Technical Reporthttps://www.epa.gov/sites/production/files/ 2016-03/documents/r02002.pdf. United States Environmental Protection Agency (2005). Average carbon dioxide emissions resulting from gasoline and diesel fuel. Technical Reporthttps://nepis.epa.gov/Exe/ ZyPDF.cgi/P1001YTF.PDF?Dockey=P1001YTF.PDF. United States Environmental Protection Agency (2015). Advancing sustainable materials management: 2014 fact sheet. Technical Reporthttps://www.epa.gov/sites/production/ files/2016-11/documents/2014_smmfactsheet_508.pdf. United States Environmental Protection Agency (2016). Sources of greenhouse gas emissions. (Online) Accessed 14.03.19 https://www.epa.gov/ghgemissions/ sourcesgreenhouse-gas-emissions. World Commission on Environment and Development (1987). Our common future. (Online) Accessed 30.05.19 http://www.ask-force.org/web/Sustainability/ Brundtland-Our-Common-Future-1987-2008.pdf. Yadav, V., Bhurjee, A., Karmakar, S., & Dikshit, A. (2017). A facility location model for municipal solid waste management system under uncertain environment. Science of The Total Environment, 603-604, 760–771. https://doi.org/10.1016/j.scitotenv.2017.02.207. Yadav, V., Karmakar, S., Dikshit, A., & Bhurjee, A. (2018). Interval-valued facility location model: An appraisal of municipal solid waste management system. Journal of Cleaner Production, 171, 250–263. https://doi.org/10.1016/j.jclepro.2017.09.233. Yadav, V., Karmakar, S., Dikshit, A., & Vanjari, S. (2016). A feasibility study for the locations of waste transfer stations in urban centers: A case study on the city of Nashik, India. Journal of Cleaner Production, 126, 191–205. https://doi.org/10.1016/ j.jclepro.2016.03.017. Yu, H., Solvang, W., Yuan, S., & Yang, Y. (2015). A decision aided system for sustainable waste management. Intelligent Decision Technologies, 9, 29–40. https://doi.org/10. 3233/IDT-140203. Yu, H., & Solvang, W. D. (2017). A multi-objective location-allocation optimization for sustainable management of municipal solid waste. Environment Systems and Decisions, 37, 289–308. https://doi.org/10.1007/s10669-017-9632-y.

Anderson, L., & Nigam, A. (1968). A mathematical model for the optimization of a waste management system. Technical Report. Sanitary Engineering Research Laboratory, SERL Report. Antunes, A. P. (1999). Location analysis helps manage solid waste in central Portugal. INFORMS Journal on Applied Analytics, 29, 32–43. https://doi.org/10.1287/inte.29.4.32. Asefi, H., & Lim, S. (2017). A novel multi-dimensional modeling approach to integrated municipal solid waste management. Journal of Cleaner Production, 166, 1131–1143. https://doi.org/10.1016/j.jclepro.2017.08.061. Asefi, H., Lim, S., Maghrebi, M., & Shahparvari, S. (2019). Mathematical modelling and heuristic approaches to the location-routing problem of a cost-effective integrated solid waste management. Annals of Operations Research, 273, 75–110. https://doi. org/10.1007/s10479-018-2912-1. Badran, M., & El-Haggar, S. (2006). Optimization of municipal solid waste management in Port Said – Egypt. Waste Management, 26, 534–545. https://doi.org/10.1016/j. wasman.2005.05.005. Barth, M., Younglove, T., & Scora, G. (2005). Development of a heavy-duty diesel modal emissions and fuel consumption model. Technical Report UCB-ITS-PRR-2005-1. California PATH Program. Institute of Transportation Studies, University of California at Berkeley. Bektaş, T., & Laporte, G. (2011). The pollution-routing problem. Transportation Research Part B: Methodological, 45, 1232–1250. https://doi.org/10.1016/j.trb.2011.02.004. Bérubé, J. F., Gendreau, M., & Potvin, J. Y. (2009). An exact ϵ-constraint method for biobjective combinatorial optimization problems: Application to the traveling salesman problem with profits. European Journal of Operational Research, 194, 39–50. https:// doi.org/10.1016/j.ejor.2007.12.014. Bing, X., Bloemhof, J. M., Ramos, T. R. P., Barbosa-Povoa, A. P., Wong, C. Y., & van der Vorst, J. G. (2016). Research challenges in municipal solid waste logistics management. Waste Management, 48, 584–592. https://doi.org/10.1016/j.wasman.2015.11.025. Costi, P., Minciardi, R., Robba, M., Rovatti, M., & Sacile, R. (2004). An environmentally sustainable decision model for urban solid waste management. Waste Management, 24, 277–295. https://doi.org/10.1016/S0956-053X(03)00126-0. Demir, E., Bektaş, T., & Laporte, G. (2011). A comparative analysis of several vehicle emission models for road freight transportation. Transportation Research Part D: Transport and Environment, 16, 347–357. https://doi.org/10.1016/j.trd.2011.01.011. Demir, E., Bektaş, T., & Laporte, G. (2014). The bi-objective pollution-routing problem. European Journal of Operational Research, 232, 464–478. https://doi.org/10.1016/j. ejor.2013.08.002. Dukkanci, O., Bektas, T., & Kara, B. Y. (2019). Green network design problems. In J. Faulin, S. E. Grasman, A. A. Juan, & P. Hirsch (Eds.). Sustainable transportation and smart logistics (pp. 169–206). Elsevier. https://doi.org/10.1016/B978-0-12-8142424.00007-7. Edalatpour, M. A., hashem, S. M. A., Karimi, B., & Bahli, B. (2018). Investigation on a novel sustainable model for waste management in megacities: A case study in Tehran municipality. Sustainable Cities and Society, 36, 286–301. https://doi.org/10.1016/j. scs.2017.09.019. Eiselt, H. (2007). Locating landfills-optimization vs. reality. European Journal of Operational Research, 179, 1040–1049. https://doi.org/10.1016/j.ejor.2005.11.039. Erkut, E., Karagiannidis, A., Perkoulidis, G., & Tjandra, S. A. (2008). A multicriteria facility location model for municipal solid waste management in North Greece. European Journal of Operational Research, 187, 1402–1421. https://doi.org/10.1016/ j.ejor.2006.09.021. Farahani, R. Z., Fallah, S., Ruiz, R., Hosseini, S., & Asgari, N. (2019). OR models in urban service facility location: A critical review of applications and future developments. European Journal of Operational Research, 276, 1–27. https://doi.org/10.1016/j.ejor. 2018.07.036. Ferri, G. L., de Lorena Diniz Chaves, G., & Ribeiro, G. M. (2015). Reverse logistics network for municipal solid waste management: The inclusion of waste pickers as a Brazilian legal requirement. Waste Management, 40, 173–191. https://doi.org/10.1016/j. wasman.2015.02.036. Fiorucci, P., Minciardi, R., Robba, M., & Sacile, R. (2003). Solid waste management in urban areas: Development and application of a decision support system. Resources, Conservation and Recycling, 37, 301–328. https://doi.org/10.1016/S0921-3449(02)00076-9. Franceschetti, A., Honhon, D., Woensel, T. V., Bektaş, T., & Laporte, G. (2013). The timedependent pollution-routing problem. Transportation Research Part B: Methodological, 56, 265–293. https://doi.org/10.1016/j.trb.2013.08.008. Ghiani, G., Laganà, D., Manni, E., Musmanno, R., & Vigo, D. (2014). Operations research in solid waste management: A survey of strategic and tactical issues. Computers & Operations Research, 44, 22–32. https://doi.org/10.1016/j.cor.2013.10.006. Global Petrol Prices (2018). Diesel prices around the world. (Online) Accessed 26.10.18 https://www.globalpetrolprices.com/diesel_prices/. Habibi, F., Asadi, E., Sadjadi, S. J., & Barzinpour, F. (2017). A multi-objective robust optimization model for site-selection and capacity allocation of municipal solid waste facilities: A case study in Tehran. Journal of Cleaner Production, 166, 816–834. https://doi.org/10.1016/j.jclepro.2017.08.063. Haimes, Y., Lasdon, Y. L., & Wismer, A. D. (1971). On a bicriterion formulation of the problems of integrated system identification and system optimization. IEEE Transactions on Systems, Man, and Cybernetics – TSMC, 1, 296–297. https://doi.org/ 10.1109/TSMC.1971.4308298. Heidari, R., Yazdanparast, R., & Jabbarzadeh, A. (2019). Sustainable design of a municipal solid waste management system considering waste separators: A real-world application. Sustainable Cities and Society, 47, 101457. https://doi.org/10.1016/j.scs. 2019.101457.

16