Policy-driven municipal solid waste network optimization under carbon regulation: a risk-informed MILP framework for a post-conflict recovery city Authors: Jamil Hallak, AH Abdul Hafez Published: 2026 ABSTRACT AbstractMunicipal solid waste planning in a rebuilding city is not only a question of cost or technology choice. It also depends on access, operational feasibility, environmental exposure, and the policy conditions under which the system can be rebuilt. This study develops a policy-driven mixed-integer linear programming (MILP) model for designing a low-carbon municipal solid waste (MSW) network that includes transfer stations, mechanical–biological treatment (MBT) facilities, waste-to-energy (WtE) plants, and landfills. The model evaluates four planning concerns together: total cost, carbon emissions, implementation feasibility risk, and site environmental risk. In addition to the four-goal baseline formulation, the study explicitly examines three carbon-policy regimes: carbon-tax internalization, carbon tax combined with processing subsidy, and cap-and-trade. The model is tested in Idlib Governorate in northwest Syria, where formal municipal data are scarce. For this reason, the case-study dataset combines field-based information, input from eleven experts, geographic information system (GIS)-based risk assessment, fuzzy Full Consistency Method (Fuzzy-FUCOM) weighting, and weighted goal programming. The baseline solution opens three transfer stations, two MBT facilities, one WtE plant, and two landfills, generating about 141,128 MWh of electricity per year at an annual cost of USD 216.8 million. The sensitivity results show that waste supply has the largest effect on cost, while the MBT-to-WtE fraction has the strongest effect on emissions and energy recovery. The policy scenarios show a mainly financial effect under the tested settings: carbon tax increases the monetized cost of emissions, processing subsidy reduces part of the treatment cost, and cap-and-trade creates either permit costs or trading revenue depending on the cap. Overall, the framework provides a practical planning structure for MSW network design where data, infrastructure, and governance capacity are limited.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58566-z.Keywords: Municipal solid waste management, Reverse logistics network, Waste-to-energy, Carbon, Goal programming, Fuzzy-FUCOM, GIS, Risk assessmentSubject terms: Engineering, Environmental sciences, Environmental social sciences Introduction Municipal solid waste management is a basic public service and a planning problem of environmental and economic consequences. Cities are not the same in terms of waste conditions and there is no universal design of the system. The choice of collection, transfer, treatment, recovery and disposal depends on local infrastructure, regulations, available technologies, markets for recovered materials and cost conditions. This has resulted in a growing study of MSW as an integrated system rather than a disposal activity. This is especially relevant where landfilling is still common, as landfill-based systems are associated with long-term environmental impacts and greenhouse-gas emissions1–4. Municipal solid waste (MSW) is defined in this study as the waste generated by residential, commercial, governmental, and institutional sources (U.S. EPA). When this boundary is set, the main question of planning is how to trade off competing goals. The cheapest network is not always the lowest-carbon network and greener options often require extra investment, infrastructure or operational capacity. This is in line with the broader sustainability outlook in MSW studies that calls for the integration of economic, environmental and social dimensions (U.S. EPA, 2023; WCED, 19874,5;. These trade-offs become more apparent when the system is visualized as a network. Waste goes through a series of interconnected stages: collection zones, transfer stations, treatment facilities, recovery processes and final disposal sites. Thus, decisions on facility location, capacity, routing and allocation need to be considered in an integrated way. Transport is also a big part of the problem, in terms of cost and emissions. This was demonstrated in the case of Ankara, where a bi-objective model was used to minimize the total system cost and transport-related carbon emissions. Other integrated network studies have also demonstrated the advantage of modeling siting, capacity, transport, and waste flows in one framework6–9. City-level studies also show why local adaptation matters. In Port Said, Egypt, Badran and El-Haggar10used mixed-integer programming to improve the municipal waste system, including collection stations. Rathi11, developed an integrated model for Mumbai that compared waste-management options using economic and environmental costs and benefits. Tan et al.12, modeled the MSW process network in Iskandar Malaysia, focusing on how treatment and disposal pathways interact. Other studies examined narrower components, such as transfer-station location and capacity planning in Nashik, India13, and recycling-network planning in Tehran, Iran14. Together, these cases show that MSW optimization depends heavily on local waste quantities, road access, technology choices, infrastructure limits, and policy priorities. Disposal to landfill continues to be widely practiced in many MSW systems, but it poses well-known environmental and health risks, including groundwater contamination, landfill gas emissions, soil degradation and public health exposure. Therefore, MSW planning is being increasingly considered as a comprehensive system where prevention, recovery, treatment and final disposal are assessed at the same time8,15,16. WtE is often taken into account in this system, as it can reduce the reliance on landfill and recover value from residual waste. For example, Mavrotas et al.17, generated a set of Pareto optimal alternatives for the MSW system of Athens, from lower-cost to lower-emission solutions. Their work shows that the planning task is not to find the least expensive solution, but to find a compromise that can be justified when the environmental costs and benefits are considered. Evidence from landfill dominated systems also indicates that the performance of WtE is dependent on the position in the wider waste network, not just the technology itself1,17. Policy adds another layer to the decision. Landfill-diversion targets, environmental regulations, emission controls, and technology requirements can influence which facilities are feasible and how waste moves through the system. Therefore, optimization models are most useful when they make these trade-offs visible rather than reducing the problem to only cost minimization5,16–18. The post-conflict recovery context makes this planning problem more difficult than in a conventional urban setting. In many cities, optimization is used to improve an existing MSW system; in a rebuilding city, however, the first challenge may be to restore a service network that has been disrupted by damaged roads, weak institutions, limited equipment, unsafe routes, and incomplete records. UNDP (2016) notes that crisis and post-crisis settings can face both increased waste pressures and reduced local capacity, especially where displacement, damaged infrastructure, and limited municipal resources affect service delivery. Disaster and conflict-related waste may also block access, delay reconstruction, increase exposure to hazardous materials, and encourage uncontrolled dumping when regular collection and disposal systems are interrupted (OCHA et al., 2011). Evidence from post-conflict urban areas in Sri Lanka similarly shows that MSW management can be constrained by weak regulation, poor source separation, irregular collection, lack of technical capacity, and limited resources19. These conditions explain why MSW planning in a recovery city cannot be treated only as a cost or technology-selection problem. It also requires attention to access, implementation feasibility, environmental exposure, data scarcity, and governance constraints. Against this background, this study develops a policy-driven MILP decision-support framework for MSW network design in a post-conflict recovery city. The model jointly determines facility opening decisions and waste flows across transfer stations, mechanical–biological treatment facilities, waste-to-energy plants, and landfills. Its main contribution is not limited to adding another optimization model for MSW planning, but lies in adapting the model structure to a rebuilding context where operational feasibility, environmental vulnerability, and policy constraints are central to the decision. For this reason, the framework includes implementation feasibility risk and site environmental vulnerability risk alongside cost and carbon emissions. It also represents policy measures such as route restrictions, landfill limits, mandatory waste-to-energy participation, carbon taxation, processing subsidies, and cap-and-trade through model constraints or scenario parameters. This makes it possible to examine how different planning and policy choices affect the balance between cost, emissions, risk, and recovery-oriented infrastructure development. Literature review MSW management as a reverse-logistics network design problem Recent studies increasingly treat municipal solid waste management as an integrated, multi-echelon network in which collection points, intermediate facilities, treatment options, and final disposal sites should be designed together rather than addressed separately. This view treats MSW management as more than disposal. MSW planning is therefore a system problem. It includes transport, treatment, recovery, disposal, and environmental performance, and it has to balance economic, environmental, and social concerns3–5. Network-design models are useful because these decisions are connected. Facility location, treatment selection and waste-flow assignment cannot be optimized independently without the risk of a poor overall design. Many studies of MSW therefore use reverse-logistics or supply-chain formulations, where the waste flows from generation points to consolidation, treatment, recovery and final disposal. Multi-echelon models are based on this logic that links facility siting, material flows and downstream processing decisions2,6,8,9. Transfer stations are particularly important because they connect local collection with larger treatment and disposal facilities. Their location influences transport distance, fleet requirements and the viability of downstream options. This was addressed by Chatzouridis and Komilis20 with a binary-programming model for transfer-station location and design. More generally, facility location remains central in MSW planning because it shapes both cost and network feasibility21. The Ankara case adds one more point: transport decisions can affect total cost and carbon emissions, thus flow decisions need to be explicitly modeled7. Treatment portfolios, MBT coupling, and the role of WtE Most MSW systems use a combination of treatment routes, not just one technology. Recycling, composting, MBT, anaerobic digestion, WtE and landfill disposal may all have a role to play, depending on the waste stream and local operating conditions. The planning question is thus how to put these options together in one network. For this reason, several studies evaluate treatment portfolios rather than isolated technologies6,8,22. In this case, MBT is important because it can recover part of the waste and prepare the remaining stream for WtE or landfill disposal8,17. WtE is often included due to its potential to reduce reliance on landfill and to produce energy from residual waste. But its performance depends on the larger network. Mavrotas et al.17, show that WtE must be evaluated with respect to its cost and environmental trade-offs, and Aracil et al.1, demonstrate that its economic and greenhouse-gas performance depends on its integration in a system where landfills dominate. The argument is central to the present model in which transfer stations, MBT plants, WtE plants and landfills are connected by material flows. MBT is not only a treatment facility in this structure, it also controls the amount of material that is recovered, sent to WtE or finally disposed of. Cost–carbon trade-offs, externalities, and decision support The MSW optimization is no longer restricted to the cheapest system design. Cost is still a factor but lower emissions, landfill diversion and environmental protection may yield different flow and facility choices. Hence, several studies have employed bi-objective or multi-objective models to simultaneously address economic and environmental objectives7,15,23. For example, Mohsenizadeh et al.7, simultaneously minimized the total cost and transport-related carbon emissions in an MSW facility-planning model. Their results provided a range of compromise solutions, rather than a single definitive solution. Mavrotas et al.17, followed a similar logic comparing MSW alternatives across cost and greenhouse-gas objectives. This type of modeling is useful because municipal planners often need to understand what is being gained and what is being sacrificed in each option, not only which design is cheapest17,18. There are two main ways carbon can enter the model. It can be optimized as a separate objective or can be included in the cost function as a price. Mavrotas et al.17, showed that adding external costs and benefits can change the relative attractiveness of treatment options, including WtE. This supports the use of carbon-pricing logic in MSW network models, where emissions are either minimized directly or translated into monetary terms7,15,17. Policy and governance constraints in MSW network design Policy can affect an MSW network in practical ways. It may restrict some routes, promote certain treatment options, limit landfill disposal, or place a cost on emissions. The European Union is a useful example. Its amended Landfill Directive limits municipal waste sent to landfill to 10% or less by 2035, while the amended Waste Framework Directive sets preparation-for-reuse and recycling targets of 55%, 60%, and 65% for 2025, 2030, and 2035, respectively (European Parliament and Council24,25,. In an optimization model, these targets can be written as constraints, scenario assumptions, or aspiration levels. This policy dimension appears in several MSW decision-support studies. Mavrotas et al.17, included external costs and benefits in a multi-objective MSW model, which changed the relative attractiveness of WtE options. Tan et al.12, compared treatment pathways in the Iskandar Malaysia MSW network, and Ooi et al.26, showed that waste-separation rates can affect both cost and greenhouse-gas emissions at the national scale in Malaysia. These studies suggest that MSW optimization should not be limited to cost minimization; it also needs to reflect environmental targets and treatment preferences. Other work links modeling more directly with policy targets. Chew et al.27, used Carbon Emission Pinch Analysis to connect national greenhouse-gas reduction targets with local waste-policy targets. Malinauskaite28, also show that WtE development in Europe is shaped by circular-economy policy, landfill restrictions, recycling priorities, and energy-recovery rules. This matters for networks that combine MBT, WtE, and landfill disposal, because policy can influence not only the cost of a route, but also whether that route is allowed or desirable. In post-conflict or rebuilding-city contexts, governance constraints become even more decisive. The issue is not only whether a policy target exists, but whether local authorities have the capacity, equipment, financing, data, and institutional stability to implement it. UNDP (2016) notes that local governments may be unable to resume municipal solid waste services after a crisis, creating public-health risks and hampering recovery efforts. Evidence from Libya’s post-conflict waste and construction-demolition waste context also shows that recovery-oriented waste management can be constrained by shortages of specialized manpower, lack of equipment and facilities, weak policy and governance arrangements, limited empirical data, and low public recognition of post-conflict waste as a recoverable resource29. Similar policy and operational gaps have been observed in post-conflict urban areas in Sri Lanka, where MSW management was affected by weak regulation, limited resources, irregular collection, poor source separation, and lack of technical capacity19. For this reason, policy-sensitive optimization in a rebuilding city should not be limited to carbon pricing or treatment targets. It should also represent route restrictions, minimum service requirements, landfill limits, technology participation rules, and risk-related siting constraints that reflect the practical conditions under which the system can actually be rebuilt. Positioning of the present study Building on this literature, the present study contributes a policy-driven MSW network model for a post-conflict recovery city where conventional cost–emission optimization is not sufficient. The framework combines transfer stations, MBT facilities, WtE plants, and landfills in one MILP model. It keeps implementation feasibility risk and site environmental vulnerability risk as separate dimensions, rather than treating risk as a single general term. This matters in a rebuilding city, where access, security, data gaps, environmental exposure, and weak institutional capacity can determine whether a proposed network is actually workable. The model also includes policy conditions such as carbon taxation, processing support, cap-and-trade, route restrictions, landfill limits, and mandatory WtE participation. As a result, it can examine facility selection, waste-flow allocation, risk-informed siting, and policy scenarios within the same planning model. Comparison with previous studies Compared with earlier MSW optimization studies, this work adds a broader network and a more explicit recovery-context perspective. Mohsenizadeh et al.7,, for example, focused on transfer-station planning with cost and transport-related CO₂ objectives supported by GIS analysis. The present study goes beyond that setting by linking transfer stations, MBT facilities, WtE plants, and landfills in one policy-driven MILP model, while also adding implementation feasibility risk and site environmental vulnerability risk as separate decision dimensions. Yousefloo and Babazadeh8, proposed an integrated MSWM network with economic and risk objectives, including transfer stations and WtE options; however, the current study goes further by separating GIS-based implementation feasibility risk from expert-based environmental risk and by embedding the model in a conflict-affected rebuilding-city context. Mavrotas et al.17, highlighted the importance of multi-objective planning and external cost considerations in WtE-related MSW systems, especially in revealing cost–emission trade-offs. Building on this line, the present study introduces three explicit carbon-policy regimes—carbon tax, processing subsidy, and cap-and-trade—and examines how these policies influence network cost, emissions, and risk within a unified decision-support framework30–34. Methodology Figure 1 presents the methodological workflow used to develop the risk-informed and policy-driven MSW reverse logistics network optimization model. The workflow includes data preparation, Delphi-based goal definition, risk assessment, Fuzzy-FUCOM weighting, MILP model formulation, solution analysis, sensitivity analysis, and carbon-policy scenario evaluation. Methodological workflow for the risk-informed and policy-driven MSW network optimization framework. Data collection in conflict areas The case-study dataset was prepared as a planning dataset for strategic MSW network design in Idlib Governorate. First, the spatial structure of the network was defined by identifying 26 collection zones representing the main neighborhoods in Idlib, together with 14 candidate transfer stations, 5 candidate mechanical–biological treatment facilities, 3 waste-to-energy sites, and 3 landfill sites. Candidate facility locations were identified in consultation with local councils and local technical informants. These locations were organized in a GIS environment and linked to an updated road-network layer to generate origin–destination distance matrices for the active transport links in the model, including collection-to-transfer, transfer-to-MBT, MBT-to-WtE, MBT-to-landfill, and WtE-to-landfill movements. Second, operational and economic inputs were converted into annual model parameters, including waste quantities, facility capacities, fixed opening costs, handling and processing costs, disposal costs, and technical conversion parameters. Because formal municipal records in the study area are limited, the dataset was cross-checked using field-based information, local technical judgement, expert review, and parameter assumptions informed by comparable service and cost conditions, particularly those observed in Gaziantep, Türkiye. The dataset should therefore be read as a validated planning dataset rather than as a synthetic dataset or a complete official municipal registry. Table 1 summarizes the main input categories. The detailed parameter table and a small excerpt from the base-case dataset, including sample waste quantities, facility data, distance values, and risk scores, are provided in Supplementary Information S1. Table 1Main categories of input data used in the Idlib MSW network model.Data categoryUnitSource or derivationModel roleNetwork nodes and candidate facilitiescountCase-study spatial dataset, local councils, local technical informants, and GIS location preparationDefines collection zones and candidate transfer, MBT, WtE, and landfill facilitiesAnnual waste supplyton/yearField-based estimation and model input fileDefines the annual waste quantity that must be routed from each collection zoneFacility capacitieston/yearCase-study parameterization informed by local judgement and comparable operating conditionsLimits the annual throughput of opened transfer, MBT, WtE, and landfill facilitiesFixed facility-opening costsUSD/yearCase-study parameterization and comparable service/cost conditionsRepresents the annual fixed cost of opening candidate facilitiesHandling, processing, and disposal costsUSD/tonCase-study parameterization and model input fileCalculates variable operating costs at transfer stations, MBT facilities, WtE plants, and landfillsTransport distanceskmGIS origin–destination matrices generated from the updated road-network layerCalculates transport cost and transport-related emissions between network levelsRecovery and conversion parametersfraction; MWh/tonTechnical assumptions used in the model and sensitivity analysisDetermines recyclable recovery, MBT-to-WtE allocation, WtE electricity generation, and ash generationEmission factorstCO₂/ton or tCO₂/ton-kmModel assumptions and emission-parameter settingsCalculates transport, treatment, WtE, and landfill-related emissionsImplementation feasibility risk scoresindex, 1–5GIS-based risk assessmentRepresents operational/spatial feasibility risk for candidate facilitiesSite environmental vulnerability risk scoresindex, 1–5Structured expert judgement through the Delphi processRepresents environmental exposure and contamination-related risk for candidate facilitiesPolicy and scenario parametersUSD/tCO₂; cap level; subsidy rate; weightsCarbon-policy assumptions, Fuzzy-FUCOM weights, and sensitivity settingsDefines carbon tax, processing subsidy, cap-and-trade, and goal-programming scenarios Main categories of input data used in the Idlib MSW network model. Delphi-based goal definition and expert elicitation The model goals and expert-based inputs were refined through a two-round Delphi process with eleven participants. The panel included specialists and local technical informants familiar with municipal services, waste management, local operations, environmental risk, logistics, and GIS-supported planning. The Delphi process was used because several parts of the dataset and model structure depended on local judgement rather than formal records. In the first round, the experts reviewed the initial goals, candidate sites, feasible links, and risk factors. Some goals were combined, some were removed, and several criteria were adjusted to fit the Idlib context. The revised list was then shared in a second round, after which the panel agreed on four goals for the model: total cost, carbon emissions, implementation feasibility risk, and site environmental vulnerability risk. The four goals were kept separate because each describes a different part of the planning problem. Cost represents the annual burden of opening and operating the network. Carbon emissions are linked to transport, treatment, WtE, and landfill disposal. Implementation feasibility risk reflects whether a facility or route can realistically function under local access, security, road, and workforce conditions. Site environmental vulnerability risk captures possible exposure of groundwater, soil, surface water, nearby receptors, agricultural land, and sensitive ecological areas. The expert panel also reviewed the risk criteria and supported the scoring of site environmental vulnerability. Median values were used to combine expert scores, while the interquartile range was used to check agreement. Risk assessment Candidate facilities were assessed through two risk streams in the Idlib MSW reverse logistics network: Implementation feasibility–spatial risk quantified through GIS, reflecting feasibility and continuity of operations under access and security constraints. Implementation feasibility–spatial risk quantified through GIS, reflecting feasibility and continuity of operations under access and security constraints. Site environmental risk, assessed by experts because current spatial data on camps, informal settlements, and water-use patterns are incomplete. Site environmental risk, assessed by experts because current spatial data on camps, informal settlements, and water-use patterns are incomplete. Implementation feasibility risk (GIS-based). The selected factors were based on location, so GIS was utilized for implementation feasibility risk. Each potential site was rated on four factors: Accessibility and proximity to roads: poor access, poor or low class roads, bottlenecks and long detours which can delay collection and haulage. Accessibility and proximity to roads: poor access, poor or low class roads, bottlenecks and long detours which can delay collection and haulage. Security exposure: insecure routes or high risk areas that could interrupt movement, affect the safety of staff or expose assets to damage. Security exposure: insecure routes or high risk areas that could interrupt movement, affect the safety of staff or expose assets to damage. Proximity to the front line: Proximity to unstable areas or recent frontlines, where there may be restrictions on access, renewed hostilities or military activity. Proximity to the front line: Proximity to unstable areas or recent frontlines, where there may be restrictions on access, renewed hostilities or military activity. Workforce availability: shortage of skilled or semi-skilled workers, high turnover, or weak local availability of labor which can affect daily operation. Workforce availability: shortage of skilled or semi-skilled workers, high turnover, or weak local availability of labor which can affect daily operation. Figure 2 presents the implementation feasibility risk maps. For each criterion , the relevant spatial layer was prepared, such as a distance raster, density surface, or network-based accessibility layer. The layers were then reclassified to a common risk scale from 1 to 5. For each candidate site, the criterion score was extracted from its location. The composite GIS-based implementation feasibility risk score was then calculated as follows: Implementation feasibility risk maps. where are criterion weights. The resulting provides a comparable operational risk score across all candidate sites (higher values indicate higher implementation risk). Site environmental vulnerability risk (expert-based, 11 experts). Some of the exposure factors in Idlib are not well captured in current spatial datasets, and so expert judgement was used to assess site environmental risk. This is particularly the case with changing camp locations, informal settlements and water-use patterns. Eleven experts, with experience in environmental health, waste management, water resources, local services, safety and logistics, reviewed the candidate sites. Six environmental factors were used: Groundwater and soil vulnerability: possible leachate movement through soil and subsurface layers into groundwater. Groundwater and soil vulnerability: possible leachate movement through soil and subsurface layers into groundwater. Surface-water proximity and watershed sensitivity: possible pollutant movement toward rivers, streams, canals, wetlands, or drainage paths through runoff, seepage, or accidental releases. Surface-water proximity and watershed sensitivity: possible pollutant movement toward rivers, streams, canals, wetlands, or drainage paths through runoff, seepage, or accidental releases. Flood and drainage risk: flooding or poor stormwater drainage that may cause overtopping, liner damage, contaminant movement, or interruption of facility operation. Flood and drainage risk: flooding or poor stormwater drainage that may cause overtopping, liner damage, contaminant movement, or interruption of facility operation. Terrain stability and landslide susceptibility: Risk of slope failure, settlement, and geotechnical instability that can compromise containment systems and facility structures. Terrain stability and landslide susceptibility: Risk of slope failure, settlement, and geotechnical instability that can compromise containment systems and facility structures. Proximity to sensitive receptors: Risk of local human exposure and nuisance impacts due to closeness to vulnerable receptors such as schools, hospitals, dense residential areas, and camps. Proximity to sensitive receptors: Risk of local human exposure and nuisance impacts due to closeness to vulnerable receptors such as schools, hospitals, dense residential areas, and camps. Protected areas, biodiversity, and agricultural/food-system sensitivity: Ecological degradation or food-system impacts when facilities are located near protected habitats, high-biodiversity zones, productive agricultural land and irrigation networks. Protected areas, biodiversity, and agricultural/food-system sensitivity: Ecological degradation or food-system impacts when facilities are located near protected habitats, high-biodiversity zones, productive agricultural land and irrigation networks. Each expert assigns a score on a five-point ordinal scale (1 very low to 5 very high) for each site and criterion . Scores are aggregated (median): For each site sand criterion m: For each site sand criterion m: Use as the aggregated score. We used IQR to show agreement where IQR is used as a “consensus” indicator. We used IQR to show agreement where IQR is used as a “consensus” indicator. A composite environmental risk score is then computed as: where are environmental criterion weights (expert-derived weights). The resulting expresses the environmental sensitivity of each candidate site (higher values indicate higher environmental risk). At the end of the risk assessment stage, each candidate site shas two independent indicators: : GIS-derived implementation feasibility risk score. : GIS-derived implementation feasibility risk score. : expert-derived site environmental risk score. : expert-derived site environmental risk score. These are kept separate because they represent different mechanisms (operational continuity vs. environmental) and rely on different evidence sources (spatial computation vs. expert judgement). In the optimization stage, the two risk components have been introduced as separate objectives/goals alongside economic and carbon metrics. The goal programming in this study is to consider four goals: total cost, total carbon emissions, implementation risk, and environmental risk. Prioritizing risks and objectives using Fuzzy-FUCOM The final goal and risk-criterion priorities were converted into weights using fuzzy-FUCOM. This method was selected because it represents expert preferences under uncertainty, which is important in a data-scarce post-conflict setting. The risk-criterion weights were used to calculate the composite implementation feasibility and site environmental vulnerability scores, while the objective weights were used in the weighted goal-programming formulation. The resulting weights were then incorporated into the optimization model described below, while the complete algebraic formulation is provided in Supplementary Information S1. MILP model structure and assumptions The optimization model was formulated as a single-period annual MILP for strategic MSW network design. It does not model daily vehicle routing. The model decides two main issues: which facilities to open and how annual waste flows should be allocated between them. Accordingly, all quantities, capacities, costs, emissions, and revenues are treated on an annual basis. The main flow-balance, capacity, cost, and emission relationships are presented in Eqs. 4–22. The network has five node types: collection zones, transfer stations, MBT facilities, WtE plants, and landfills. Collection zones represent the main neighborhoods in Idlib, while transfer stations consolidate waste before treatment. MBT facilities recover recyclable material and prepare the remaining stream for WtE or landfill disposal. WtE plants generate electricity and ash from part of this residual stream, and landfills receive the non-recovered residues and WtE ash. The related input data, including capacities, costs, distances, and risk scores, are summarized in Table 1. In the baseline case, waste follows the selected recovery route: collection zones → transfer stations → MBT facilities, and then to recycling, WtE, or landfill disposal. Ash from WtE is also sent to landfill. This routing structure is captured by the collection-zone balance, transfer-station balance, MBT input and material-balance equations, WtE input and ash-balance equations, and landfill input equation (Eqs. 8, 9, 11, 13, 16, 18, and 19). Although the general formulation includes direct collection-to-MBT, collection-to-landfill, and transfer-to-landfill flows, these flows are fixed to zero in the baseline policy setting to enforce the transfer–MBT recovery pathway. Facility-opening decisions are represented by binary variables, while annual waste movements are represented by continuous flow variables. A transfer station, MBT facility, WtE plant, or landfill can receive flow only when it is opened. This link between opening decisions and usable capacity is enforced through the transfer, MBT, WtE, and landfill capacity constraints (Eqs. 10, 12, 17, and 20). The first objective minimizes total annual system cost. As shown in Eq. 4, this cost includes fixed annual opening costs for selected transfer stations, MBT facilities, WtE plants, and landfills; transport costs calculated from annual flows, GIS-derived distances, and the unit transport cost; handling, treatment, and disposal costs; and revenues from recovered recyclables and electricity generated by WtE. The cost equation therefore connects the infrastructure decision with the spatial structure of the network and the economic parameters listed in Table 1. The second objective minimizes total carbon emissions. Equation 5 includes transport-related emissions, calculated from flows, distances, and the transport emission factor, together with facility-related emissions from MBT processing, WtE processing, and landfill disposal. This separates the carbon effect of moving waste through the network from the carbon effect of treating or disposing of it. In the baseline goal-programming model, total carbon emissions are treated as a separate objective, while in the carbon-policy scenarios they are also used to calculate carbon-tax costs or cap-and-trade balances. The third and fourth objectives represent the two risk dimensions. Implementation feasibility risk is calculated in Eq. 6 by summing the GIS-derived implementation-risk scores of the opened facilities. These scores reflect spatial and operational concerns such as road accessibility, security exposure, frontline proximity, and workforce availability. Site environmental vulnerability risk is calculated in Eq. 7 by summing the expert-derived environmental-risk scores of the opened facilities. These scores reflect groundwater and soil vulnerability, surface-water sensitivity, flood and drainage risk, terrain stability, sensitive receptors, and agricultural or ecological sensitivity. Keeping Eqs. 6 and 7 separate allows the model to distinguish between operational feasibility and environmental exposure instead of merging both into one broad risk indicator. The material-recovery and energy-conversion relationships are represented through fixed linear yield assumptions. Recovered recyclables from each MBT facility are calculated as a fixed fraction of MBT input in Eq. 14, while the quantity sent from MBT to WtE is determined by the MBT-to-WtE fraction in Eq. 15. Total electricity generation is then calculated in Eq. 21 by multiplying WtE input by the electricity conversion factor. These fixed-yield relationships keep the model linear and appropriate for strategic planning. They also make it possible to test key technical assumptions, such as recyclable recovery, MBT-to-WtE allocation, and WtE electricity conversion, through sensitivity analysis. The policy structure is represented through additional routing and technology requirements. In the baseline recovery policy, direct collection-to-MBT, direct collection-to-landfill, and transfer-to-landfill flows are fixed to zero, so the main route becomes collection → transfer → MBT → recycling/WtE/landfill. In addition, Eq. 22 requires at least one WtE facility to be opened. This requirement reflects the policy assumption that energy recovery should be included in the rebuilding network rather than being evaluated only as an optional technology. Because the model contains four competing objectives, it is solved using non-preemptive weighted goal programming. First, the model is solved separately for each individual objective to obtain aspiration levels for total cost, carbon emissions, implementation feasibility risk, and site environmental vulnerability risk. These aspiration levels are then used as target values in Eqs. 23–26. The final goal-programming objective in Eq. 27 minimizes the weighted sum of normalized positive deviations from these targets. Normalization by the target values allows deviations measured in different units, such as USD/year, tCO₂/year, and risk-index points, to be compared in a single compromise model. The weights used in Eq. 27 are derived from the Fuzzy-FUCOM procedure described earlier. The complete notation and detailed algebraic formulation, including all sets, parameters, variables, objective functions, and constraints, are provided in Supplementary Information S1. Solution procedure and sensitivity analysis The model was solved in two main stages. In the first stage, each objective was optimized separately to obtain the aspiration levels for total cost, carbon emissions, implementation feasibility risk, and site environmental vulnerability risk. The resulting values were used as targets in the weighted goal-programming model. The final solution then identified the selected facilities, waste-flow allocation, electricity generation, emissions, and risk levels. Sensitivity analysis was then used to test how the baseline solution changes under selected technical, economic, and policy assumptions. The tested factors included MBT recovery, MBT-to-WtE allocation, electricity conversion, transport cost, waste supply, objective weights, carbon tax, processing subsidy, and cap-and-trade settings. Case study in Idlib/Syria The framework is tested in Idlib Governorate in northwest Syria, where conflict has limited access to formal municipal records. The case study is structured as a five-level MSW reverse logistics network, with 26 collection zones, 14 candidate transfer stations, 5 candidate MBT facilities, 3 candidate WtE sites, and 3 candidate landfills. Together, these nodes describe the movement of waste from collection to consolidation, treatment, energy recovery, and final disposal. Data preparation, expert validation, and parameter sources are described in Sect. 3. Results and discussion Aspiration levels The model was first solved separately for each objective to define the aspiration levels used in the weighted goal-programming formulation. Table 2 shows that the four objectives lead to different network choices. The minimum-cost solution gives the lowest annual cost, but it does not provide the lowest implementation or environmental risk. The minimum-carbon solution reduces emissions only slightly compared with the minimum-cost case, but it requires a larger infrastructure configuration, with more transfer stations, MBT facilities, and WtE facilities. This raises both risk scores. For a post-conflict recovery city, this trade-off is important: a lower-carbon solution may be less practical if it depends on more facilities, higher-risk sites, or a more complex operating structure. Table 2Aspiration matrix and solution indicators under single-objective optimization.Metric (unit)Min Total CostMin CO₂Min IF riskMin SE riskTotalCost (USD/year)207,547,343220,338,519220,740,523220,740,523CO₂ (tCO₂/year)497,391496,834509,269.0509,642.3IFRisk (index points)25.874.0 20.2 22.9SERisk (index points)26.273.520.2 17.9 Electricity, E (MWh/year)141,127141,127141,127141,127Opened Transfer stations (#)3.05.03.03.0Opened MBT facilities (#)2.03.02.02.0Opened WtE facilities (#)1.02.01.01.0Opened Landfills (#)2.02.02.02.0 Aspiration matrix and solution indicators under single-objective optimization. Comparison with the epsilon-constraint method The model was also solved using the epsilon-constraint method to compare the weighted goal-programming solution with a threshold-based formulation. In this version, total cost was minimized, while carbon emissions, implementation feasibility risk, and site environmental vulnerability risk were imposed as upper bounds. In this formulation, the model looked for the lowest-cost network that could still remain within selected carbon and risk limits. Table 3 compares the WGP baseline with four epsilon-constraint cases. In EC-1, the limits were set close to the WGP results: CO₂ ≤ 506,000, IF risk ≤ 23, and SE risk ≤ 19. The model then produced a lower-cost solution, USD 213.11 million, while keeping emissions and risk values close to the baseline. EC-2 used a stricter carbon limit, CO₂ ≤ 500,000, but allowed wider risk limits. This reduced cost to USD 208.55 million and emissions to 498,319 tCO₂/year, but increased the site environmental risk to 23.49. EC-2 therefore gives an important warning: a solution can look better in terms of cost and carbon, but still be less suitable if it shifts the network toward higher-risk sites. Table 3Comparison between weighted goal programming and epsilon-constraint solutions.Method/caseOptimization logicEpsilon boundsTotal cost USD/yearCO₂ tCO₂/yearIF riskSE riskElectricity MWh/yearFeasibilityWGP baselineMinimize weighted normalized deviationsNot applicable216,812,789505,99522.3018.90141,128FeasibleEC-1 balanced boundsMinimize total costCO₂ ≤ 506,000; IF ≤ 23;SE ≤ 19213,113,893502,56022.5218.94141,128FeasibleEC-2 low-carbon/relaxed-risk caseMinimize total costCO₂ ≤ 500,000; IF ≤ 26;SE ≤ 24208,546,086498,31923.7323.49141,128FeasibleEC-3 strict-risk caseMinimize total costCO₂ ≤ 510,000; IF ≤ 21;SE ≤ 18—————InfeasibleEC-4 moderate-risk caseMinimize total costCO₂ ≤ 510,000; IF ≤ 22;SE ≤ 19212,436,211501,93121.9818.94141,128Feasible Comparison between weighted goal programming and epsilon-constraint solutions. CO₂ ≤ 506,000; IF ≤ 23; SE ≤ 19 CO₂ ≤ 500,000; IF ≤ 26; SE ≤ 24 CO₂ ≤ 510,000; IF ≤ 21; SE ≤ 18 CO₂ ≤ 510,000; IF ≤ 22; SE ≤ 19 EC-3 was infeasible when the risk limits were tightened to IF risk ≤ 21 and SE risk ≤ 18. This means that, with the available sites, capacities, routing structure, and mandatory WtE requirement, the model could not meet both risk limits at the same time. The epsilon-constraint comparison therefore confirms that threshold-based targets are useful for policy testing, but overly strict risk limits may not be feasible in the current rebuilding context. This supports the need for a compromise-based approach in post-conflict MSW planning, where the technically desirable targets may not all be attainable within the same infrastructure configuration. After slightly relaxing the risk limits in EC-4 to IF risk ≤ 22 and SE risk ≤ 19, the model became feasible. The resulting solution had a total annual cost of USD 212.44 million, CO₂ emissions of 501,931 tCO₂/year, IF risk of 21.98, and SE risk of 18.94. Compared with the WGP baseline, this solution reduced total cost and CO₂ while slightly improving implementation feasibility risk. However, the environmental risk remained close to the upper bound and slightly above the WGP value. This confirms that the epsilon-constraint method is useful for testing threshold-based policy choices, but the results depend strongly on how the epsilon bounds are selected. Electricity generation remained almost unchanged across all feasible cases, at about 141,128 MWh/year. This is expected because the WtE throughput is mainly controlled by the fixed MBT-to-WtE fraction and the requirement that at least one WtE facility must be opened. Therefore, the epsilon-constraint cases mainly changed facility/site selection and risk-cost trade-offs rather than the total energy-recovery level. Overall, the epsilon-constraint cases confirm that moderate bounds can produce feasible lower-cost alternatives, while strict risk limits may make the model infeasible. Sensitivity analysis MBT recyclable recovery fraction . A sensitivity analysis was conducted on the MBT recyclable recovery fraction (± 10%). As increases, the model forces a higher share of MBT inflow to be recovered as recyclables, thereby reducing MBT residue delivered to landfills and lowering the landfill-emissions component of total CO₂. Consequently, CO₂ decreases consistently across the tested range. Total cost decreases from the low to the baseline recovery level due to higher recycling revenues and reduced landfill disposal, but exhibits a slight rebound at the upper recovery level. This non-monotonic cost response is explained by the weighted goal-programming formulation, which minimizes normalized positive deviations from aspiration levels; thus, the optimizer may accept marginal cost increases when they yield larger reductions in carbon and/or risk deviations. Facility opening decisions and energy generation remain stable across the tested range, indicating that moderate changes in influence the solution primarily through flow composition rather than structural network reconfiguration. the results are shown in Table 4. Table 4Sensitivity results for MBT recyclable recovery fraction . TotalCost (USD/year)CO₂ (tCO₂/year)IFRisk (index)SERisk (index)E (MWh/year)Transfer (#)MBT (#)WtE (#)Landfill (#)0.05221,427,983532,70921.422.8141,127.832120.10216,812,789505,99522.318.9141,127.832120.15218,211,577484,86423.118.9141,127.83212 Sensitivity results for MBT recyclable recovery fraction . MBT recyclable recovery fraction . Table 5 shows that increasing the MBT-to-WtE fraction shifts a larger share of MBT output from landfill disposal to energy recovery. As a result, annual electricity generation increases substantially, from about 70,564 MWh/year at to 141,128 MWh/year at the baseline value and 188,170 MWh/year at . However, CO₂ emissions do not decrease with higher WtE allocation. In the model inputs, the WtE process-emission factor is relatively high, so the additional emissions from sending more waste to WtE outweigh the landfill-emission reduction achieved through diversion. Accordingly, CO₂ increases from about 494,223 tCO₂/year at the low WtE fraction to 505,995 tCO₂/year at the baseline and 520,329 tCO₂/year at the high WtE fraction. Total cost decreases slightly at the baseline because of higher electricity revenue, but it rises sharply at , mainly because a second WtE facility is opened and additional WtE processing costs are incurred. This indicates that the MBT-to-WtE fraction is one of the most influential technical levers in the model, affecting energy recovery, emissions, cost, and facility-opening decisions. Table 5Sensitivity results for MBT-to-WtE fraction . TotalCost (USD/year)CO₂ (tCO₂/year)IFRisk (index)SERisk (index)E (MWh/year)Transfer (#)MBT (#)WtE (#)Landfill (#)0.15221,567,498494,22321.518.670,563.932120.30216,812,789505,99522.318.9141,127.832120.40265,626,972520,32924.422.3188,170.43222 Sensitivity results for MBT-to-WtE fraction . Electricity conversion factor (MWh/ton of WtE input). The electricity conversion factor mainly affects energy output and revenue. When increases, electricity generation rises almost linearly because WtE throughput is still fixed by the MBT-to-WtE allocation. The additional electricity revenue lowers the net annual cost, as shown in Table 6. CO₂ changes only slightly because is not included in the emissions calculation; the small differences reflect minor flow adjustments in the weighted goal-programming solution. Facility openings remain unchanged, so the effect of is mainly economic, not structural. Table 6One-at-a-time sensitivity results for electricity conversion factor .(MWh/t)TotalCost (USD/year)CO₂ (tCO₂/year)IFRisk (index)SERisk (index)E (MWh/year)Transfer (#)MBT (#)WtE (#)Landfill (#)0.5219,419,599506,88723.419.5117,606.532120.6216,812,789505,99522.318.9141,127.832120.7213,211,162504,18023.018.5164,649.13212 One-at-a-time sensitivity results for electricity conversion factor . Transport cost. Transport cost sensitivity (ckm, USD/ton-km): Increasing ckm raises the TotalCost almost linearly (≈ 214.3 M → 216.8 M → 233.7 M USD/y), because transport cost enters the objective directly as . Table 7 shows only minor changes in CO₂ emissions and no change in facility openings. This is expected because changes transport cost, not the emissions equation. The small CO₂ differences result from limited flow reallocation within the weighted goal-programming solution, while the selected numbers of transfer stations, MBT facilities, WtE plants, and landfills remain unchanged. Table 7sensitivity results for transport cost coefficient ckm.(ckm) (USD/ton-km)TotalCost (USD/year)CO₂ (tCO₂/year)IFRisk (index)SERisk (index)E (MWh/year)Transfer (#)MBT (#)WtE (#)Landfill (#)0.10214,333,503508,84621.518.6141,127.832120.14216,812,789505,99522.318.9141,127.832120.20233,746,082512,44721.219.1141,127.83212 sensitivity results for transport cost coefficient ckm. Waste generation sensitivity (Supply ± 10%). As total supply increases, Total Cost increases (≈ 214.7 M → 216.8 M → 226.8 M USD/y) because higher waste volumes drive higher transport, handling, processing, and disposal flows through the whole network. The results are presented in Table 8. Table 8sensitivity results for waste generation Supply.Supply changeTotalCost (USD/year)CO₂ (tCO₂/year)IFRisk (index)SERisk (index)E (MWh/year)Transfer (#)MBT (#)WtE (#)Landfill (#)−10%214,683,828457,58021.518.6127,015.032120%216,812,789505,99522.318.9141,127.83212+ 10%226,797,704561,70523.620.6155,240.63212 sensitivity results for waste generation Supply. CO₂ increases strongly (≈ 457.6k → 506.0k → 561.7k tCO₂/y) since emissions are flow-based (transport + MBT + WtE + landfill), so more waste directly produces more emissions. Electricity generation rises (≈ 127.0k → 141.1k → 155.2k MWh/y) because WtE input is proportional to total treated waste under the fixed MBT-to-WtE fraction. Facility-opening counts stay constant, indicating existing capacities can absorb ± 10% demand without structural changes. Figure 3 presents the percentage deviation from the base-case solution when selected parameters are varied at low and high levels. The results show that the total cost is most sensitive to the WTE-related parameter, gammaWTE, where the high-level scenario increases total cost by 22.5%, representing the largest cost deviation among all tested parameters. The transport-cost parameter, ckm, also has a noticeable effect on total cost, with a 7.8% increase under the high-level scenario, while supply variation causes a smaller cost increase of 4.6%. In contrast, alpha and gammaRec have relatively limited effects on total cost. Sensitivity of total cost and Co2. For CO₂ emissions, the strongest sensitivity is observed for supply variation, where emissions increase by 11.0% in the high-level scenario and decrease by 9.6% in the low-level scenario. The recycling-related parameter, gammaRec, also has a clear influence on emissions, as the low-level scenario increases CO₂ emissions by 5.3%, while the high-level scenario reduces them by 4.2%. The WTE-related parameter, gammaWTE, has a moderate effect on CO₂ emissions, ranging from − 2.3% to 2.8%, whereas alpha and ckm show comparatively smaller effects. Overall, the sensitivity analysis indicates that cost outcomes are mainly driven by WTE-related assumptions, while emission outcomes are mainly driven by supply levels and recycling-related changes. Weight-scenario analysis. Table 9 shows how the weighted goal-programming solution changes under different FUCOM-derived weight profiles. The facility configuration remains the same in all cases: three transfer stations, two MBT facilities, one WtE facility, and two landfills. Electricity generation also remains almost unchanged at about 141,128 MWh/year. This suggests that, with the enforced transfer–MBT route and the current capacity and yield assumptions, the selected network is not very sensitive to changes in stakeholder priorities. The weights mainly affect the balance among cost, carbon, and risk values rather than the facility-opening decisions. Table 9Weight-scenario analysis based on FUCOM preferences.Preference scenarioWeights (stacked)TotalCost (USD/year)CO₂ (tCO₂/year)IFRiskSERiskE (MWh/year)TransferMBTWtELandfillCarbon-dominant wCost 0.10 wCarb 0.60 wIF 0.10 wSE 0.20 215,740,997505,000 22.3 19.4 141,127.8 3 2 1 2 Cost-dominant wCost 0.50 wCarb 0.25 wIF 0.13 wSE 0.12 215,668,836504,933 22.3 18.4 141,127.8 3 2 1 2 Operational-risk-dominant wCost 0.19 wCarb 0.05 wIF 0.57 wSE 0.19 217,362,743 506 506 21.1 20.0 141,127.8 3 2 1 2 Environmental-risk-dominant wCost 0.22 wCarb 0.11 wIF 0.22 wSE 0.45 222,466,395511,245 21.1 18.8 141,127.8 3 2 1 2 Weight-scenario analysis based on FUCOM preferences. wCost 0.10 wCarb 0.60 wIF 0.10 wSE 0.20 wCost 0.50 wCarb 0.25 wIF 0.13 wSE 0.12 wCost 0.19 wCarb 0.05 wIF 0.57 wSE 0.19 wCost 0.22 wCarb 0.11 wIF 0.22 wSE 0.45 The carbon-dominant profile gives the lowest CO₂ value among the tested weight scenarios, while total cost remains close to the baseline. The cost-dominant profile produces the lowest total cost, with only small changes in emissions and risk. When implementation feasibility risk receives the highest weight, the model slightly improves the IF risk score, but this comes with a small increase in cost and CO₂. The environmental-risk-dominant profile has the highest cost and CO_2 among the tested cases. This is a reflection of the trade-offs inherent in prioritizing environmentally safer site choices within the available candidate network. Taken together, the sensitivity results show that the selected infrastructure configuration is relatively stable under moderate changes in technical and economic parameters. In most sensitivity tests, the same facility pattern was selected: three transfer stations, two MBT facilities, one WtE plant, and two landfills. This does not mean the tested parameters have little value. Rather, it suggests that the candidate sites, routing rules, and capacity limits define a core network that is difficult to change under moderate parameter shifts. Among the tested parameters, waste supply has the strongest volume effect because any increase in waste quantity passes through transport, treatment, disposal, emissions, and electricity-generation calculations at the same time. By contrast, changes in the electricity conversion factor mainly affect energy revenue and generated electricity, with limited influence on facility selection. The MBT-to-WtE fraction is more influential than the electricity conversion factor because it changes the quantity of material sent to energy recovery rather than only changing the energy output per ton. This explains why MBT-to-WtE allocation has a stronger effect on both emissions and energy recovery. Carbon policies In this section, we have applied three goals, implementation feasibility risk, site environmental risk and total cost. Simultaneously, we monetized the carbon inside the cost goal. Policy 1: carbon tax. In Policy 1, CO₂ is not treated as a separate optimization goal. Instead, emissions are internalized by applying a carbon tax and adding the resulting carbon cost to the total cost objective. The model therefore assumes the municipality pays a tax for each ton of CO₂ emitted, while the corresponding CO₂ quantities are reported as performance indicators. we solved Policy 1 (carbon tax) three times with different carbon prices: 55, 60, 65 $/tCO₂. we solved Policy 1 (carbon tax) three times with different carbon prices: 55, 60, 65 $/tCO₂. In each run, you used 3 goals only (Goal Programming): Operational risk (goal).Environmental risk (goal).Total cost (goal) and this total cost already includes carbon cost. In each run, you used 3 goals only (Goal Programming): Operational risk (goal).Environmental risk (goal).Total cost (goal) and this total cost already includes carbon cost. Operational risk (goal). Operational risk (goal). Environmental risk (goal). Environmental risk (goal). Total cost (goal) and this total cost already includes carbon cost. Total cost (goal) and this total cost already includes carbon cost. Carbon is not a goal in this experiment. It is reported as an output indicator. Carbon is not a goal in this experiment. It is reported as an output indicator. So carbon affects the solution indirectly through the cost goal, because the cost term contains . Table 10 shows what’s happening when carbon price increases From 55 → 60 → 65: Table 10Policy 1 — Carbon tax (carbon monetized inside the cost goal).Model/policyCarbon price (USD/tCO₂)Goal: TotalCost (USD/year)ΔCost vs. WGPReported: CO₂ (tCO₂/year)ΔCO₂ vs. WGPGoal:IFRisk (index)Goal:SERisk (index)E (MWh/year)Four goals (Cost, CO₂, IFRisk, SERisk)—216,812,789-505,995-22.318.9141,127.8Three goals (Policy 1): CO₂ monetized in cost55247,651,981+ 30.84 M USD508,654+ 2,65921.118.8141,127.8Three goals (Policy 1): CO₂ monetized in cost60250,243,443+ 33.43 M USD508,696+ 2,70121.118.8141,127.8Three goals (Policy 1): CO₂ monetized in cost65246,982,268+ 30.17 M USD503,613−2,38221.819.0141,127.8 Policy 1 — Carbon tax (carbon monetized inside the cost goal). Under Policy 1 (carbon-tax internalization), the objective penalizes emissions through an added term . As a result, the reported total cost increases relative to the four-goal formulation because it now includes explicit carbon payments (≈ 28–33 MUSD/year). Under Policy 1 (carbon-tax internalization), the objective penalizes emissions through an added term . As a result, the reported total cost increases relative to the four-goal formulation because it now includes explicit carbon payments (≈ 28–33 MUSD/year). Changes in CO₂ across carbon-price levels are modest, reflecting that the mass-flow splits at MBT and the WtE feed are largely governed by fixed yield constraints; therefore, carbon pricing primarily alters the monetized cost while inducing only marginal adjustments in routing and facility assignment. The facility-opening pattern remains stable across scenarios. Changes in CO₂ across carbon-price levels are modest, reflecting that the mass-flow splits at MBT and the WtE feed are largely governed by fixed yield constraints; therefore, carbon pricing primarily alters the monetized cost while inducing only marginal adjustments in routing and facility assignment. The facility-opening pattern remains stable across scenarios. Policy 2: Processing subsidy. In this policy also, CO₂ is not treated as a separate optimization goal. Instead, emissions are internalized by applying a carbon tax and adding the corresponding carbon cost to the total cost objective, while CO₂ is reported as a performance indicator. In addition, the model assumes a government subsidy that reduces effective processing costs, provided as a proportion of the waste treated through transfer, Treating (MBT) and WtE recycling activities. Table 11 compares the four-goal formulation with Policy 2, where CO₂ is internalized through a carbon tax and processing is subsidized at transfer, MBT, and WtE. Across all subsidy levels (20–30%), the selected infrastructure configuration remains unchanged (3 transfer stations, 2 MBTs, 1 WtE, and 2 landfills), and electricity generation remains constant (≈ 141,128 MWh/year). Table 11Policy comparison: Four-goal weighted GP vs. Policy 2 (carbon tax embedded + processing subsidy).Model/PolicySubsidy rateGoal:TotalCost (USD/year)ΔCost vs. WGPReported:CO₂ (tCO₂/year)Goal:IFRisk (index)ΔCO₂ vs. WGPGoal:SERisk (index)E (MWh/year)Four goals (Cost, CO₂, IFRisk, SERisk)—216,812,789505,995.122.318.9141,127.8Three goals (Policy 2): carbon tax + subsidy on T/M/W20%238,097,079.7+ 21.28 M503,598.622.1−2,396.518.5141,127.8Three goals (Policy 2): carbon tax + subsidy on T/M/W25%243,014,361.9+ 26.20 M509,320.122.9+ 3,325.017.9141,127.8Three goals (Policy 2): carbon tax + subsidy on T/M/W30%241,404,634.8+ 24.59 M509,300.721.6+ 3,305.618.9141,127.8 Policy comparison: Four-goal weighted GP vs. Policy 2 (carbon tax embedded + processing subsidy). The reported TotalCost is higher under Policy 2 than under the four-goal case because TotalCost now includes the explicit carbon payment term . The subsidy reduces the operating costs of transfer, MBT and WtE, but the carbon-cost component is still high because emissions remain high. So the annual net cost is still above the four-goal benchmark. CO 2 changes slightly over the subsidy levels, indicating the tested subsidy range is not strong enough to move waste to different processing routes. Table 9 shows that Policy 2 changes the economic outcome more clearly than the physical network configuration. The subsidy reduces part of the treatment cost, but total cost remains above the WGP baseline because the carbon tax is still included. The 20% subsidy case gives the lowest CO₂ value, although the difference is modest. The selected infrastructure is unchanged in all subsidy cases, with three transfer stations, two MBT facilities, one WtE facility, and two landfills. In the tested range, the subsidy changes the financial result more than the physical network or waste-flow pattern. Policy 3: Carbon cap-and-trade. Under Policy 3, carbon is regulated through a cap-and-trade mechanism, rather than through a simple tax or a hard cap alone. An emissions allowance is allocated to the system, and the operator may buy additional permits if actual emissions exceed the allowance, or sell part of the surplus if emissions remain below the cap. Additional parameters. : carbon market price (USD/tCO2). : carbon market price (USD/tCO2). : allocated allowance (tCO2). : allocated allowance (tCO2). : fraction of surplus allowances allowed to be sold. : fraction of surplus allowances allowed to be sold. Additional decision variables. : permits purchased (tCO2). : permits purchased (tCO2). : permits sold (tCO2). : permits sold (tCO2). : remaining allowances after compliance. : remaining allowances after compliance. : surplus allowance. : surplus allowance. : deficit allowance. : deficit allowance. These variables satisfy: The net carbon trading cost is defined as: Thus, may be positive when permits must be purchased, or negative when surplus permits are sold. This policy complemented the previous two scenarios. Policy 1 internalizes emissions through a carbon tax, Policy 2 promotes processing through subsidies, while Policy 3 introduces a market-based carbon mechanism. In practice, if emissions exceed the cap, the system incurs extra cost through permit purchases; if emissions remain below the cap, part of the surplus can be sold, reducing net system cost. Table 12 reports the results of Policy 3 (cap-and-trade) under different allocated carbon allowances. The results show the expected economic behavior of a functioning trading mechanism. When the cap is set below actual emissions (450,000 and 500,000 tCO₂), the system faces an allowance deficit and must purchase permits, generating a positive carbon trading cost of about 2.91 million USD and 0.32 million USD, respectively. In contrast, when the cap is set above actual emissions (550,000 tCO₂), the system generates a surplus of allowances, sells the permitted share, and records a negative carbon trading cost (i.e., net revenue) of about 2.04 million USD, while retaining a small unused balance. Table 12Policy 3 results under alternative carbon caps.PolicyCarbon cap tCO₂TotalCost USD/yearΔCost vs. WGPCO₂ tCO₂/yearΔCO₂ vs. WGPCap positionBuy tCO₂Sell tCO₂Unused tCO₂Carbon trading cost USDWGP baseline—216,812,789—505,995.1——————Policy 3: cap-and-trade450,000212,489,684−4.32 M508,128+ 2,133Deficit58,128——2,906,384Policy 3: cap-and-trade500,000208,001,889−8.81 M506,364+ 369Deficit6,364——318,188Policy 3: cap-and-trade550,000203,937,629−12.88 M504,775−1,220Surplus—40,7024,522−2,035,124 Policy 3 results under alternative carbon caps. Tighter caps increase the net cost to the system in financial terms because the model has to buy permits. The reverse is true for a looser cap : the excess permits can be sold and the net cost is reduced. Table 12 shows how the cap-and-trade setting changes the financial result. When the cap is set at 450,000 or 500,000 tCO₂, emissions are above the allowance, so the model must buy permits. When the cap is raised to 550,000 tCO₂, emissions are below the allowance and part of the surplus can be sold. Emissions do not change much compared with the WGP baseline because the physical network is still constrained by the fixed MBT-to-WtE fraction, mandatory WtE participation, and the available candidate sites. The main change is therefore in the carbon-trading balance, not in the waste-flow structure. Interpretation of carbon-policy responsiveness The three carbon-policy scenarios enter the model in different ways. The carbon tax prices each ton of CO₂ and therefore mainly increases the reported system cost. The processing subsidy reduces selected treatment costs, but it does not change feasible routes, facility capacities, or the fixed MBT-to-WtE allocation. The cap-and-trade case changes the financial balance according to whether emissions exceed or fall below the allocated cap. In all three cases, the policy effect is mainly financial unless the policy setting becomes strict enough to change the feasible network. The limited change in facility selection is therefore linked to the model structure. The baseline network already follows a recovery-oriented route through transfer stations and MBT facilities, requires at least one WtE facility, uses fixed recovery and MBT-to-WtE fractions, and is limited to the available candidate sites. Under these conditions, moderate carbon prices or subsidies mainly change the economic accounting of the selected network. Larger structural changes would be expected when a policy lever directly changes the feasible region or makes the current configuration economically unattractive, for example through a stricter emissions cap, a landfill-disposal cap, a higher landfill tax, a higher required WtE/recovery share, stronger subsidies for low-emission treatment, or additional candidate sites with lower carbon and risk profiles. Table 13 summarizes these policy levers and their expected model responses. Table 13Carbon-policy levers and expected model responses.Policy leverHow it enters the modelMain expected responseWhen it can change the network structureCarbon taxAdds to the cost expressionIncreases monetized system cost and may favor lower-emission optionsWhen the carbon price is high enough to offset the fixed-cost and transport-cost advantages of the current networkProcessing subsidyReduces effective processing costs at selected facilitiesImproves the economics of subsidized treatment routesWhen the subsidy is large enough to make alternative MBT/WtE configurations cheaperCap-and-tradeAdds permit-buying cost or permit-selling revenue based on the emissions capCreates compliance cost under tight caps or trading revenue under loose capsWhen the cap is strict enough to force lower-emission flows or facility choicesLandfill cap or landfill taxLimits or penalizes disposal quantitiesPushes waste away from landfill and toward recovery or WtEWhen landfill disposal becomes capacity-limited or financially unattractiveMinimum WtE/recovery shareRequires a minimum share of waste to enter WtE or recoveryIncreases recovery or energy-generation pressureWhen the required share exceeds the baseline MBT-to-WtE allocationAdditional low-carbon sitesExpands the candidate facility setGives the model more alternativesWhen new candidates have better cost, emission, or risk profiles Carbon-policy levers and expected model responses. Discussion The results show that the proposed framework is most useful as a decision-support tool for comparing feasible recovery pathways under constrained rebuilding conditions, rather than as a tool for identifying a single technically optimal waste-management design. In the Idlib case, the weighted goal-programming solution selected a stable core network of three transfer stations, two MBT facilities, one WtE plant, and two landfills. The Idlib results show why a compromise model is needed in this type of planning problem. The results show that the compromise solution is not simply a middle point between objectives. It avoids some of the sharper trade-offs that appeared in the single-objective runs. A cheaper or slightly lower-carbon solution may look attractive in numerical terms, but it can be less practical if it requires more facilities, higher-risk sites, or routes that are difficult to operate under post-conflict conditions. The epsilon-constraint cases reinforce this point. When the carbon and risk limits were moderate, the model found lower-cost alternatives. When the risk limits became too strict, the model was infeasible under the available sites, capacities, routing assumptions, and WtE requirement. This is useful from a planning perspective because it separates two issues: choosing a balanced solution based on expert preferences, and testing whether specific policy thresholds can actually be met. The sensitivity analysis also suggests that the selected network has a stable core. Moderate changes in the tested parameters did not usually alter the facility pattern, although they did affect cost, emissions, and energy output. Waste supply had the widest system effect because it changes the volume moving through every stage of the network. The MBT-to-WtE fraction was also important, since it controls how much residual waste moves toward energy recovery instead of landfill disposal. Electricity conversion, by contrast, mainly changed revenue and energy output. The carbon-policy results point in the same direction. In the tested scenarios, carbon tax, subsidy, and cap-and-trade changed the financial side of the model more than the physical network. To produce a stronger structural response, the policy would likely need to be more binding, for example through stricter emission caps, landfill-disposal limits, higher recovery or WtE requirements, stronger subsidies, or additional candidate sites with better carbon and risk profiles. For planners working in data-scarce recovery settings, the framework is useful because it does three things at once. It identifies a workable network, highlights the assumptions that need better data, and separates policies that change accounting results from policies that could actually reshape the waste system. In a post-conflict city, this supports a staged approach: establish a basic transfer–MBT–landfill/WtE backbone first, then improve the system as better data, stronger institutions, and more effective policy tools become available. Conclusion This study presented a policy-driven MILP model for designing an MSW recovery network in a post-conflict setting. The model links facility-location and waste-flow decisions across transfer stations, MBT facilities, WtE plants, and landfills. It also brings cost, carbon emissions, implementation feasibility, and site environmental vulnerability into the same planning problem. This was necessary in the Idlib case, where the choice of a waste-management network is shaped not only by technical and economic factors, but also by limited data, access constraints, uncertain operating conditions, and policy requirements. The case-study results selected a recovery-oriented network with three transfer stations, two MBT facilities, one WtE plant, and two landfills. The single-objective runs showed that improving one objective can weaken another. Lower cost or lower emissions, for example, may come with higher implementation or environmental risk. The epsilon-constraint results led to a similar conclusion. Moderate carbon and risk limits were feasible, but stricter risk limits could not be met with the available candidate sites, capacities, routing structure, and WtE requirement. The sensitivity results showed that waste supply and the MBT-to-WtE fraction are the most influential drivers of system performance. The carbon-policy scenarios also showed an important distinction between financial effects and structural network changes. Carbon tax, processing subsidy, and cap-and-trade changed the cost structure and carbon-related payments, but they did not substantially change the selected infrastructure. Larger structural changes would likely require stronger levers, such as stricter carbon caps, landfill-disposal limits, higher recovery requirements, stronger subsidies, or additional candidate sites with better carbon and risk profiles. The model has some limitations. It is a single-period strategic planning model and does not include daily vehicle routing, seasonal waste variation, or phased facility expansion. Some technical and economic parameters were also estimated under data-scarce conditions and should be updated when better local data become available. Future work could extend the model to multi-period investment planning, uncertain waste generation and road access, more detailed life-cycle emissions, and social-acceptability indicators. Despite these limitations, the framework provides a practical planning structure for cities that need to rebuild MSW services under constrained data, infrastructure, and governance conditions. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Author contributions Conceptualization, (J.H.). and (A.H.A.H.); methodology, (J.H.).; software, (A.H.A.H.); validation, (A.H.A.H.); formal analysis, (J.H.); investigation, (A.H.A.H.).; resources, (A.H.A.H.).; data curation, (A.H.A.H.); writing—original draft preparation, (J.H.).; writing—review and editing, (A.H.A.H.).; visualization, (J.H.).; supervision, (J.H.).; project administration, (A.H.A.H.). funding acquisition, (A.H.A.H.)All authors have read and agreed to the published version of the manuscript. Funding This work was supported by the Deanship of Scientific Research, Vice Presidency for Graduate Studies and Scientific Research, King Faisal University, Saudi Arabia, under Grant KFU263261. Data availability Data can be shared upon reasonable request. Declarations Competing interests The authors declare no competing interests. Ethics and informed consent This study used anonymous expert elicitation for research purposes and did not involve patients, clinical records, biological materials, photographs, or any identifiable personal data. Prior to participation, all experts were informed about the purpose of the study, the voluntary nature of participation, their right to withdraw at any time, and the anonymous and confidential treatment of their responses. Access to the questionnaire was conditional on explicit agreement to participate. Thus, informed consent was obtained from all participants. Since no identifying information or images are included in this manuscript, consent for publication of identifying information was not applicable. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Jamil Hallak, Email: jamil.hallak@hku.edu.tr. A.H. Abdul Hafez, Email: aabdulhafiz@kfu.edu.sa References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement Data can be shared upon reasonable request.