An Integrative Decision-Making Framework for Sustainable Urban Water Governance: The Case of Yerevan City

Abstract

Sustainable urban water governance in rapidly transforming cities requires integrative decision-making frameworks capable of balancing social equity, economic efficiency, and environmental resilience. This study develops an Integrative Decision-Making Framework (IDMF) for optimizing urban water policy in Yerevan, Armenia, built upon AI- and GIS-assisted diagnostics and incorporating a Governance Readiness Index (GRI) together with spatial hotspot overlay analysis. The framework employs an AHP–TOPSIS multi-criteria structure to evaluate five policy alternatives—leakage reduction, demand-side management, decentralized reuse, green–blue infrastructure, and data-driven governance—based on normalized quantitative indicators across social, economic, and ecological domains. Results show that Leakage Reduction (A1) and Data-Driven Governance (A5) consistently rank as the top-performing strategies across both baseline and sensitivity scenarios, while equity-prioritized weightings enhance social outcomes without significantly compromising economic performance. The approach also demonstrates robustness under ±10–20% weight variations. Acknowledging limitations related to data availability and expert-based judgments, the study outlines the minimum governance and data-readiness conditions required for transferability. The IDMF thus advances decision-support science in urban water management by integrating governance feasibility with spatial diagnostics and provides adaptable guidance for mid-income cities facing institutional and environmental constraints.

1. Introduction

Sustainable urban water governance has become one of the most pressing challenges of the twenty-first century, particularly in rapidly transforming cities of the Global South and post-Soviet regions. Accelerated urbanization, uneven infrastructure development, and increasing climate pressures have intensified the need for integrative planning models that can balance environmental resilience, economic efficiency, and social equity [1,2,3]. The United Nations’ Sustainable Development Goals, especially SDG 6 (Clean Water and Sanitation) and SDG 11 (Sustainable Cities and Communities), emphasize the necessity of data-driven, participatory, and adaptive governance approaches [4,5,6]. Despite global progress, many medium-income cities continue to face institutional fragmentation, data limitations, and sectoral trade-offs that hinder the implementation of sustainable water policies [7,8].

Decision-making in urban water management represents a complex multi-objective problem involving diverse stakeholders, uncertain information, and interlinked hydrological, economic, and social subsystems [9,10,11]. Traditional top-down planning methods often fail to capture these interactions, leading to inefficiencies and inequitable outcomes [12]. To overcome these limitations, scholars increasingly advocate the use of multi-criteria decision-making (MCDM) and systems-thinking approaches that integrate quantitative assessment with stakeholder participation [13,14,15]. Techniques such as the Analytic Hierarchy Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) enable planners to identify optimal strategies across competing objectives while maintaining transparency and traceability of results [16,17,18,19]. These approaches are particularly valuable in urban contexts where resource scarcity, institutional inertia, and socio-environmental tensions coexist [20,21,22].

Recent advances in artificial intelligence (AI), geospatial analytics, and data integration frameworks have expanded the methodological toolkit available for sustainable urban planning [23,24,25,26]. Machine-learning-based spatial diagnostics, Internet-of-Things (IoT) sensors, and remote-sensing indicators now allow real-time monitoring of water availability, quality, and consumption [27,28,29]. However, the translation of such diagnostics into actionable governance decisions remains a critical gap in both theory and practice [30,31]. Bridging this gap requires integrative frameworks that link empirical data with value-based decision criteria—embedding social, economic, and ecological dimensions into a unified policy model [32,33,34].

In this regard, Yerevan City provides a particularly relevant context for investigating integrative decision-making. As a mid-income capital with rapid demographic growth and outdated water infrastructure, Yerevan faces increasing stress on both surface and groundwater systems [35,36,37]. Recent studies have documented rising non-revenue water losses, uneven service accessibility, and declining per capita green space [38,39,40]. Building on our earlier research integrating AI- and GIS-based diagnostics of urban sustainability [41], the present study moves beyond spatial assessment toward decision optimization, developing a multi-criteria framework that evaluates and ranks alternative policy options for sustainable water management in Yerevan. By combining environmental performance metrics, economic cost indicators, and social participation indices within an MCDM structure, the proposed model provides a replicable pathway for evidence-based and equity-oriented water governance.

From an institutional perspective, Yerevan’s urban water governance is characterized by a multi-actor setting in which responsibilities are fragmented across national and municipal levels. A single regulated utility operates the bulk water supply and distribution network under long-term contractual arrangements with the central government, while tariff setting and service quality oversight fall under an independent regulatory authority. At the same time, the Ministry of Environment is responsible for basin-level planning and ecological standards, whereas the Yerevan Municipality oversees land-use planning, green space maintenance, and local infrastructure projects. This division of mandates often results in coordination gaps, delayed implementation of investment programs, and limited integration between water supply, drainage, and urban greening policies. Recent reform efforts have focused on reducing non-revenue water and upgrading aging infrastructure, but governance bottlenecks—such as overlapping competencies, constrained municipal budgets, and weak participatory mechanisms—continue to hinder the transition towards more adaptive and accountable water governance in Yerevan.

However, the scope of this study is constrained by several contextual and methodological limitations. First, the evaluation relies on available municipal, hydrometeorological, and utility datasets, which, despite triangulation and quality checks, remain subject to temporal gaps and reporting inconsistencies typical of mid-income urban systems. Second, the AHP–TOPSIS model captures relative performance across selected sustainability indicators but does not replace detailed engineering simulations, hydrological forecasting, or cost–recovery modeling. Third, the governance feasibility dimensions—operationalized through the GRI—reflect institutional conditions specific to Yerevan and should not be generalized without contextual adaptation. Fourth, GIS-derived ecological indicators represent a snapshot of 2018–2024 conditions and may shift under longer-term climatic or land-use changes. These limitations define the analytical boundaries of the framework, which is intended as a decision-support tool under informational constraints rather than a comprehensive infrastructure design model.

The study contributes to the literature in three major ways. First, it operationalizes the decision-making dimension of sustainable urban water governance by linking empirical data with structured evaluation models. Second, it demonstrates how multi-criteria analysis can be localized to the institutional realities of post-Soviet cities, integrating governance readiness and technical feasibility factors. Third, it extends the ongoing dialog on adaptive urban governance, offering an analytical framework transferable to other emerging urban systems. Overall, this work advances the scientific understanding of how integrative decision-making can enhance urban resilience, resource efficiency, and social inclusion in the context of water management.

2. Literature Review

The management of urban water resources has long evolved from a purely technical domain toward an interdisciplinary field that integrates environmental, economic, and social dimensions of decision-making [1,2,3]. Early approaches to water planning were largely grounded in hydraulic engineering and cost–benefit analyses, prioritizing physical efficiency over governance or equity considerations. However, the paradigm began shifting in the late 20th century with the emergence of Integrated Water Resources Management (IWRM) and sustainable urban water systems (SUWM) frameworks, which emphasized participatory governance, demand management, and ecosystem services [4,5,6]. These frameworks recognized that water systems are socio-ecological complexes where institutional and behavioral factors are as critical as technical ones.

Recent scholarship has underlined the need for decision-support models that can operationalize sustainability principles within this integrated context [7,8,9,10]. Such models should allow stakeholders to weigh trade-offs among competing objectives—equity, efficiency, and environmental integrity—while accounting for uncertainty and data limitations. The Multi-Criteria Decision-Making (MCDM) family of methods has become particularly influential in this regard. Techniques such as the Analytic Hierarchy Process (AHP), Analytic Network Process (ANP), and TOPSIS have been widely adopted in environmental and infrastructure planning to evaluate alternatives based on quantitative and qualitative indicators [11,12,13,14,15]. Scholars have demonstrated their capacity to enhance transparency and stakeholder engagement by translating complex technical criteria into structured comparative judgments [16,17].

Parallel to methodological advancements, the literature on urban sustainability and governance has emphasized adaptive, data-informed, and participatory approaches. The concept of adaptive governance—initially developed in environmental policy—has been extended to urban water systems to capture the dynamic feedbacks between ecological change, technological innovation, and institutional learning [18,19,20,21]. Governance studies show that integrated decision frameworks can strengthen accountability and resilience, particularly when coupled with digital tools and open data infrastructures [22,23,24]. Yet, despite the theoretical progress, empirical implementation in medium-income or post-Soviet cities remains limited, primarily due to institutional inertia, fragmented responsibilities, and low public participation [25,26,27].

Another significant body of research addresses the integration of artificial intelligence (AI), geospatial analytics, and real-time data systems in water management. These technologies enable predictive modeling of consumption patterns, detection of system losses, and simulation of policy impacts under multiple scenarios [28,29,30,31]. For instance, machine-learning algorithms and remote-sensing data have been used to monitor drought stress, map groundwater recharge, and assess spatial inequities in service delivery [32,33,34]. Nevertheless, most AI- and GIS-based studies remain diagnostic in nature—they reveal problems rather than guiding structured decisions. Bridging this diagnostic–decision gap represents a key research frontier.

To address this gap, emerging literature advocates for hybrid frameworks that combine data-driven analytics with normative decision theory [35,36,37,38]. Integrating MCDM with spatial modeling and governance assessment allows for multi-layered evaluation of policies where social, economic, and ecological criteria interact dynamically [39,40]. Examples include coupling AHP with GIS for watershed prioritization, applying fuzzy TOPSIS for climate adaptation planning, or integrating Delphi-based expert weighting with sustainability indices [41,42,43,44]. These approaches illustrate the feasibility of quantitatively expressing stakeholder preferences while maintaining contextual sensitivity to institutional and cultural settings.

The INFFEWS (Integrated Food–Energy–Water Systems) framework also provides a relevant example of urban water-sensitive design, emphasizing cross-sectoral interactions, co-benefit valuation, and scenario-based planning. Its emphasis on integrated sustainability assessment aligns closely with our governance-sensitive MCDM approach, highlighting the importance of coupling environmental performance with institutional feasibility [45].

Recent applications of TOPSIS and its fuzzy extensions have demonstrated their effectiveness in sustainability-oriented water resource planning. For example, fuzzy TOPSIS has been used to evaluate drought mitigation strategies, prioritize watershed management options, and rank integrated water supply alternatives under uncertainty. Such studies highlight the suitability of TOPSIS-type methods for balancing environmental, economic, and social criteria when data are imprecise or partially subjective, reinforcing the relevance of our AHP–TOPSIS approach for urban water governance [46,47].

An important contribution relevant to groundwater vulnerability assessment is the integration of GIS with AHP in DRASTIC-based zoning, which demonstrates how multi-criteria spatial modeling can identify groundwater risk hotspots under data-scarce conditions [48].

While a substantial body of research has applied AHP, ANP, TOPSIS, and their fuzzy variants to water resource planning, most existing studies operationalize MCDM in a narrowly technical manner—prioritizing alternatives based solely on hydraulic performance, water allocation efficiency, or drought mitigation metrics. For instance, TOPSIS-based evaluations in hydrological modeling, watershed prioritization, and urban flood resilience generally treat governance parameters as exogenous conditions rather than internal components of the decision structure. Similarly, AHP applications commonly focus on engineering criteria or environmental indicators without explicitly integrating institutional feasibility or socio-spatial inequalities. As a result, the majority of MCDM water studies remain limited to analytic optimization rather than governance-oriented decision integration.

However, most published applications are situated in high-income or data-rich contexts. Few studies have explored decision-making for sustainable water governance in transition economies, where uncertainty, data scarcity, and governance fragmentation prevail [45,46,47]. Post-Soviet urban systems, such as those of Armenia, Georgia, or Kazakhstan, exemplify this challenge. Rapid development, legacy infrastructure, and evolving legal frameworks require adaptable models capable of functioning under informational constraints [48,49]. Thus, there is a clear research need for integrative decision-making frameworks that are not only technically robust but also institutionally feasible and socially legitimate.

The present study advances this literature by positioning MCDM within a governance-sensitive and spatially explicit framework. Unlike conventional AHP–TOPSIS models, our approach integrates: (a) a Governance Readiness Index (GRI) that adjusts rankings based on regulatory maturity, coordination efficiency, and data transparency; (b) AI–GIS diagnostic layers that spatially anchor MCDM results in district-level ecological vulnerability and service accessibility; and (c) socio-economic equity indicators tailored to the institutional context of a post-Soviet, data-constrained urban environment. To our knowledge, no prior MCDM study in the urban water domain has jointly embedded governance feasibility, spatial diagnostics, and socio-economic equity into a unified decision-support architecture.

Building on this gap, the present study contributes to the literature by synthesizing insights from MCDM theory, adaptive governance, and sustainability science into a unified decision-making framework for urban water management. It extends prior AI–GIS diagnostic work on Yerevan City [50] toward a decision-oriented model capable of ranking and optimizing policy alternatives under competing objectives. This approach situates decision-making as both a quantitative evaluation and a governance process, thereby linking analytical rigor with participatory policy design.

3. Methodology

3.1. Conceptual Framework

The methodological foundation of this study rests on an integrative decision-making framework (DMF) designed to evaluate and optimize urban water governance under the triple dimensions of social equity, economic efficiency, and environmental sustainability. Building on the systems-thinking perspective [1], the DMF recognizes water governance as a complex socio-technical system where institutional capacities, financial constraints, and ecological risks interact dynamically. The framework synthesizes three analytical pillars:

1.

Diagnostic layer—This is based on spatial and institutional assessments from prior AI–GIS studies of Yerevan [2], which provide indicators of infrastructure condition, environmental pressure, and governance readiness.

2.

Data Sources, Temporal Coverage, and Quality Control—The empirical analysis relies on a consolidated dataset representing the social, economic, and environmental dimensions of Yerevan’s urban water system. The primary data sources include:

(a)

The Statistical Committee of Armenia (household income, service affordability, demographic indicators);

(b)

Yerevan Water Utility (non-revenue water levels, operational efficiency, service coverage),

(c)

The Hydrometeorological Service of Armenia (hydrological balance, supply variability, and climatic stress indicators).

The temporal coverage spans 2018–2024, capturing recent infrastructural rehabilitation efforts, demand fluctuations, and ecological trends. All quantitative variables were compiled using their standard measurement units (e.g., % for NRW, m³/year for water balance, mg/L for pollution metrics, GIS-based indices for ecological risk). To ensure data reliability, multiple quality control procedures were applied, including cross-agency validation (triangulating utility and hydrometric records), temporal smoothing of incomplete time series, and harmonization of inconsistent reporting formats. For indicators with missing values, linear interpolation and expert-verified estimation were performed following OECD data-quality guidelines. This multi-step procedure enhances the internal validity of the case study and ensures consistency across the integrated MCDM evaluation.

3.

Decision layer—employing multi-criteria decision-making (MCDM) techniques to structure, weight, and rank policy alternatives.

4.

Policy synthesis layer—integrating scenario-based evaluation and sensitivity testing to translate model outputs into actionable strategies for adaptive governance.

The framework’s logic flow is illustrated in Figure 1 (Conceptual Diagram), linking data inputs, criteria hierarchy, MCDM computation, and scenario outputs.

The diagram illustrates the logical structure of the proposed framework, linking the Diagnostic Layer (AI–GIS data and institutional baseline), the Decision Layer (MCDM AHP–TOPSIS weighting and ranking), and the Policy Synthesis Layer (scenario testing, governance feasibility, and policy recommendations).

Arrows indicate the directional flow of data and analytical results toward actionable governance strategies for Yerevan City.

This conceptual diagram illustrates the logic of the proposed decision-making framework, linking data inputs, criteria hierarchy, MCDM computation, and scenario synthesis.

It demonstrates how diagnostic information (AI–GIS data and institutional baseline) flows into the decision layer (AHP–TOPSIS weighting and ranking), and finally into the policy synthesis layer (scenario testing, governance feasibility, and recommendations).

3.2. Criteria Selection and Structure

Following the sustainability triangle approach, the evaluation criteria are organized into three dimensions: social, economic, and environmental. Each dimension includes specific indicators used to assess the performance of alternative policy strategies for Yerevan City. As summarized in

Table 1. Evaluation criteria for sustainable urban water governance within the proposed decision-making framework *.

Dimension	Criteria	Indicators	Data Source
Social	S1. Accessibility to safe water	% of population with 24 h service	Yerevan Water Utility reports
	S2. Affordability	Household water expenditure/income	NSS Armenia
	S3. Public participation	Frequency of community feedback mechanisms	Municipal surveys
Economic	E1. Operational efficiency	Non-revenue water %	Utility monitoring
	E2. Investment feasibility	Cost/benefit ratio of interventions	Engineering estimates
	E3. Infrastructure resilience	Average downtime/year	SCWS datasets
Environmental	ENV1. Ecological risk index	Derived from LST × NDVI layers	Remote sensing (Sentinel-2)
	ENV2. Water resource stability	Annual variability of supply vs. demand	Hydrological statistics
	ENV3. Pollution intensity	BOD/COD concentration in urban runoff	Environmental monitoring data

* Source: Author’s elaboration based on sustainability assessment methodology and empirical datasets (Mkhitaryan, 2025 [31]).

, all indicators are normalized on a 0–1 scale to ensure comparability across units. The environmental indicators were derived from medium- and high-resolution remote sensing datasets. NDVI values were extracted from Sentinel-2 Level-2A imagery at 10 m spatial resolution. To minimize cloud contamination and seasonal noise, monthly cloud-free composites were generated and aggregated into annual median NDVI values for 2018–2024. LST layers were obtained from Landsat-8 and Landsat-9 TIRS observations at 30 m resolution; summer-season composites (June–August) were computed to capture peak thermal stress relevant to urban ecological vulnerability. All raster datasets were resampled to a common 30 m grid, atmospherically corrected, co-registered, and normalized prior to calculating the Ecological Risk Index (ERI). The ERI was computed by combining z-score–standardized LST and NDVI layers, where higher LST and lower NDVI jointly indicate elevated ecological stress.

All indicators were normalized to a 0–1 scale using linear transformation (benefit/cost criteria differentiation) to ensure comparability across units [5].

These formulas follow standard MCDM normalization procedures for benefit- and cost-type indicators.

No additional mathematical transformations were applied to the raw indicators beyond the min–max scaling procedures described above. Specifically, the analysis did not employ logarithmic transformations, z-score standardization, winsorization, or temporal smoothing, as these procedures would have altered the original distributional properties of the administrative and environmental datasets. All indicators were assessed for outliers using IQR-based thresholds and histogram inspection; however, no values exceeded exclusion limits. Thus, the normalized decision matrix is directly derived from the raw inputs following linear transformation only.

r_{ij} = \frac{\max(x_j) − x_{ij}}{\max(x_j) − \min(x_j)}

r_{ij} = \frac{x_{ij} − \min(x_j)}{\max(x_j) − \min(x_j)}

These transformations ensure that all criteria enter the decision matrix on a consistent 0–1 scale, preserving relative variation across heterogeneous units such as percentages, currency ratios, downtime hours, and raster-derived indices. No logarithmic or z-score transformations were required after outlier screening.

To enhance methodological transparency, each of the nine indicators was operationalized using standardized measurement procedures. Service accessibility was computed as the district-level share of households receiving continuous (24 h) supply based on monthly utility records. Affordability was defined as the ratio of mean quarterly household water expenditure to disposable income using NSS socio-economic datasets. Public participation was quantified through the annual frequency of formal consultation events and community feedback submissions aggregated from municipal open-government channels. Operational efficiency (NRW) relied on audited utility loss reports, while investment feasibility was assessed using engineering cost–benefit ratios derived from feasibility studies. Infrastructure resilience captured the annual average downtime hours per service zone. The Ecological Risk Index (ERI) combined normalized LST and NDVI values extracted from Sentinel-2 raster data. Water resource stability reflected annual supply–demand variability from hydrometeorological archives, and pollution intensity was calculated using BOD/COD concentrations from certified environmental monitoring stations.

3.3. AHP–TOPSIS Weighting and Evaluation

To evaluate and prioritize alternative policy strategies, this study adopts an enhanced AHP–TOPSIS integration specifically tailored to the institutional and spatial realities of Yerevan’s water governance system. Unlike generic applications of AHP–TOPSIS in resource management, the present framework embeds the weighting and ranking procedures into a governance-sensitive structure, allowing social equity, economic feasibility, and ecological vulnerability indicators to interact dynamically.

In this context, expert judgments were elicited not only to capture relative indicator importance but also to reflect the governance bottlenecks characteristic of mid-income, post-Soviet cities. This resulted in pairwise comparison matrices that embody contextual asymmetries across social participation, affordability, institutional coordination, and environmental stress. The subsequent TOPSIS evaluation incorporates Yerevan’s spatial risk layers, enabling policy alternatives to be ranked in terms of both multidimensional performance and district-level vulnerability.

Step 1: AHP weighting.

Pairwise comparison matrices were developed based on expert judgments from 12 specialists (university researchers, water utility engineers, and municipal planners).

The panel of 12 experts was intentionally composed to ensure disciplinary diversity and institutional representation. Four experts were senior utility engineers specializing in distribution networks and NRW diagnostics. Three were environmental and GIS specialists with experience in ecological risk mapping and spatial analytics. Three participants were municipal water governance officers with direct responsibility for regulatory oversight, service monitoring, and public engagement processes. The remaining two experts were academic researchers in sustainability policy and decision-support modeling. All experts possessed 7–20 years of domain experience and were actively involved in water management or urban infrastructure planning, thereby ensuring informed and context-sensitive pairwise judgments.

Consistency of all AHP pairwise comparison matrices was assessed using Saaty’s Consistency Ratio (CR). A threshold of CR ≤ 0.10 was applied following standard methodological practice. Matrices with acceptable consistency were retained without modification. When CR exceeded 0.10, the corresponding expert was contacted and asked to revisit the inconsistent judgments. This iterative refinement process continued until all matrices satisfied the consistency requirement. Consequently, no matrix with CR ≥ 0.10 was included in the final weighting calculations.

Inter-expert variability was addressed through the standard AHP aggregation procedure. After each expert’s pairwise comparison matrix passed the CR ≤ 0.10 threshold, the individual matrices were combined using the geometric mean, which is the recommended method for multi-expert AHP synthesis. The geometric mean preserves proportional judgment structures and minimizes distortion caused by outlier values. Variability across experts was examined by comparing dispersion patterns in corresponding matrix entries; however, no expert judgments exhibited excessive divergence warranting exclusion. The resulting aggregated matrix was subsequently used to derive the final priority weights for all criteria.

The Saaty scale (1–9) was used to express relative importance among criteria. Consistency ratios (CR < 0.1) were verified for reliability [6]. Final weights reflected the average of individual assessments.

3.3.1. Construction of the AHP Judgment Matrix and Consistency Verification

To ensure methodological transparency, the construction of the AHP pairwise judgment matrix followed a structured and replicable procedure. Twelve experts—representing academic researchers, water utility engineers, and municipal water planners—were invited to independently assess the relative importance of the nine evaluation indicators. The assessments were based on Saaty’s 1–9 scale, where 1 denotes equal importance and 9 represents extreme importance.

Individual pairwise comparison matrices were first constructed separately for each expert. These matrices were then aggregated into a single group matrix using the geometric mean method, which is recommended for multi-expert synthesis in AHP. The aggregated matrix served as the basis for deriving the normalized weight vector through principal eigenvalue computation.

To evaluate the internal logical consistency of expert judgments, we computed the Consistency Index (CI), Random Index (RI), and Consistency Ratio (CR) for each comparison matrix. According to Saaty (2008) [11], a CR value below 0.10 indicates acceptable consistency. In our analysis, all CR values fell within this acceptable range (CR = 0.041 for the main criteria level; CR values ranging from 0.052 to 0.067 for sub-criteria), confirming that the expert judgments were coherent and statistically reliable.

Detailed matrix values, consistency calculations, and eigenvector normalizations are available from the authors upon reasonable request for reproducibility.

Step 2: TOPSIS ranking.

Using the normalized decision matrix, the weighted Euclidean distance of each policy alternative was calculated relative to the ideal (best) and negative-ideal (worst) solutions [7]. The closeness coefficient (C*) for each alternative was derived as:

$$\mathit{C}\_\mathit{i}^\mathit{\ast }=\backslash \mathit{frac}\{\mathit{D}\_\mathit{i}^-\}\{\mathit{D}\_\mathit{i}^++\mathit{D}\_\mathit{i}^-\}$$

where D_i⁺ and D_i⁻ denote distances to positive and negative ideal solutions, respectively.

Higher C_i* values indicate better performance.

Step 3: Scenario analysis.

Two weighting scenarios were tested:

• Balanced scenario—equal emphasis on social, economic, and environmental dimensions;
• Equity-prioritized scenario—40% social, 30% environmental, 30% economic weighting.

Scenario comparison allowed exploration of trade-offs and robustness.

A Monte Carlo simulation was performed to assess the robustness of the Consistency Ratio (CR) under stochastic perturbation of the pairwise comparison matrices. A preliminary pilot run of 500 iterations was used to estimate computational load; however, all final reported results are based on a 1000-iteration simulation, following established AHP robustness-testing practice. Increasing the number of iterations beyond 1000 (e.g., 2000 or 5000) did not produce significant changes in the empirical distribution of CR values, confirming that 1000 iterations were sufficient to ensure stability and reliability of the consistency assessment.

3.3.2. Sensitivity Analysis and Treatment of Data Uncertainty

To assess the robustness of the derived indicator weights and the stability of the final rankings, a sensitivity analysis was incorporated into the methodological framework. Two complementary procedures were applied.

First, a one-way deterministic sensitivity test was conducted by varying each criterion weight by ±10% and ±20% while holding the remaining weights constant. This allowed us to examine the degree to which individual dimensions (social, economic, environmental) influence the closeness coefficients and ranking outcomes.

Second, a multi-scenario perturbation analysis was performed by generating alternative weighting schemes beyond the balanced and equity-prioritized scenarios, including environmentally dominant and economically dominant configurations. Across these scenarios, the rank positions of the top alternatives demonstrated high stability, indicating that the model does not exhibit rank reversals under realistic fluctuations of expert-derived weights.

Given that indicator uncertainty may arise from measurement gaps, expert scoring variance, or seasonal variability in hydrological data, we also evaluated the potential propagation of these uncertainties into the MCDM results. Following OECD and JRC guidelines for composite indicator uncertainty assessment, we applied a Monte Carlo resampling of the weight vector (n = 500 iterations), producing randomized variations within the plausible intervals of expert judgment. The resulting distribution of rankings confirmed that A1 and A5 remain consistently within the top positions, thereby validating the robustness of the evaluation under stochastic uncertainty.

This combined deterministic–stochastic approach improves the methodological reliability of the framework and explicitly links uncertainty sources to their impacts on decision stability.

Step 4: Sensitivity analysis.

A global ±10% perturbation of criterion weights and a Monte Carlo 1000-iteration resampling were applied to test model stability [8]. Rank reversals were analyzed to assess the reliability of policy recommendations.

3.4. Policy Alternatives

Five strategic policy options were identified in consultation with Yerevan’s municipal water management department and academic experts:

• A1—Leakage reduction program (infrastructure rehabilitation and smart metering);
• A2—Demand-side management (progressive tariffs, awareness campaigns);
• A3—Decentralized water reuse (greywater recycling at district level);
• A4—Green-blue infrastructure expansion (urban wetlands and infiltration parks);
• A5—Data-driven governance system (digital twin and open-data dashboard).

Each alternative was evaluated against the nine criteria using both quantitative datasets and expert scoring (1–9 scale). The combined dataset formed the decision matrix for MCDM analysis.

3.5. Integration with Governance Feasibility

Recognizing that technically optimal solutions may not always be institutionally feasible, the model incorporates a Governance Readiness Index (GRI) adapted from previous AI–GIS work [9].

The Governance Readiness Index (GRI) was developed to evaluate the institutional feasibility and implementation capacity associated with each policy alternative. The index covers four governance dimensions commonly used in policy diagnostics: (1) regulatory clarity, (2) inter-agency coordination, (3) data transparency, and (4) implementation capacity. Each dimension was assessed using 3–5 qualitative–quantitative indicators rated by the same panel of 12 experts on a 0–4 ordinal scale.

All indicator scores were normalized to the 0–1 range using min–max scaling, after which a weighted additive model was applied. Weights for the governance dimensions were derived through an AHP sub-module, ensuring consistency (CR ≤ 0.10) and inter-expert agreement.

The final Governance Readiness Index for alternative k is computed as:

GGRI_k = w_1 g_{1k} + w_2 g_{2k} + w_3 g_{3k} + w_4 g_{4k}

GRI_k = w_1 g_1k + w_2 g_2k + w_3 g_3k + w_4 g_4k

where w_dare the AHP-derived weights of the four governance dimensions, and g_dkare their normalized scores for alternative k. Higher GRI values indicate stronger institutional readiness.

In the final stage, the GRI score of each alternative was used as a multiplicative adjustment to the TOPSIS performance score to incorporate governance feasibility into the ranking results.

The weighting scheme for the Governance Readiness Index (GRI) was established using an Analytic Hierarchy Process (AHP) sub-module. The four governance dimensions—regulatory clarity, inter-agency coordination, data transparency, and implementation capacity—were evaluated through pairwise comparison matrices completed by 12 experts. The geometric mean was used to aggregate individual matrices, and consistency was ensured by applying the CR ≤ 0.10 criterion. The resulting normalized weights for the four dimensions are denoted as w_1, w_2, w_3, and w_4, where ∑_(d = 1)⁴ w_d = 1. These weights reflect the relative governance importance assigned by experts and are directly used in the computation of the Governance Readiness Index.

To evaluate the sensitivity of the final ranking to the assumptions embedded in the GRI construction, two robustness checks were performed. First, the AHP-derived governance weights were perturbed by ±10% while maintaining the normalization constraint ∑w_d = 1. Second, an alternative scenario was constructed using equal weights across the four governance dimensions. In both robustness checks, the highest-ranked policy alternative remained unchanged, and the relative ordering of the middle-ranked options exhibited only marginal variation (rank shifts of at most one position). These findings indicate that the final TOPSIS ranking is robust to reasonable variations in GRI weights and expert judgments, demonstrating the stability of the decision-support model.

The Governance Readiness Index (GRI) is primarily expert-based in its construction. Although several governance indicators contain empirical components—such as the existence of regulatory documents, frequency of public reporting, or availability of municipal open-data platforms—the majority of the scoring reflects qualitative institutional assessments provided by the 12 experts. This approach aligns with governance diagnostics commonly used in data-limited and transition-economy settings, where many institutional processes are not captured through standardized quantitative metrics. Accordingly, the final GRI values are derived from normalized expert scores aggregated through the AHP-based weighting scheme.

All expert scores used in the governance indicators were originally elicited on a 0–4 ordinal scale. For comparability with other components of the model, these values were transformed into a 0–1 interval using the standard min–max normalization: s_”norm” = (s − 0)/(4 − 0). No additional scaling procedures, reclassification rules, or threshold adjustments were applied. This direct linear normalization ensures transparency and preserves the relative distances among expert judgments while maintaining methodological consistency with the normalization applied to quantitative indicators.

Detailed scoring rules were applied to each indicator included in the model. Governance indicators (e.g., regulatory clarity, coordination efficiency, implementation capacity) were evaluated by experts on a 0–4 ordinal scale according to predefined qualitative anchors. Environmental indicators—such as pollution intensity, ecological vulnerability, and groundwater stress—were derived from administrative datasets and geospatial layers; values were converted into categorical scores (0–4) based on percentile thresholds and spatial severity. Social indicators, including community feedback, equity of service access, and responsiveness to public complaints, were operationalized using municipal service data and scored through a structured rubric reflecting frequency, coverage, and quality of engagement. All indicator scores were subsequently transformed to a 0–1 scale through min–max normalization to ensure full comparability within the multi-criteria decision matrix.

All indicators included in the decision-support framework were derived from clearly documented and operationalized data sources. Environmental indicators were constructed from administrative datasets provided by the Yerevan Water Utility, including district-level non-revenue water statistics, outage frequency logs, and water quality compliance reports. These were supplemented with GIS layers on surface pollution hotspots, groundwater vulnerability zones, and spatial accessibility of green infrastructure. Social indicators were computed from municipal service datasets capturing complaint frequencies, community feedback reports, and district-level service accessibility. Governance indicators were derived from structured expert assessments collected through a standardized 0–4 scoring rubric. All datasets underwent a harmonization process, including unit standardization, missing-value screening, and spatial resampling for GIS layers. Following harmonization, values were transformed into comparable numerical indicators through min–max normalization, enabling their integration into the multi-criteria decision matrix.

3.6. Model Validation and Reproducibility

The methodological process follows MDPI’s reproducibility standards. All computations were executed in Python 3.11 using open-source libraries (numpy, pandas, scikit-criteria). The workflow and input datasets will be made publicly available through a Zenodo repository upon publication, ensuring transparency and replicability.

Diagnostic → Decision → Policy Synthesis | AI–GIS baseline → MCDM weighting → Scenario testing → Governance adjustment → Actionable recommendations.

4. Results

Integration of ERI and Accessibility Layers․ To construct the final hotspot surface, the Environmental Risk Index (ERI) and the Accessibility Index were integrated through a standardized GIS-based weighted overlay procedure. First, each input layer was normalized to the 0–1 interval using min–max scaling to ensure comparability. The ERI raster was generated by aggregating three environmental sub-layers—pollution intensity, ecological vulnerability, and groundwater stress—while the Accessibility Index captured the spatial distribution of service distance and district-level coverage. Following normalization, the two layers were aligned to the same spatial grid and combined using a cell-by-cell weighted sum (0.5 ERI + 0.5 Access). This composite raster highlights districts simultaneously exposed to high ecological stress and low accessibility.

Robustness of Hotspot Maps to Classification Thresholds․ To assess the robustness of hotspot delineation, multiple classification schemes were tested, including quantile, natural breaks (Jenks), and equal-interval thresholds. Across all schemes, core hotspot zones remained spatially stable, with only marginal boundary shifts. Quantile classification (used in the final results) provided the clearest discrimination of high-risk cells while preserving spatial consistency across methods.

4.1. Base Scenario (Balanced Weights)

Under the balanced weighting configuration—where social, economic, and environmental dimensions are assigned equal importance—the MCDM evaluation produces a coherent and transparent ranking of policy alternatives. The results reveal how water service accessibility, affordability, ecological-risk reduction, operational efficiency, and investment feasibility interact to shape overall sustainability performance.

The analysis indicates that A1 (Leakage reduction) and A5 (Data-driven governance) emerge as top-performing strategies. Both options demonstrate strong multi-dimensional performance by simultaneously improving service reliability, reducing non-revenue water losses, enhancing monitoring accuracy, and optimizing financial efficiency.

A2 (Demand-side management) and A3 (Decentralized reuse) perform well in environmental and social terms but are moderately constrained by cost-effectiveness due to higher capital requirements. A4 (Green-blue infrastructure expansion) achieves notable ecological benefits but obtains lower short-term feasibility and affordability scores.

These findings suggest that technologically advanced infrastructure rehabilitation and digital governance integration represent the most synergistic pathway for simultaneously achieving social equity, economic efficiency, and environmental sustainability within Yerevan’s urban water system.

The table presents the comparative results of the MCDM evaluation under two weighting configurations: the balanced scenario (equal weights for social, economic, and environmental dimensions) and the equity-prioritized scenario (40–30–30). Higher C* values indicate stronger sustainability performance (see

Table 2. Alternative Ranking—Balanced vs. Equity-Prioritized Scenarios *.

Policy Alternative	C* (Balanced)	Rank (Balanced)	C* (Equity)	Rank (Equity)
A1—Leakage reduction program	0.74	1	0.70	1
A5—Data-driven governance system	0.71	2	0.69	2
A2—Demand-side management	0.65	3	0.68	3
A3—Decentralized water reuse	0.59	4	0.57	5
A4—Green-blue infrastructure expansion	0.56	5	0.60	4

* Source: Author’s MCDM computations (AHP–TOPSIS) based on the proposed methodological framework (Mkhitaryan, 2025 [31]).

).

The outcomes in

Table 3. Governance-Adjusted Ranking of Policy Alternatives *.

Policy Alternative	Base Rank	GRI Score (0–1)	Adjusted Rank	Key Governance Remarks
A1—Leakage reduction	1	0.92	1	Strong technical and regulatory readiness
A5—Data-driven governance	2	0.88	2	Supported by digital infrastructure and open-data initiatives
A2—Demand-side management	3	0.84	3	Politically feasible; requires sustained community engagement
A4—Green-blue infrastructure	5	0.80	4	Compatible with municipal development plans
A3—Decentralized reuse	4	0.72	5	Limited regulatory support and high CAPEX burden

* Source: Author’s calculations based on Governance Readiness Index integration (Mkhitaryan, 2025 [31]).

are complemented by Figure 2, which visualizes the multi-criteria performance profiles of the leading alternatives․

The blue triangle represents the aggregated sustainability profile combining social, economic, and environmental scores derived from the AHP–TOPSIS evaluation. Risk interpretation categories are defined as follows:

• Low risk (0.0–0.25): minimal vulnerability and high feasibility.
• Medium-low risk (0.25–0.50): manageable risks under moderate governance support.
• Medium-high risk (0.50–0.75): significant challenges requiring targeted intervention.
• High risk (>0.75): limited feasibility and high susceptibility to system shocks.

This figure illustrates the comparative sustainability performance of the leading policy alternatives (A1—Leakage reduction, A2—Demand-side management, and A5—Data-driven governance) across the three dimensions of sustainability: social, economic, and environmental.

Each axis represents normalized scores (0–1) derived from the AHP–TOPSIS model.

The radar chart demonstrates that A1 and A5 exhibit balanced, high-level performance across all dimensions, while A2 achieves the highest social impact, reflecting its relevance for inclusive urban water management.

4.2. Governance-Adjusted Ranking

While the base MCDM results identify technically optimal alternatives, real-world implementation depends heavily on institutional capacity, regulatory coherence, and organizational readiness.

To bridge the gap between analytical outcomes and policy feasibility, the evaluation framework integrates a Governance Readiness Index (GRI).

The GRI functions as a feasibility multiplier, scaling the idealized performance of each option according to three governance attributes:

(1)

Legal and regulatory maturity;

(2)

Inter-agency coordination;

(3)

Data-transparency and monitoring capacity.

Adjusted Outcomes

After incorporating the GRI values (ranging from 0 to 1), the ranking remains largely consistent with the base scenario, though minor adjustments emerge:

❖

A1 (Leakage reduction) and A5 (Data-driven governance) retain the top two positions. Their technical feasibility is reinforced by existing utility-level expertise, digital monitoring systems, and clear regulatory frameworks.

❖

A2 (Demand-side management) remains stable, supported by strong policy legitimacy and moderate institutional demands.

❖

A3 (Decentralized reuse) experiences a slight decline due to high regulatory complexity and the absence of clear guidelines for gray-water treatment and reuse.

❖

A4 (Green-blue infrastructure) improves marginally, as it aligns with municipal planning instruments and external donor priorities.

These shifts demonstrate that governance maturity and implementation capacity significantly influence sustainability outcomes.

Even highly efficient or innovative solutions can be downgraded if they face administrative fragmentation or data-sharing constraints.

4.3. Analytical Interpretation of Empirical Results

The governance-adjusted analysis underscores that technical excellence alone is insufficient for sustainable policy adoption.

High-performing alternatives such as A1 and A5 succeed not merely because they are efficient but because they align with Yerevan’s existing institutional architecture and regulatory pathways.

Conversely, A3 illustrates the innovation paradox—conceptually strong yet institutionally fragile.

By embedding governance feasibility into the MCDM framework, the model provides a realistic, decision-oriented tool for prioritizing measures that are both impactful and implementable in a middle-income urban context.

4.4. Sensitivity and Robustness Analysis

To verify the stability of the MCDM results, a sensitivity analysis was conducted by systematically varying the weighting distribution of the three sustainability dimensions by ±10% and ±20%, as summarized in

Table 4. Sensitivity Summary and Rank Stability.

Scenario Type	Weight Variation	Rank Reversals	Most Stable Alternative(s)	Stability Score (%)
Balanced (Base)	0%	–	A1, A5	100
Social +10%	+10% S/−5% E, −5% Env	1	A1, A2	95
Economic +10%	+10% E/−5% S, −5% Env	0	A1, A5	97
Environmental +10%	+10% Env/−5% S, −5% E	1	A5, A4	93
Monte Carlo (n = 500)	Random ±20%	2	A1, A5	88

Source: Author’s sensitivity analysis using stochastic weight variation and Monte Carlo simulation (Mkhitaryan, 2025 [31]).

.

This procedure evaluates how rank positions respond to changes in the relative importance of the social (S), economic (E), and environmental (Env) criteria.

Additionally, a Monte Carlo simulation (n = 500 runs) was employed to generate random weight combinations and assess the overall robustness of the ranking outcomes.

Key Findings

The analysis reveals high robustness in the top-performing alternatives.

• A1 (Leakage reduction) and A5 (Data-driven governance) maintain their leading positions in over 85% of all simulation runs, even under strong perturbations of weighting priorities.
• A2 (Demand-side management) occasionally rises to second place under socially dominant scenarios but rarely displaces the top two.
• A3 (Decentralized reuse) and A4 (Green-blue infrastructure) display moderate sensitivity, suggesting dependency on contextual weighting schemes and external funding feasibility.

These results confirm that the integrated AHP–TOPSIS framework remains methodologically stable and decision-resilient, with minimal rank reversals under realistic uncertainty intervals.

The table summarizes the results of the sensitivity and robustness analysis, demonstrating the stability of alternative rankings under varying weighting configurations. Weight variations of ±10% and ±20%, as well as Monte Carlo simulations (n = 500), were conducted to evaluate rank reversals and identify the most stable policy alternatives.

As illustrated in Figure 3, this figure displays the rank stability index (RSI) for each policy alternative across multiple weighting configurations.

Bars represent the percentage of simulation runs in which each alternative remained within the top-three ranking positions.

The figure demonstrates that A1 and A5 exhibit consistently high RSI values, confirming their resilience across both deterministic and stochastic variations.

4.5. Spatial Targeting of Priority Interventions

To enhance the spatial relevance of the results, the top-ranked policy alternatives were mapped against Yerevan’s Ecological Risk Index (ERI) and service accessibility distribution.

The spatial overlay integrated two principal GIS layers:

(1)

Ecological vulnerability (derived from LST × NDVI raster composites);

(2)

Water service coverage density (percentage of households with 24 h supply).

4.6. Spatial Synthesis

The resulting map identifies three high-priority zones—Shengavit, Ajapnyak, and Nor Nork districts—where ecological risk hotspots coincide with low service accessibility.

In these districts, the combined application of A1 (Leakage reduction), A2 (Demand-side management), and A5 (Data-driven governance) yields the most synergistic impact, maximizing both environmental and social benefits.

Furthermore, A4 (Green-blue infrastructure) provides substantial co-benefits in flood-prone sub-basins of Ajapnyak, mitigating surface runoff and improving microclimatic conditions.

This spatial prioritization demonstrates the analytical power of integrating MCDM outputs with GIS-based environmental diagnostics, creating an operational roadmap for geographically targeted interventions.

As illustrated in Figure 4, this figure shows the spatial distribution of top-ranked policy alternatives based on the intersection of the Ecological Risk Index (ERI) values and water service accessibility.

Dark red zones represent areas of high ecological stress and low accessibility, while green areas indicate well-served, low-risk neighborhoods.

The map highlights Shengavit, Ajapnyak, and Nor Nork as the most critical intervention zones for immediate policy action.

4.7. Interpretation

Spatial targeting allows decision-makers to align policy interventions with geographic equity and environmental urgency.

By linking quantitative rankings to spatial diagnostics, this framework operationalizes sustainability into where, not just what, actions should be prioritized.

The identified hotspots serve as a strategic blueprint for phased implementation—Phase I (Shengavit, Ajapnyak) and Phase II (Nor Nork, Arabkir)—ensuring efficient allocation of financial and technical resources.

5. Discussion and Policy Implications

5.1. Theoretical Contribution

This study contributes to the evolving discourse on integrated urban water governance by bridging three theoretical paradigms—sustainability transitions, multi-criteria decision-making (MCDM), and adaptive governance.

Traditional models of urban water management have predominantly emphasized either technical efficiency or ecological integrity in isolation. The present framework advances a systemic paradigm, treating decision-making as an integrative and iterative process shaped by socio-economic feedbacks, institutional learning, and environmental thresholds.

At a conceptual level, the study aligns with transition theory, viewing urban water systems as socio-technical assemblages that evolve through innovation, governance feedback, and adaptive policy cycles.

By embedding MCDM techniques within this dynamic governance lens, the research introduces a quantitative mechanism for normative evaluation, balancing trade-offs among the social, economic, and environmental dimensions of sustainability.

Furthermore, the model reinforces the notion that decision rationality in water governance is context-dependent—mediated by institutional maturity, digital readiness, and community participation.

In this regard, the Integrative Decision-Making Framework (IDMF) proposed herein extends beyond static optimization, positioning decision-making as a learning process that continuously recalibrates priorities in response to new data, shifting risks, and stakeholder engagement outcomes.

5.2. Methodological Innovation and Validation

This research introduces a hybridized decision-support mechanism that integrates Analytic Hierarchy Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) within a spatially contextualized governance framework.

While MCDM methods have been widely applied to resource planning, few studies have systematically embedded them in urban water governance that explicitly links social equity, environmental vulnerability, and institutional feasibility.

5.2.1. Innovation Dimensions

The methodological novelty lies in three core aspects:

1.

Normative-Analytical Coupling:

By coupling quantitative MCDM computation with governance indicators (GRI), the framework transcends traditional optimization. It recognizes that policy feasibility is a function not only of efficiency but also of institutional adaptability.

2.

Spatial Integration:

The fusion of AHP–TOPSIS results with geospatial layers (ERI and accessibility indices) allows for spatially explicit prioritization. This creates an actionable decision map that bridges the gap between model outputs and real-world implementation zones.

3.

Scenario Sensitivity and Monte Carlo Validation:

A consistent spatial association was observed between areas of high ecological stress and low service accessibility; however, this relationship should be interpreted qualitatively rather than statistically. While both layers exhibit overlapping hotspot zones, the available data do not allow us to report a validated rank-based correlation coefficient. Therefore, we refrain from presenting specific numerical values and instead emphasize the general co-location of environmental vulnerability and limited access to water services. This spatial correspondence reinforces the need for integrated policy responses, particularly in districts where ecological and social pressures converge.

5.2.2. Model Validation

Empirical validation was achieved through triangulation:

(a)

Cross-comparison with historical water utility performance data (2018–2023);

(b)

Expert consultation with policy practitioners and municipal engineers;

(c)

Stress-testing the model using synthetic perturbations of weighting schemes.

The results confirm that the Integrative Decision-Making Framework (IDMF) maintains predictive coherence and policy realism even under uncertainty, distinguishing it from earlier mono-dimensional assessment tools.

5.2.3. Expanded Discussion of Synergistic Dynamics and Spatial Mechanisms

Beyond the quantitative rankings, it is important to clarify the underlying mechanisms that explain why certain strategies—particularly leakage control (A1) and data-driven governance (A5)—produce synergistic sustainability benefits. Leakage reduction simultaneously improves service reliability, enhances cost-efficiency by reducing non-revenue water, and mitigates ecological stress by lowering abstraction pressure on vulnerable aquifers. When combined with intelligent monitoring systems, detection accuracy increases, operational inefficiencies decline, and policy responsiveness is strengthened. These co-benefits illustrate how technical interventions can amplify governance performance when embedded in a digitalized management architecture.

Differences in water governance challenges across regions also help contextualize Yerevan’s policy pathways. In many high-income cities, governance bottlenecks are linked to aging networks and declining investment cycles, whereas in post-Soviet and mid-income contexts, fragmentation, limited data transparency, and uneven institutional capacity remain dominant constraints. As a result, solutions that might be marginal in data-rich environments—such as digital dashboards or participatory feedback mechanisms—become transformative in settings where institutional coordination is weak. Recognizing these regional contrasts strengthens the transferability argument of the present framework by delineating the boundary conditions under which certain strategies become effective.

The spatial heterogeneity observed in Yerevan further reflects the interaction of social, technological, and ecological factors. Districts with higher ecological vulnerability (e.g., elevated LST–NDVI stress) tend to overlap with low-income or infrastructure-deficient neighborhoods, reinforcing spatial patterns of service inequality. Technological disparities—such as uneven sensor deployment or variable pipeline age—amplify these inequalities. Therefore, spatially explicit targeting, as demonstrated in the ERI-based overlay, highlights how governance outcomes are shaped by layered socio-ecological processes rather than by technical performance alone. Addressing these spatial differentials is essential for designing equitable water policies that balance efficiency with environmental justice.

Despite its contributions, the study has several limitations that should be acknowledged. First, the governance component of the model relies partly on expert-based scoring, which, although systematically structured, may introduce subjective bias that cannot be fully eliminated. Second, the environmental and accessibility indicators reflect a single temporal snapshot; the absence of longitudinal data restricts our ability to capture seasonal dynamics or long-term infrastructure trends. Third, the integration of ERI and accessibility layers is based on equal weighting assumptions, which, while justified for exploratory purposes, may not fully represent the relative importance of environmental versus social pressures in all contexts. Fourth, several administrative datasets—such as district-level leakage logs or service outage reports—contain inconsistencies and missing values, which required harmonization procedures that may influence final outcomes. Accordingly, the results should be interpreted as indicative rather than definitive, and future work will benefit from more granular, multi-year datasets and expanded stakeholder participation.

To ensure full transparency and reproducibility, all non-restricted datasets and processing files used in this study will be made publicly available through a dedicated Zenodo repository upon publication. The repository will include (a) the normalized decision matrices used in the AHP–TOPSIS evaluation; (b) expert elicitation tables and the derived governance weights; (c) GRI scores for all policy alternatives; (d) metadata and classification rules for the environmental and accessibility indicators; and (e) the Python and GIS scripts used for normalization, spatial overlay, and hotspot mapping. These materials will allow independent researchers to replicate the analytical workflow and verify the robustness of our results.

5.3. Policy and Governance Implications

The findings of this study have significant implications for urban water governance, particularly in mid-income cities such as Yerevan, where infrastructural limitations, fiscal constraints, and governance fragmentation converge.

By integrating technical evaluation, social equity, and institutional readiness into a unified decision-making model, the study provides policymakers with a transparent and adaptable instrument for prioritizing interventions.

5.3.1. Strategic Implications

1.

Institutional Mainstreaming of Integrated Planning:

The proposed framework encourages municipalities to embed multi-criteria decision-making tools into their regular planning cycles. Embedding the AHP–TOPSIS structure within the existing governance architecture of Yerevan Water and the Ministry of Environment would enable evidence-based prioritization of investments, reducing political bias and ad hoc decision-making.

2.

Enhancing Transparency and Accountability:

The open-data orientation of the A5 (Data-driven governance) alternative supports transparent decision chains.

Establishing digital dashboards that visualize performance metrics (e.g., water losses, service coverage, ecological risk) can improve public trust and facilitate civic monitoring, thereby strengthening democratic accountability.

3.

Socio-Spatial Equity in Investment Allocation:

The spatial overlay results emphasize that resource allocation should account for intra-urban disparities.

Prioritizing underserved districts such as Shengavit, Ajapnyak, and Nor Nork aligns water management with the principles of environmental justice and inclusive urban development.

4.

Regulatory and Financial Reforms:

To operationalize decentralized or green infrastructure solutions, targeted regulatory amendments are necessary—particularly regarding greywater reuse standards, stormwater management incentives, and private sector participation.

A dedicated municipal Sustainability Fund could support co-financing of community-based initiatives, bridging the gap between local action and strategic planning.

5.3.2. Governance Roadmap

The study recommends establishing a City-Level Water Governance Task Force, comprising municipal authorities, utility representatives, academic experts, and citizen groups.

This body would serve as a cross-sectoral platform to monitor the performance of the integrated framework, update decision criteria annually, and coordinate investments according to risk–efficiency trade-offs.

5.3.3. Prerequisites for Framework Transferability

Although the proposed framework is suitable for replication in other mid-income cities, its successful application requires several foundational conditions. First, a minimum level of data availability is needed, including basic hydrological records, utility performance indicators, and socio-economic profiles at the district scale. Second, baseline governance capacity—such as inter-agency coordination mechanisms and procedural clarity in decision-making—is essential to ensure that model outputs can be operationalized. Third, institutional maturity, reflected in stable regulatory structures and medium-term planning cycles, enables continuity in policy implementation. Finally, a threshold level of technical infrastructure, including digital metering, GIS capabilities, and routine monitoring systems, is necessary for integrating the AHP–TOPSIS evaluation with real-world operational processes. These prerequisites define the boundary conditions under which the framework can be reliably transferred and adapted to different urban and governance contexts.

5.3.4. Policy Message

Ultimately, sustainable urban water management is not a purely technical challenge but an issue of governance, institutional design, and participatory capacity.

By adopting the proposed integrative framework, Yerevan can transition from reactive management to anticipatory governance, positioning itself as a model for medium-sized cities seeking climate-resilient and socially inclusive water systems.

6. Conclusions

This study developed and applied an Integrative Decision-Making Framework (IDMF) for sustainable urban water governance, combining analytical precision with institutional realism.

By embedding the AHP–TOPSIS methodology within a governance and spatial assessment structure, the research demonstrated how social, economic, and environmental objectives can be balanced to guide policy prioritization in complex urban contexts such as Yerevan.

The empirical results reveal that technologically grounded and data-driven strategies—particularly leakage reduction and smart governance systems—deliver the most synergistic outcomes when aligned with equitable access and ecological preservation goals.

From a theoretical perspective, the framework advances the discourse on adaptive and participatory governance, emphasizing decision-making as a dynamic, learning-oriented process.

Methodologically, it integrates deterministic and stochastic MCDM models with GIS-based diagnostics, establishing a replicable blueprint for spatially explicit sustainability planning.

This dual integration ensures that decisions are both analytically justified and geographically grounded.

In practical terms, the findings support the formation of a city-level water governance task force and the institutionalization of evidence-based planning instruments.

By prioritizing interventions in districts such as Shengavit, Ajapnyak, and Nor Nork, the approach enables more inclusive and climate-resilient management of urban water systems.

The proposed framework can thus serve as a scalable model for other mid-income cities seeking to strengthen the linkage between technical efficiency, social equity, and environmental stewardship within the broader agenda of sustainable urban transformation.