Evaluating Spatial Equity in Urban Park Accessibility: A GIS‑Based Analysis of the İzmir Gulf Region, Turkey

ARCHITECTURAL SCIENCES AND SUSTAINABLE APPROACHES: URBAN RESILIENCE Editors Prof. Dr. Ömer ATABEYOĞLU Prof. Dr. Ertan DÜZGÜNEŞ October 15, 2025 Copyright © 2025 by İKSAD publishing house All rights reserved. No part of this publication may be reproduced, distributed or transmitted in any form or by any means, including photocopying, recording or other electronic or mechanical methods, without the prior written permission of the publisher, except in the case of brief quotations embodied in critical reviews and certain other noncommercial uses permitted by copyright law. Institution of Economic Development and Social Researches (The Licence Number of Publicator: 2014/31220) TÜRKİYE TR: +90 342 606 06 75 USA: +1 631 685 0 853 E mail: [email protected] www.iksadyayinevi.com It is responsibility of the author to abide by the publishing ethics rules. Iksad Publications – 2025© Architectural Sciences and Sustainable Approaches: Urban Resilience ISBN: 978-625-378-337-2 Cover Design: Prof. Dr. Ertan DÜZGÜNEŞ October 15, 2025 Ankara / Türkiye Size = 16x24 cm PREFACE Dear Professors and Colleagues, We are pleased bring to life that Architectural Sciences and Sustainable Approaches: Urban Resilience, which was published as an e-book by IKSAD Publishing House with the editors Prof. Dr. Ömer ATABEYOĞLU and Prof. Dr. Ertan DÜZGÜNEŞ. This book project, entitled “Architectural Sciences and Sustainable Approaches: Urban Resilience,” aims to address sustainability-oriented approaches to urban resilience from theoretical, methodological, and practical perspectives. The volume seeks to establish a multi-layered platform of discussion, ranging from the scale of individual buildings to the entirety of the urban fabric. Within this framework, it welcomes contributions from scholars and researchers working in architecture, urban design, landscape architecture, urban and regional planning, environmental engineering, and related disciplines. With the valuable contributions of our chapter authors working in the professional disciplines of landscape architecture, architecture, city and regional planning, urban design and sustainability, we have completed Architectural Sciences and Sustainable Approaches: Urban Resilience book study has been completed with 24 book chapters. We would like to thank you, our esteemed authors, for their contributions to the preparation of the book. We would also like to thank the editorial board and IKSAD Publishing House. We wish to continue this process we have started in the coming years. In addition, we would like to express our sincere appreciation to Prof. Dr. Atila GÜL, the book coordinator of IKSAD Publishing House, for his guidance and support throughout the publication process. We hope that our book ‘Architectural Sciences and Sustainable Approaches: Urban Resilience’ will be helpful to the readers. Best regards. 15.10.2025 EDITORS Prof. Dr. Ömer ATABEYOĞLU Prof. Dr. Ertan DÜZGÜNEŞ EDITORS Prof. Dr. Ömer ATABEYOĞLU Prof. Dr. Ertan DÜZGÜNEŞ AUTHORS The authors were listed in alphabetical order Alper ÇABUK Ayça GÜLTEN Ayşe ÖZYETGİN ALTUN Ayşe Özge ŞİMŞEK SOYSAL Ayşegül TANRIVERDİ KAYA Demet EROL Deniz DEMİRARSLAN Ebru Vesile ÖCALIR Eda ŞENTÜRK Elif Kübra ÖZTÜRK Emine BAYDAN Esra KESKİN Feran AŞUR Feyza Sena ŞENOCAK Filiz KARAKUŞ Furkan AKDEMİR Gencay ÇUBUK Gülşah BİLGE ÖZTÜRK Halil DUYMUŞ Hamza ALTAŞ Hande AKARCA İnci OLGUN Kemal Mert ÇUBUKÇU Kumru ÇILGIN Mehmet Akif IRMAK Mehmet Emin DAŞ Mehtap ÖZENEN KAVLAK Merve ALICI AKA Mesut GÜZEL Muhammed Akif AÇIKGÖZ Muhammed Emir GÖRAL Murat YEŞİL Olcay Türkan YURDUGÜZEL Özge DÜZGÜN EREKİNCİ Pervin YEŞİL Rabia Nurefsan ACIKGOZ Sedef ŞENDOĞDU Seher Simay KUŞOĞLU Serim DİNÇ Sevilay YILDIZ Sinem SEYHAN Şevval ERGİNDOĞAN Şuheda ALTUNOK Temuçin Göktürk SEYHAN Tuba Nur OLĞUN Tuna BATUHAN Ufuk Teoman AKSOY Yusuf Eminoğlu REVIEWER LIST The authors were listed in alphabetical order Aslıhan TIRNAKÇI Nevşehir Hacı Bektaş Veli University Atila GÜL Süleyman Demirel University Ayşe Kalaycı ÖNAÇ İzmir Katip Çelebi University Bige ŞİMŞEK İLHAN İstanbul Medipol University Burcu YILMAZEL Eskişehir Technical University Eda KOÇAK Siirt University Ekrem BAHADIR Ankara Yıldırım Beyazıt University Elif KUTAY KARAÇOR İstanbul Technical University Hakan ARSLAN Ondokuz Mayıs University Hilal TURGUT Karadeniz Technical University Meliha AKLIBAŞINDA Nevşehir Hacı Bektaş Veli University Murat AKTEN Süleyman Demirel University Nihan Sümeyye GÜNDOĞDU Atlas University Okan Murat DEDE Amasya University Ömer Lütfü ÇORBACI Recep Tayyip Erdoğan University Selcen Nur Erikci Çelik Beykoz University Sibel AKTEN Isparta Unıversıty Of Applıed Scıences Sinem ÖZDEDE Pamukkale University Şeyma ŞENGÜR Ordu University Turgut KALAY Kütahya Dumlupınar University Tendü Hilal GÖKTUĞ Aydın Adnan Menderes University 313 1. Introduction 1.1. Research Motivation and Problem Context Urban green spaces (UGS) serve as critical infrastructure for ecological sustainability, climate resilience, and public well-being. In rapidly urbanizing regions, however, their distribution, accessibility, and equity remain profoundly uneven—reflecting and often reinforcing broader spatial and social disparities (Li et al., 2024; S. Wu, Chen, Webster, Xu, & Gong, 2023; Xu et al., 2022). Globally, the intensification of urban heat islands (UHIs), compounded by climate change and densification, underscores the necessity of green infrastructure not only in terms of total provision but also in terms of equitable spatial allocation (Okumus & Terzi, 2023; Yang, Jin, & Li, 2024). Inequities in green space distribution disproportionately impact vulnerable groups, limiting both ecological benefits and access to essential recreational and thermal regulation services (LaReaux & Watkins, 2025; Zhao & Gong, 2024). Spatial inequity in green space accessibility has been studied from diverse perspectives—including environmental justice, spatial planning, and urban resilience—but often lacks methodological convergence. While accessibility indicators have proliferated (Semenzato, Costa, & Campagnaro, 2023), there remains limited consensus on how to incorporate spatial configuration, population heterogeneity, and networkbased proximity into a unified equity assessment (He, Wu, & Wang, 2020; Leng, Sun, Yang, & Chen, 2023). Moreover, a persistent divide is observed between the Global North and South in terms of both total green exposure and adaptation capacity, with cities in the South exhibiting 314 significantly lower average cooling benefits and greater inequities in park access (Li et al., 2024; S. Wu et al., 2023). 1.2. Conceptual and Thematic Background The debate over green space equity spans horizontal equity—equal access regardless of individual characteristics—and vertical equity—just distribution that accounts for social vulnerability (He et al., 2020; Xia, He, & Zhang, 2024). In this regard, several studies have emphasized the role of socioeconomic and demographic stratification in mediating spatial inequity, particularly among children, women, the elderly, and lowincome populations (Jin, He, Hong, Luo, & Xiong, 2023; Kaya & Eminoğlu, 2025; LaReaux & Watkins, 2025; Şenol & Atay Kaya, 2024; Zhao & Gong, 2024). As noted by Guan et al. (2023) , spatial configuration and pattern metrics also shape access disparities, revealing that accessibility is not solely a function of park area or number, but of landscape morphology and connectivity. Recent advances have refined the methodological toolkit for equity assessment, including network-based service areas (Gao, Xu, Shang, Li, & Wang, 2025), floating catchment methods (Lan, Liu, Huang, Corcoran, & Peng, 2022; Zhao & Gong, 2024), space syntax for pedestrian flow modeling (Huang, Li, Ma, & Xiao, 2023; Kahraman & Çubukçu, 2023), and spatial regression approaches such as GWR and MGWR (Adorno, Pereira, & Amaral, 2025; Wang & Guan, 2025). These approaches increasingly integrate multiple dimensions of accessibility—e.g., visibility, proximity, and service coverage—while seeking to map spatial heterogeneity in explanatory mechanisms. 315 1.3. Methodological Framing and Innovation This chapter responds to the need for a multi-method, spatially explicit framework that bridges normative concepts of green equity with empirical measures of spatial configuration and social exposure. It develops and operationalizes a network-based accessibility model for 692 validated urban parks across 237 neighborhoods in the İzmir Gulf Region, Türkiye—an urban corridor of environmental and demographic diversity. The methodology incorporates space syntax metrics, network centralities, and service area modeling at four spatial thresholds (300 m, 500 m, 1000 m, 1500 m), combining them with population-weighted Gini and Lorenz statistics to examine access inequalities. Following global calls for replicable and rigorous planning tools (Adorno et al., 2025; Martin & Conway, 2025; Rubaszek, Gubański, & Podolska, 2023), the study uses multiscale regression models (OLS, GWR, MGWR) to decompose spatial variability in park accessibility outcomes. Variables such as canopy cover, elevation, built-up density, and impervious surface area are tested across different spatial kernels to reveal local predictors of access deficits. Building on equity classification typologies (C. Wu, Wei, Huang, & Chen, 2017; J. Zhang & Tan, 2023), the results culminate in a tiered priority framework for planning interventions, where Tier 1 neighborhoods exhibit the highest urgency across multiple spatial and social dimensions. 1.4. Research Gap and Study Contribution Although spatial equity assessments have become increasingly common in urban planning research, most studies continue to rely on Euclidean buffers, omit population heterogeneity, or fail to examine accessibility as 316 a function of network constraints (Leng et al., 2023; Semenzato et al., 2023). This study addresses those limitations by (i) grounding accessibility within real pedestrian networks, (ii) applying population-weighted equity diagnostics for key demographic cohorts, and (iii) integrating spatial regression models to understand place-based drivers of inequity. Importantly, the İzmir Gulf Region presents a relevant empirical case due to its pronounced topographic variation, dense coastal urbanization, and active green space planning initiatives. Drawing upon lessons from comparative studies in China, Europe, and North America (Adorno et al., 2025; Lan et al., 2022; Martin & Conway, 2025; Z. Zhang, Cenci, & Zhang, 2024), the study contributes an adaptable, scalable framework for evaluating access equity that can inform SDG 11.7 and local sustainability agendas. The study is further distinguished by its integration of spatial diagnostics (LISA, Gi*), quantitative equity scores, and regression-based decomposition into a single analytical framework, facilitating a policyrelevant synthesis. The resulting classification of high-priority zones provides actionable guidance for urban planners, environmental policymakers, and equity-driven park design interventions (Jin et al., 2023; J. Zhang & Tan, 2023). Figure 2 illustrates the methodological workflow, detailing data inputs, network-based processing, regression modeling, and spatial diagnostics across the chapter’s analytical framework. 317 2. Material and Method 2.1. Study Area and Units of Analysis The study focuses on the “İzmir Gulf Region”, a contiguous urban corridor encompassing coastal districts of İzmir Province, Türkiye. Defined by the spatial extent of pedestrian networks and urban settlement forms surrounding the inner Gulf, the analysis targets neighborhoods that are functionally integrated with the metropolitan coastline and accessible open space systems. Following spatial filtering and validation, the final analytical scope includes 237 neighborhoods (mahalle). These units represent the lowest tier of official administrative geography in Türkiye and are widely used for urban diagnostics, planning, and service delivery. They also align with the resolution of available demographic and spatial datasets, supporting consistent aggregation and analysis. As shown in Figure 1, (i) the administrative district layout of İzmir Province (top left), (ii) the bounding box of the Gulf region overlaid on the full provincial extent (middle left), (iii) true-color satellite imagery underlaying the road network and park locations within the study area (bottom left), and (iv) a digital elevation model (DEM) overlay emphasizing topographical patterns and fluvial corridors within the region . Together, these components provide a spatial overview of the landscape conditions and infrastructural context in which park accessibility is evaluated. 318 Figure 1. Study area and base layers (İzmir Gulf Region) All variables in this study are aggregated or attributed at the neighborhood level, including accessibility measures, demographic cohorts, network statistics, environmental indices, and model outputs. Where point representations are required (e.g., for regression kernels or cluster diagnostics), projected neighborhood centroids derived from polygon geometries are used to preserve spatial accuracy. This unit-of-analysis framework ensures compatibility with planning practice, protects the confidentiality of population data, and facilitates reproducibility across urban studies grounded in official territorial hierarchies. 2.2. Data and Variables This study integrates spatial and sociodemographic datasets to assess pedestrian access to urban parks in the İzmir Gulf Region. All layers were reprojected to a common metric coordinate system (UTM Zone 35N; 319 EPSG:32635) and processed within a unified GeoPackage. The complete inventory of data sources, resolutions, and preprocessing steps is provided in Table 1. Table 1. Datasets and sources. Layer Geometry / resolution Source (year) Preprocessing summary Neighborhoods Polygon (admin) OpenStreetMap extract (2025-07-20) Attribute cleaning; code harmonization; reprojection. Urban parks Polygon IMM Open Data Portal Validation; dissolve by park; area recomputation. Pedestrian road network Line (graph) OpenStreetMap extract (2025-07-20) Filter to pedestrian‑permitted links; snap park entrances to nearest node. Space‑syntax segments Line (graph) Author‑derived using space‑syntax toolkit Segment metrics aggregated to neighborhood median. Network centralities Line (graph) Author‑derived (GrassGIS v.net) Node/edge centralities summarized to neighborhood median. Building footprints Polygon OpenStreetMap extract (2025-07-20) Coverage % computed by neighborhood. Impervious surface Raster/10m GEE Dynamic World Total impervious area per neighborhood. Tree canopy Raster/1m Meta and WRI, Tolan et al 2023 Total canopy area per neighborhood. Elevation (DEM) Raster/30m DEM Copernicus2025-06 Neighborhood median elevation. Land surface temperature Raster/30m GEE Landsat 8-9 2025-05:07 Median Neighborhood median LST. The primary dependent variable is park-per-capita (PPC), calculated for each neighborhood as the cumulative area of reachable parks divided by the neighborhood’s population, at four walking-distance thresholds: 300 m, 500 m, 1000 m, and 1500 m. This measure captures network-based accessibility to parkland rather than proximity alone, and is sensitive to the number and size of accessible parks, not merely their presence. Equity is assessed by computing the population-weighted Gini index for PPC across 320 all neighborhoods. Gini coefficients are calculated both for the total population and three vulnerable cohorts: children under five, women, and adults aged 65 and older. Lorenz curves are also derived to visually depict the cumulative distribution of accessible park area. These metrics are interpreted by radius and population group to evaluate patterns of distributive justice in green space provision. The study includes a set of theoretically grounded neighborhood-level predictors hypothesized to influence PPC. These variables are grouped into four thematic domains: • Network configuration: space-syntax integration (`INT_med`) and graph-based centrality metrics (`c_closeness_med`, `c_betweennes_med`), derived from segment-based and nodebased analyses of the pedestrian network. • Built environment: neighborhood-level building coverage (`building_coverage_pct`) and total impervious area (`imp_total_m2`), both of which are expected to constrain open space availability. • Green structure: total tree canopy area (`canopy_area_m2`), serving as a proxy for green infrastructure and ecological character. • Topography and thermal environment: median elevation (`elev_med`) and median land surface temperature (`lst_med`), reflecting possible constraints on access or vegetation presence. To ensure model validity, variables exhibiting zero variance were excluded, and predictor pairs with Pearson correlation coefficients r > 0.95 were screened to avoid multicollinearity. Count and area-based predictors were log-transformed using log1p, then z-standardized prior to modeling. 321 Population totals and subgroup counts (children under 5, women, and adults 65+) are sourced at the neighborhood level and used both as denominators in PPC calculation and as weights in the Gini and Lorenz metrics. This enables stratified assessments of accessibility equity, emphasizing variation in park provision across vulnerable populations. The variable definitions, coding, and expected relationships are summarized in Table 2 and Table 3. Table 2. Dependent (access/equity) variables. Code Definition Unit Interpretation PPC_300:1500_tot, Reachable park area per capita at distance d m² per person Higher indicates greater accessible park supply via the network. Ginid_d (total, u5, women, 65+) Population‑weighted Gini of PPC at d 0–1 Larger values denote greater inequality of access. Lorenzd_d Lorenz curve of PPC at d — Cumulative distribution of access (visual companion to Gini). Table 3. Explanatory (driver) variables used in OLS/GWR/MGWR Code Description Sign Rationale INT_med Space‑syntax integration (median) + Greater through‑movement potential supports practical proximity to parks. c_closeness_med Network closeness (median) + Shorter network distance to many destinations facilitates access. c_betweennes_med Network betweenness (median) ± Through‑corridor effects can aid or hinder local access. building_coverage_pct Building land‑cover share - Higher coverage reduces open/green availability. imp_total_m2 Impervious surface total - Hardscape correlates with lower green provision and thermal burden. canopy_area_m2 Tree canopy area total + Proxy for green infrastructure and adjacent parkland. elev_med Median elevation ± Topography may constrain pedestrian routes; sign is context‑dependent. lst_med (optional) Median land surface temperature - Heat‑prone surfaces tend to coincide with scarce green access. 322 2.3. Network‑Based Service Areas and Park‑per‑Capita (PPC) To evaluate pedestrian accessibility, service areas were computed around each of the 692 validated urban parks using network distances of 300 m, 500 m, 1000 m, and 1500 m. A shortest-path graph was constructed to generate service areas reflecting real walking conditions. Each park was linked to its nearest walkable node and assumed to have at least one entrance. Where minor roads bisected parks, polygons were dissolved to preserve continuity. While the single-entrance assumption simplifies modeling, its implications for large parks are noted in Section 4. Neighborhood-level park-per-capita(PPC) is computed for network-based walking distances 𝑑∈ {300, 500, 1000, 1500}m by attributing reachable park area to each neigborhood and dividing by its population. Formally, for neighborhood 𝑗, 𝑃𝑃𝐶𝑑𝑗 =∑𝐴𝑝𝑝∈𝑃(𝑑𝑗) 𝑁𝑗 (1) where 𝐴𝑝 is park 𝑝’s area and 𝑃(𝑑,𝑗) is the set of parks whose 𝑑 service area intersects neighborhood 𝑗; 𝑁𝑗 is the neighborhood population. Cohortspecific PPC measures were also derived using subgroup population counts. These indicators form the basis for the equity metrics and modeling analyses described in subsequent sections. 2.4. Equity Metrics, Spatial Diagnostics, Modeling Framework This section integrates the procedures for evaluating distributive equity, detecting spatial clustering, modeling explanatory relationships, and synthesizing outcomes for planning relevance. The analytical steps described here follow the structure summarized in Figure 2. 329 identify areas of spatial autocorrelation and statistically significant hotspots or coldspots of urban park access. The global Moran’s I statistic, computed for PPC at each radius, confirms a strong spatial structure in access values. Spatial clustering is minimal at 300 m (I = 0.15) but rises sharply at 500 m (I = 0.40), peaking at 1000 m (I = 0.56) and remaining high at 1500 m (I = 0.54). All results are significant at p < 0.01, indicating that neighborhoods with similar access levels tend to be spatially proximate at larger distances. Figure 5 displays the LISA significance maps for high-high (HH) and lowlow (LL) clusters across radii. At 300 m (Panel A), no statistically significant local clusters are detected. At 500 m (Panel B), a small number of HH clusters emerge near central districts (n = 5), while a few LL clusters (n = 3) appear in low-density peripheries. At 1000 m (Panel C), clustering intensifies with 17 HH neighborhoods in Konak, and 27 LL neighborhoods in outer Karşıyaka, Balçova, Bayraklı and Bornova. At 1500 m (Panel D), this pattern consolidates further, showing 19 HH and 48 LL neighborhoods. Notably, LL clusters align with areas of lower walkability and sparse park coverage, reinforcing prior Gini results. 330 Figure 5. LISA cluster significance maps for PPC by network radius. Figure 6 shows the standardized Gi* z-scores for each neighborhood across radii. High positive z-scores indicate statistically significant accessibility hotspots, while negative values signal coldspots. At 300 m, Gi* clustering is weak and patchy, but becomes more structured at 500 m and stabilizes at 1000 m and 1500 m. Hotspots consistently emerge in core districts such as Alsancak, Bornova, and central Konak, where both park supply and network connectivity are high. Coldspots align with outlying hillslopes and disconnected peripheries in northern Karşıyaka and southern Balçova. The results from both LISA and Gi* diagnostics confirm that spatial equity is not uniformly distributed across the Gulf Region. Instead, sharp intraurban gradients persist even under extended service assumptions. These 331 localized disparities justify geographically disaggregated modeling in the subsequent section. Figure 6. Getis-Ord Gi* z-scores for PPC by network radius. 3.4. Model-Based Insights To explore the structural determinants of spatial access to urban parks, this study applied a stepwise regression framework incorporating Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Multi-Scale GWR (MGWR). Model performance and spatial heterogeneity were assessed across four network service distances (d = 300, 500, 1000, and 1500 m), using PPC for the total population as the dependent variable. All predictors were standardized, and skewed variables were log-transformed prior to model fitting. 332 Global model comparison reveals that MGWR consistently outperforms both OLS and single-bandwidth GWR across all radii. MGWR achieves superior R² values—ranging from 0.463 at 300 m to 0.821 at 1500 m— while maintaining lower AICc scores. In contrast, GWR yields relatively modest fits at lower radii, especially at 300 m (R² = 0.17), but improves with scale. These gains demonstrate the importance of multiscale modeling to capture non-stationary spatial processes underlying park accessibility. Figure 7 visualizes the GWR local R² surfaces across the four radii. At 300 and 500 m, model fits are uniformly low, with minimal spatial variation. As the radius increases, spatial heterogeneity becomes more pronounced. At 1500 m, central and southern districts display significantly higher local R² values—approaching 0.9 in some neighborhoods—indicating stronger model explanations at broader accessibility thresholds. Figure 7. Local GWR R² surfaces across service distances. To unpack these drivers, MGWR coefficient maps are provided in Figure 8 for four key predictors (elev_med, INT_med, building_coverage_pct, and canopy_area_m2) at d = 500 m (top row) and 1500 m (bottom row). Visual inspection reveals distinct spatial patterns. For instance, at 1500 m, the positive effect of tree canopy area (Panel H) is concentrated in the central- 333 southern corridor, while elevation exhibits stronger negative associations in coastal areas (Panel E). These patterns reflect the compound influence of both terrain and green infrastructure on perceived park accessibility. Complementing these maps, Figure 9 provides a comprehensive beeswarm-style plot of local MGWR coefficients across neighborhoods for all eight standardized predictors and all four radii. Violin curves summarize the coefficient distribution, while dot colors indicate the normalized z-score of each predictor per neighborhood. This visualization highlights both the direction and magnitude of local effects. Notably, the influence of impervious surface (imp_total_m2) and canopy cover (canopy_area_m2) is more consistent across scales, whereas metrics such as network closeness and betweenness centrality show greater local fluctuation. Figure 8. Key MGWR coefficients mapped at d = 500 m and 1500 m. Taken together, the results underscore the critical role of urban form, spatial configuration, and green infrastructure in shaping the equity of park access. They also confirm that spatial non-stationarity is scale-dependent, 334 reinforcing the need for multiscale analytical strategies in accessibility planning. Figure 9. Beeswarm plots of MGWR local coefficients across predictors and radii. Panels (A–D) represent d = 300 m, 500 m, 1000 m, and 1500 m, respectively. 3.5. Synthesis for Planning To translate the analytical findings into actionable planning guidance, this study developed a composite prioritization framework to stratify neighborhoods based on their accessibility deficits and contextual impediments to urban park access. The resulting classification delineates a four-tier system: Tier 1 (highest urgency), Tier 2, Tier 3, and Monitor (low or uncertain need), as shown in Figure 10A. 335 Figure 10. Priority classification for spatial equity interventions. The prioritization logic integrates three key indicators: (i) populationweighted park-per-capita (PPC) deficit at d = 1500 m, (ii) presence of spatial clustering (Local Indicators of Spatial Association, LISA) indicating low-low access patterns at d = 1000 and 1500 m, and (iii) the average standardized strength of negative MGWR coefficients for impervious surface (β_imp), representing environmental constraints to equitable park provision. All variables were normalized and aggregated into a composite score, with rule-based thresholds guiding tier assignment. Tier 2 neighborhoods exhibit substantial PPC deficits (mean = 0.88) and localized clustering of poor access conditions, alongside moderate levels of imperviousness-related constraint (mean β_imp = 0.03), as detailed in Figure 10B. Tier 3 neighborhoods have milder access gaps (mean PPC deficit = 0.53) but still warrant future interventions, particularly where β_imp effects are stronger (mean = 0.21). Monitor areas show lower 336 urgency across all metrics, though select neighborhoods within this group may exhibit latent needs due to high imperviousness or low canopy cover. Importantly, no neighborhood met all threshold conditions for Tier 1, reflecting either a dispersion of extreme vulnerability or mitigating effects of surrounding green infrastructure. The population-weighted distribution confirms this, with Tier 2 comprising 23.8% and Tier 3 comprising 17.3% of the study area’s population. This stratified output facilitates targeted investment planning by enabling planners to spatially differentiate urgency and intervention type. For example, Tier 2 neighborhoods may be prioritized for both new park development and impervious surface mitigation, while Tier 3 areas may benefit from retrofitting access paths or expanding entrances to existing parks. 4. Conclusion and Suggestions This chapter has presented a spatially explicit, network-based analysis of urban park accessibility and equity in the İzmir Gulf Region, Türkiye. By integrating high-resolution spatial datasets with advanced GIS modeling techniques—including network-based service area analysis, populationweighted equity diagnostics, spatial autocorrelation statistics (LISA and Gi*), and multiscale regression models (OLS, GWR, MGWR)—the study advances the methodological rigor of urban green space equity assessments. The findings demonstrate significant inequalities in neighborhood-level access to parks, with persistent spatial clustering of highand low-access zones, as well as localized predictors driving these disparities. Network-constrained accessibility modeling revealed considerable variation in park-per-capita (PPC) values across neighborhoods and 337 distance thresholds. At closer radii (e.g., 300 m and 500 m), accessibility inequities were particularly acute, with population-weighted Gini coefficients exceeding 0.70 for the total population and remaining above 0.62 even for prioritized demographic groups such as children and elderly residents. The composite Lorenz curves further revealed disproportionate concentrations of park resources among higher-access neighborhoods. These distributional disparities were reinforced by spatial clustering patterns. Significant high–high (HH) and low–low (LL) clusters—detected through LISA—emerged more prominently at 1000 m and 1500 m distances, indicating persistent spatial segregation in park accessibility. Gi* analyses complemented this interpretation by identifying statistically significant cold spots of access that align with socioeconomically vulnerable neighborhoods, thus suggesting spatial injustice in green resource allocation (Adorno et al., 2025; He et al., 2020; Huang et al., 2023; Zhao & Gong, 2024). Regression modeling confirmed that park accessibility is not spatially random but is associated with distinct urban morphological and environmental variables. MGWR outperformed OLS and GWR in explaining PPC variations, with adjusted R² values increasing from 0.46 at 300 m to 0.82 at 1500 m. Locally varying coefficients demonstrated that indicators such as space syntax integration (INT_med), building coverage percentage, impervious surface, and canopy area exerted differential effects across the study area, reflecting strong spatial heterogeneity in the equity landscape. These findings resonate with global evidence showing that physical form and environmental context substantially condition the 338 distribution and performance of green infrastructure (Lan et al., 2022; Leng et al., 2023; S. Wu et al., 2023). A policy-relevant synthesis was constructed by generating a composite priority score from MGWR local diagnostics, persistent low-access clusters, and PPC deficits. This score was used to classify neighborhoods into three intervention tiers. Tier 1 zones—exhibiting both structural access deficits and weak model fit—are proposed as the highest priority for immediate action. The resulting spatial tiers offer a replicable tool for guiding resource allocation in urban greening interventions. Importantly, this approach aligns with the principles of vertical equity by emphasizing neighborhoods with overlapping environmental and socio-spatial vulnerabilities (Kaya & Eminoğlu, 2025; LaReaux & Watkins, 2025; Martin & Conway, 2025). From a planning perspective, the study highlights the need for differentiated strategies at multiple scales. For example, high-density urban neighborhoods with limited open space may benefit from microscale interventions such as pocket parks, vertical greening, or improved connectivity through redesigned pedestrian networks. In contrast, peripheral districts with larger but underutilized parkland require improved accessibility infrastructure and targeted social outreach. These targeted strategies reflect the growing consensus that one-size-fits-all solutions fail to address complex, spatially situated inequities in green space access (Jin et al., 2023; Xia et al., 2024; J. Zhang & Tan, 2023). In closing, this chapter contributes a methodologically robust and policyrelevant framework that integrates spatial analytics and planning intelligence. It reinforces the imperative for green infrastructure planning