The global urban population is projected to reach 68% by 2050, emphasizing the importance of urban greening initiatives and nature-based solutions to improve public health, well-being, and mitigate environmental impacts. Public Green Areas (PGAs) have been linked to various health benefits, including reduced mortality rates and improved mental health. These benefits have led to the establishment of numerous green-related targets for cities by organizations such as the World Health Organization (WHO) and the incorporation of green space accessibility into the UN Sustainable Development Goals. The 3-30-300 paradigm exemplifies the multi-level approach to urban greening, highlighting the need for green spaces at various scales (three trees visible from every home, 30% tree canopy cover in every neighborhood, and a greenspace within 300 m of every home). The increasing focus on data-driven policy design necessitates the development of spatially-resolved indicators to monitor progress towards these goals. However, a universally adopted framework for measuring accessibility to urban green is currently lacking. Existing research utilizes diverse indicators, ranging from minimum distance to the nearest park to metrics evaluating total green exposure from satellite data. The definition of spatial indicators presents challenges, particularly concerning the interplay between population distribution and greenspace location, along with walkable catchment areas. While some studies incorporate behavioral data from various sources, these are often limited by data accessibility and representativeness. This study addresses the gap in research by investigating the interchangeability of accessibility patterns derived from different green accessibility indicators and their implications for policy planning.

The literature review highlights the existing diversity of indicators used to measure green accessibility in urban areas. Studies have employed various approaches, including calculating the minimum distance to the nearest park using data from administrative sources, OpenStreetMap (OSM), or satellite imagery. Other studies have focused on assessing total green exposure based on satellite data on land cover or green intensity using the Normalized Difference Vegetation Index (NDVI). Furthermore, the literature points to the increasing use of behavioral data from surveys or user-generated geographic information (UGI) derived from social media, sports tracking, mobile phone traces, and PPGIS to understand actual greenspace usage and green exposure levels. The review acknowledges the computational challenges associated with incorporating walkable catchment areas into large-scale studies, while emphasizing that such considerations are common in single-city analyses. Finally, the review identifies a gap in research on the interchangeability of these different indicators and the implications for policy design.

This study develops a computational framework to measure three families of green accessibility indicators: minimum distance, exposure, and per-person. The minimum distance indicator measures the walking distance to the nearest PGA (parameterized by minimum size and type of green). The exposure indicator measures the amount of urban green within a walking time budget (parameterized by time budget and minimum size of green features). The per-person indicator considers the per-capita availability of PGAs within a walking time budget (also parameterized by time budget, minimum size, and type of green). The framework utilizes data from the Global Human Settlement Urban Center Database (GHS-UCDB), OpenStreetMap (OSM), and the European Space Agency's 2020 World Cover database. The framework is applied to over 1040 cities worldwide. The study then evaluates the stability of these indicators to small changes in their underlying parameters, focusing on three dimensions: (i) ranking stability of geographical units using the Kendall rank correlation coefficient; (ii) stability of population subgroups at the lower end of the ranking using two targeting strategies ('naive targeting' and 'most-disadvantaged targeting'); and (iii) inequality levels using a weighted Gini indicator. Finally, the study assesses the overlap among targets established by selected institutional bodies to examine the interchangeability of accessibility pictures derived from different metrics. The geographical units of analysis are rectangular squares of side equal to 9 arc-sec (around 196 m). Walking distances are computed using the walkable street network and a base walking speed of 5 km/h, allowing for customization based on specific user capabilities. The study uses the GHS-POP for population data, OSM for PGA data, and WC-ESA 2020 for green infrastructure data. Data preprocessing includes extracting city boundaries, population distribution, OSM PGA data, WC-ESA green coverage data, and calculating a walking distance matrix using the OSRM engine.

The study's findings highlight significant instability in the ranking of both geographical units and population subgroups when parameters of the accessibility indicators are slightly altered. This instability is observed across all three families of indicators (minimum distance, exposure, and per-person). For example, concerning the minimum distance indicator, the stability level decreases as the minimum size of the PGA increases. In some large cities, substantial variability in stability is observed, emphasizing the limitation of relying on a single parameterization. The analysis using targeting strategies (naive and most-disadvantaged) reveals that under-performing populations are often found in less densely populated areas, posing challenges for policy interventions. An examination of the weighted Gini indicator shows variable impacts of parameter changes on inequality levels, with some cities showing decreases, increases, or U-shaped patterns. When comparing institutional targets, the study finds limited interchangeability among indicators from different families or those measuring different aspects of green accessibility. Even among indicators aiming to capture similar forms of accessibility, substantial discrepancies are observed. This points to the critical need for a multi-dimensional framework in evaluating green accessibility.

The findings of this study challenge the common practice of using single indicators to assess urban green accessibility. The significant instability observed across various parameterizations and indicator types underscores the inadequacy of simplistic, single-indicator approaches. Relying on fixed parameters can lead to insufficient discrimination across areas and population subgroups, potentially hindering effective policy design. The study's emphasis on consistent under-performance in less densely populated areas further complicates policymaking, as these areas might not be sufficiently populated for targeted interventions. While acknowledging the limitations of simplistic ranking-based prioritization strategies, the authors highlight the need for a multi-indicator framework to address urban inequalities in green accessibility effectively. The multi-dimensional approach accounts for the nuances of diverse metrics and institutional targets, overcoming limitations of fixed-threshold structural indicators. The results emphasize the need for a more nuanced understanding of green accessibility that goes beyond single-indicator assessments. This multi-indicator approach helps identify consistently under-performing groups across multiple measures.

This study demonstrates the limitations of using single indicators to assess green accessibility in cities. The substantial instability in rankings and the limited interchangeability between various indicators highlight the crucial need for a multi-dimensional framework. Such a framework would enable a more comprehensive assessment of green provision, considering multiple forms of accessibility and accounting for the variability associated with different parameterizations and institutional targets. Future research should focus on incorporating additional data layers to enrich the characterization of urban green areas and systematically evaluate various data sources for constructing green accessibility indicators, while carefully considering the challenges of implementing these approaches at a large scale and engaging with complex real-world factors and participatory methods.

The study's main limitation relates to the completeness of green feature mapping in OpenStreetMap (OSM). To mitigate this, the authors implemented data quality checks, significantly reducing the initial sample size. However, the accessibility metrics still depend on OSM data quality, potentially introducing bias. The study also acknowledges limitations in the definition of PGAs, resulting from heterogeneous mapping standards across countries and climate zones. The interactive platform partially addresses this by allowing users to customize green feature types, but more comprehensive efforts are needed. Finally, the study's large geographical scope prevents a detailed consideration of real-world factors like green gentrification and citizen participation.