The study area could hardly be more instructive. Sintra, covering approximately 319 square kilometres in the Lisbon Metropolitan Area and home to around 385,000 people, is a municipality of stark internal contrasts. Its north-western quadrant is dominated by the Sintra–Cascais Natural Park, a protected mountainous landscape of roughly 14,590 hectares of forested terrain and cultural landmarks. Its eastern sector, by contrast, is a dense band of suburban development—the parishes of Algueirão-Mem Martins, Agualva e Mira-Sintra, Massamá e Monte Abraão, and Rio de Mouro—where high population densities coincide with a relative scarcity of green spaces larger than one hectare. The park’s rugged topography, with elevations reaching 514 metres, and its infrastructure geared toward nature tourism mean that much of it lies outside the routine activity spaces of lowland residents, particularly those without cars, training, or mobility. On paper, Sintra looks abundantly green; functionally, its eastern neighbourhoods are “green deserts.”

To measure this discrepancy rigorously, the team first built an inventory of leisure green spaces from the official municipal cadastre, retaining only sites that either offered functional amenities for active recreation—playgrounds, sports courts, or similar facilities—or met a minimum size threshold of one hectare, in line with World Health Organization recommendations and the United Kingdom’s Accessible Greenspace Standards. Crucially, the researchers then abandoned the conventional Euclidean buffer approach, in which access is judged by straight-line distance, in favour of network-based service area analysis performed in ArcGIS Pro Network Analyst. Accessibility was computed along the actual topological structure of the pedestrian street network, using walking time as the impedance attribute derived from segment lengths and an assumed average pedestrian speed of 4.5 kilometres per hour. Service areas were generated at three cumulative travel-time cutoffs—0–5, 5–10, and 10–15 minutes—producing pedestrian isochrones that capture the real barriers imposed by street connectivity, high-speed transit corridors, and Sintra’s fragmented road system.

The results are sobering. Approximately 183,000 residents live within a five-minute walk of a qualifying leisure green space, concentrated in well-served central areas such as Queluz, Agualva-Cacém, and Algueirão-Mem Martins. Another 80,000 fall within the 5-to-10-minute band, and only about 36,700 within the 10-to-15-minute interval. Beyond that threshold, roughly 82,000 inhabitants—about 21 percent of the municipal population—remain entirely outside the effective service area of any qualifying green space, a deficit that falls disproportionately on high-density eastern neighbourhoods including Tapada das Mercês, Mem Martins, and parts of Agualva-Cacém. By intersecting the accessibility isochrones with 2021 census population density data, classified into four categories using the Jenks natural breaks method, the team produced a sixteen-class spatial typology that exposes the sharpest mismatches: areas combining medium-to-high population density with low or null walkable access. Population counts were apportioned among statistical subsections intersected by service-area rings using areal weighting, with subsections whose overlap fell below 50 percent excluded to avoid spurious inflation—a conservative reconciliation designed to keep estimates honest and non-duplicated.

With the deficit map in hand, the researchers turned to identifying where new green spaces should go. Five criteria were selected and harmonised onto a common 0-to-100 suitability scale after projecting all layers to the ETRS89/Portugal TM06 coordinate system and rasterising at a uniform one-metre resolution. The criteria were proximity to existing leisure green spaces, buffer zones around highly populated areas, land-use suitability based on the 2018 Portuguese land cover map, proximity to schools as a proxy for the social dimension of access, and the regulatory footprint of the Sintra–Cascais Natural Park. Notably, land ownership was deliberately not used as an exclusion filter: at the strategic screening stage, the model aims to identify where new green space would yield the greatest gains, leaving tenure questions—acquisition, negotiation, or regulatory instruments—to a later adaptive-planning phase.

To weight these criteria objectively, the team applied the Analytic Hierarchy Process, the pairwise comparison method introduced by Thomas Saaty. Judgements were elicited from a focus group of eight participants, including municipal planning officers from the Câmara Municipal de Sintra and researchers specialised in urban ecology and spatial analysis, with individual matrices aggregated via the geometric mean. The resulting weights reveal a clear value orientation: demographic pressure, expressed as buffers around highly populated areas, received the highest weight at 29 percent, followed closely by the pedestrian accessibility gradient at 25 percent. Land-use suitability took 20 percent, proximity to schools 16 percent, and the Natural Park criterion just 10 percent—reflecting its regulatory rather than demand-driven character. Internal consistency was confirmed with a Consistency Ratio of 0.04, comfortably below the accepted 0.10 threshold, and a one-at-a-time sensitivity analysis perturbing each weight by ±10 percent produced only marginal changes and did not alter the ranking of priority sites, indicating robust model outputs.

The weighted overlay produced a composite suitability index normalised to 0–100 percent and classified into five equal-interval categories. Only areas exceeding 80 percent suitability were retained, filtered by a minimum size of one hectare and dissolved to eliminate fragmentation, yielding 96 candidate areas totalling 967.22 hectares with a mean size of 9.6 hectares. Built-up land accounted for only 5.74 hectares—about 0.6 percent of the selected area—demonstrating that the model overwhelmingly favoured non-urbanised land such as forests, agricultural plots, and abandoned shrubland, thereby minimising conflicts with existing residential uses. No qualifying sites emerged in the western parishes of São João das Lampas e Terrugem, Colares, or Almargem do Bispo, while the highest concentration appeared precisely where demographic pressure and accessibility deficits converge: Algueirão–Mem Martins and Agualva–Mira Sintra. From these candidates, the five largest continuous parcels were selected as priority intervention areas, ranging from 14.1 to 27 hectares, located primarily in Tapada das Mercês, Vale Mourão, and Chão de Meninos.

Network-based estimates suggest that establishing these five parks would improve walkable access to leisure green spaces for approximately 68,200 residents—a de-duplicated figure derived from a single combined service-area solve, since the sites’ catchments overlap spatially. The analysis also surfaced a cautionary tale about data latency. Site 4, Vale Mourão II, was scored as shrubland based on the 2018 land cover dataset, but between 2018 and 2026 a fully sealed open-air car park was built over part of the parcel. Because the conversion postdates the land cover input, the impervious surface is invisible to the model, and the site is rated more suitable than its current physical state warrants. The researchers flag this as an exemplar of the temporal-latency limitation of static land cover data, while noting that the low-intensity built use is reversible, so the parcel retains genuine potential for green-space conversion or partial renaturalisation—a form of adaptive urbanism in which even parking infrastructure can be reclaimed for nature.

Beyond its local findings, the study makes a broader methodological argument: nominal proximity is a poor proxy for experienced accessibility. Prior research in Porto and Kunming has shown that residents may live within the nominal fifteen-minute range yet face fragmented infrastructure and limited multimodal connectivity that undermine the equity ambitions of the fifteen-minute city. The Sintra results reinforce this picture, demonstrating that simple straight-line measures would mask precisely the deficits that matter most to vulnerable, car-dependent, or mobility-limited populations. The authors are candid about their model’s limitations: the constant walking speed does not capture slope-induced effort on steeper segments; the fifteen-minute threshold may overestimate access for elderly residents; pocket parks under half a hectare are underrepresented; and population density, the model’s demand proxy, is not equivalent to socioeconomic vulnerability. Fully substantiating environmental justice claims, they acknowledge, would require integrating income data or a composite Social Vulnerability Index directly into the multi-criteria framework—an extension their weighting structure can readily accommodate.

The practical payoff is a six-stage transferable framework: inventory and assess existing green spaces; model pedestrian access on real networks; identify equity-focused deficits by overlaying accessibility with demographic data; evaluate and prioritise candidate sites with AHP-weighted spatial suitability models; and pursue adaptive planning that verifies ground conditions and integrates selected sites into municipal plans as multifunctional green infrastructure. For Sintra specifically, the recommendations point toward Tapada das Mercês and the other “Category D” zones where medium-high density meets total accessibility deficit. For planners elsewhere, the message is that the fifteen-minute city will remain an aspirational slogan until municipalities replace area-based green space quotas with people-centred, network-based analysis—treating walkable access to nature not as a statistical by-product of protected landscapes, but as a fundamental determinant of urban health to be engineered, measured, and delivered.