Quantifying the Scaling Effects of Urban Green Infrastructure on Air Quality and Greenhouse Gas Dynamics: Insights from a Multi-Site Evaluation in Athens, Greece

Abstract

Urban green infrastructure (GI) provides a promising nature-based solution to mitigate urban air pollution, particularly fine particulate matter (PM), yet its quantifiable impacts across diverse urban settings remain insufficiently explored. This study investigates pollutant dynamics in Athens, Greece, a Mediterranean megacity characterized by high anthropogenic emissions and sparse green cover. A two-week monitoring campaign was conducted at four urban locations with vegetation density ranging from 5% to 100%. Concentrations of PM_2.5, PM₁₀, NO₂, CO₂, O₃, and VOCs were measured and analyzed using statistical approaches including ANOVA, ANCOVA, and regression models to normalize meteorological influences. Results showed consistent decreases in primary pollutants (PM_2.5, PM₁₀, NO₂, VOCs, CO₂) with increasing vegetation cover, while O₃ exhibited the expected inverse pattern due to reduced NO titration. Diurnal analyses revealed sharp peaks in PM and other pollutants during morning and evening rush hours in low vegetation corridors, contrasted with flatter profiles in greener sites. These findings demonstrate that even modest increases in green cover can dampen traffic-related pollution surges, reduce daily variability, and lower human exposure. The study highlights GI as a scalable and cost-effective strategy for particulate matter reduction and sustainable urban air quality improvement.

1. Introduction

Urban air pollution remains one of the most pressing environmental and public-health challenges of the 21st century, contributing substantially to morbidity and premature mortality worldwide [1]. Fine particulate matter (PM_2.5) alone accounts for millions of premature deaths annually, while co-occurring pollutants such as nitrogen dioxide (NO₂), ozone (O₃), and volatile organic compounds (VOCs) exert additional cardiopulmonary burdens [2]. In densely populated cities, pollutant levels are driven by traffic, industrial activity, and energy use but are further modulated by meteorology and urban morphology, which together govern dispersion, chemistry, and exposure patterns [3]. Also, combustion-related carbon dioxide (CO₂) emissions intensify the climate crisis, underscoring the need for strategies that jointly address local air quality and greenhouse-gas (GHG) mitigation [4].

Urban green infrastructure (GI), including parks, street trees, vegetated walls and roofs, has been advanced as a multifunctional, nature-based solution to improve air quality [5,6]. Vegetation can remove pollutants via dry deposition to leaf surfaces and stomatal uptake, and can indirectly improve air quality by altering microclimate and enhancing turbulent mixing [7]. However, effectiveness is highly context-dependent, varying with canopy structure, spatial distribution, and emission environments. In this research, green cover defined as the proportion of vegetated area including tree canopy and ground vegetation was used as an operational proxy for certain ecosystem functions of GI. Nevertheless, GI is conceptually broader, encompassing spatial configuration, ecological connectivity, and multifunctionality beyond surface coverage alone. Recent multi-scale urban modeling studies have further demonstrated that the air quality and microclimate benefits of GI depend strongly on canopy geometry, surrounding street-canyon morphology, and atmospheric flow regimes, reinforcing the need to evaluate GI performance within its specific urban context [8]. Notably, diverging hypotheses exist: in some street-canyon settings, vegetation may reduce ventilation and trap primary pollutants; O₃ responses can be non-monotonic due to reduced NO titration and biogenic VOC influences; and city-scale mass-balance arguments suggest benefits may be modest unless paired with emission reductions [9,10]. These debates highlight the importance of empirical, multi-site evidence that captures both primary and secondary pollutants under realistic urban conditions.

Mediterranean cities for this issue present an informative testbed; for example, recurrent heat waves, high photochemical activity, shallow boundary layers, and water scarcity can amplify pollution episodes and human exposure [11]. Athens, Greece, exemplifies these challenges, dense urban form, limited vegetation, extreme summer heat, and smog while offering heterogeneous green cover from traffic-dominated corridors to large parks, enabling analysis across a gradient of GI contexts [12].

Although numerous studies have examined GI impacts on microclimate and pollutant levels, many rely on single sites, short campaigns, or modeling frameworks, limiting generalizability across space, time, and meteorological regimes [13]. Few investigations simultaneously quantify the influence of quantified vegetation cover on multiple pollutant classes spanning primary emissions and secondary photochemical species while incorporating a GHG indicator such as CO₂ and explicitly normalizing for meteorology [14,15]. This gap motivates the present work.

A two-week monitoring campaign was conducted across four urban sites in Athens, spanning 5–100% green cover. Concentrations of PM_2.5, PM₁₀, NO₂, CO₂, O₃, and VOCs, together with on-site meteorological variables, were measured, and ANOVA, ANCOVA, and regression models were applied to control for meteorological variability. The objectives were to: (i) characterize spatial and diurnal pollutant patterns across contrasting GI settings; (ii) quantify relationships between vegetation cover and pollutant concentrations under meteorology-normalized conditions; and (iii) differentiate responses of primary versus secondary pollutants. The results indicate that increasing green cover is associated with lower concentrations of primary pollutants and dampened rush hour peaks, whereas O₃ exhibits the expected inverse response, consistent with reduced NO titration. These findings provide policy-relevant evidence for deploying GI as part of an integrated strategy for particulate matter reduction, exposure mitigation, and climate-resilient urban planning in Mediterranean cities.

2. Materials and Methods

2.1. Dataset Context & Spatial Baseline

The study was conducted in Athens, Greece (37°58′ N, 23°43′ E), the nation’s capital and largest metropolitan area, which lies in the southern Balkan Peninsula along the Saronic Gulf (Figure 1a). Situated within the Mediterranean Basin—a recognized global climate hotspot—the city experiences a hot-summer Mediterranean climate (Csa, Köppen–Geiger), with prolonged dry summers, frequent heatwaves above 35 °C, and mild, wet winters dominated by cyclonic activity [16]. Persistent summer anticyclonic conditions, characterized by strong solar radiation, shallow boundary layers, and limited precipitation, promote pollutant accumulation [17,18]. The specific case study area of Athens is shown in Figure 1b.

Athens is characterized by dense building configurations, narrow street canyons, and fragmented vegetation. While large parks are concentrated in peripheral districts, central areas feature only sparse and isolated greenery [19,20,21]. The Athens metropolitan area accommodates about 3.16 million residents [22] and is strongly shaped by vehicular traffic and industrial emissions, providing a representative setting to assess how GI influences air quality and GHG dynamics.

To systematically capture the influence of spatial heterogeneity in green cover, four representative monitoring sites were strategically selected to span a gradient of vegetation coverage and contrasting land-use types. Site selection criteria included:

• Representation of a wide range of vegetation densities to allow gradient-based statistical analysis.
• Diversity in surrounding land use typologies (open green space, academic campus, mixed residential/commercial, and industrial/traffic corridor).
• Proximity to emission sources of varying intensity, particularly high traffic roads and industrial zones.
• Logistical feasibility for repeated, time-specific sampling campaigns.

The four monitoring sites included (Figure 2):

• Syggrou Grove Park (S.G.P.)—a large urban park in Marousi neighborhood, characterized by mature trees and expansive lawns, isolated from major traffic corridors, representing the upper extreme of the vegetation gradient.
• Agricultural University of Athens (A.U.A.)—a landscaped academic campus with scattered mature trees and lawns, surrounded by moderate traffic activity.
• Pl. Karaiskaki Square (PL.K.S.)—a dense urban square with limited vegetation (few street trees and shrubs), adjacent to mixed-use commercial and residential developments.
• Athens National Road (A.N.R.)—a heavily trafficked industrial corridor with negligible vegetation cover, representing the lower extreme of the gradient.

This configuration ensured a clear and quantifiable GI gradient across sites, enabling analysis of both spatial contrasts and incremental vegetation effects on pollutant behavior.

Table 1. Overview of the four monitoring sites in Athens, Greece, including land-use type, dominant vegetation, estimated green cover percentage, and coordinates.

Location	Land Use Category	Vegetation Type	Green Cover (%), Mean ± 95% CI	Latitude	Longitude
1. Syggrou Grove Park (S.G.P.)	Open green space	Mature trees & lawn	95 ± 3%	38.064064	23.823989
2. Agricultural University of Athens (A.U.A.)	Academic campus	Lawns & scattered trees	65 ± 4%	37.986746	23.707899
3. Pl. Karaiskaki Square (PL.K.S.)	Mixed residential/commercial	Few shrubs & street trees	35 ± 5%	37.985339	23.721774
4. Athens National Road (A.N.R.)	Industrial/traffic corridor	Minimal vegetation	8 ± 2%	38.014500	23.718750

provides detailed site characteristics, including land-use category, vegetation type, estimated green cover percentage, and precise geographic coordinates.

Green-cover percentages were estimated using high-resolution land-cover maps (updated to 2025) combined with in situ field surveys in June 2025. Field estimation employed a modified point intercept method adapted for urban environments, which accounted for both tree canopy and ground vegetation to produce consistent percentage cover estimates. A 95% confidence interval (CI) was calculated based on the variance across the point-intercept sampling grid (n = 200 points per site) and cross-validation with the land-cover classification outputs, and the reported values are presented as mean ± 95% CI in Table 1.

Regional air quality assessments from the Greek Ministry of Environment and Energy and the European Environment Agency (EEA) indicate that Athens frequently experiences pollutant concentrations near or above recommended health-based thresholds. Annual mean PM_2.5 levels at central urban monitoring stations typically range between 12–18 µg·m⁻³, exceeding the WHO (2021) [26] guideline of 5 µg·m⁻³, while PM₁₀ commonly ranges from 25–40 µg·m⁻³ under warm-season conditions. Episodic exceedances of the EU target value for O₃ (120 µg·m⁻³, 8-h) are recurrent during heatwave and stagnant-air episodes in summer, and traffic corridors often exhibit NO₂ annual means close to or exceeding the EU annual limit value of 40 µg·m⁻³. These background conditions provide a meaningful atmospheric context for examining how spatial variation in vegetation cover modulates pollutant exposure under real-world Mediterranean urban conditions.

2.2. Sampling Design

The monitoring campaign was carried out over two consecutive weeks (17–29 June 2025), a period corresponding to early summer in Athens, when meteorological conditions and anthropogenic activity patterns are particularly relevant for studying urban air quality. This timing was chosen deliberately for three reasons:

• Meteorological influence—Early summer conditions in Athens are characterized by recurrent high-pressure systems and strong diurnal heating, which produce pronounced day–night variation in boundary-layer structure. Morning and evening stable layers can limit vertical dispersion, while midday mixing layers enhance turbulent transport. This pronounced diurnal contrast provides a suitable framework for examining how vegetation modulates pollutant dispersion under different mixing states. It does not imply that summer boundary-layer heights are uniformly lower than in winter, but rather that diurnal variability is stronger during this period.
• Vegetation status—Urban vegetation is in peak physiological condition during this period, maximizing its capacity for pollutant removal through processes such as deposition, stomatal uptake, and shading-induced microclimatic cooling.
• Representative anthropogenic patterns—Traffic flows, energy consumption, and human outdoor activity follow consistent seasonal patterns, allowing for the isolation of diurnal and weekday–weekend variations without interference from transitional seasonal shifts.

The experimental design incorporated both spatial and temporal replication to enable robust statistical comparisons. Spatially, the four monitoring sites (Section 2.1) were selected to represent distinct points along a green cover gradient. Temporally, each site was monitored at four fixed daily intervals:

• 08:00 (morning rush hour)—corresponding to peak commuter traffic and typically shallow boundary-layer depths that limit vertical dispersion.
• 13:00 (midday maximum solar radiation)—favoring turbulent mixing and photochemical ozone formation.
• 18:00 (evening rush hour)—a second commuter traffic peak, often coupled with residual atmospheric stability before sunset.
• 22:00 (post traffic night period)—reduced traffic emissions and typically enhanced horizontal dispersion but still limited vertical mixing under stable nighttime conditions.

These time slots were applied consistently across all sites to allow direct comparison of diurnal patterns and to facilitate the detection of interactions between time of day, site characteristics, and meteorological variables.

Each measurement session lasted a minimum of five minutes of continuous sampling, during which instruments were positioned at a standardized height of 170 cm above ground level to approximate the average human breathing zone. This placement ensured that recorded pollutant concentrations were directly relevant to population exposure assessments. To minimize the influence of transient plumes (e.g., from passing vehicles), instruments were located at least 5 m away from active traffic lanes while still being representative of the surrounding microenvironment. The design also accounted for weekday–weekend contrasts by including both Tuesday–Friday (weekday) and Saturday–Sunday (weekend) measurements. This allowed the analysis of how pollutant dynamics vary with changes in anthropogenic activity patterns, particularly traffic intensity and industrial operations, which typically decline on weekends.

2.3. Measured Parameters

The monitoring program targeted a comprehensive set of atmospheric pollutants representing both primary emissions (directly emitted to the atmosphere) and secondary pollutants (formed via photochemical or atmospheric transformation processes). This combination was selected to capture the full spectrum of urban air-quality dynamics relevant to health, climate, and environmental policy.

The measured pollutants and their significance were as follows:

• Particulate Matter ≤ 2.5 μm (PM_2.5)—Fine inhalable particles predominantly originating from combustion processes, including vehicle exhaust, industrial activities, and biomass burning. Due to their small size, PM_2.5 can penetrate deep into the alveolar region of the lungs, with established links to cardiovascular and respiratory morbidity [27]. Concentrations were expressed in mg·m⁻³.
• Particulate Matter ≤ 10 μm (PM₁₀)—Coarse inhalable particles generated from mechanical abrasion, resuspension of road dust, and natural sources such as pollen. Although generally less respirable than PM_2.5, they contribute substantially to particulate load and can exacerbate respiratory conditions [28]. Concentrations were expressed in mg·m⁻³.
• Carbon Dioxide (CO₂)—A greenhouse gas emitted mainly from fossil-fuel combustion in vehicles, heating systems, and industrial processes [29]. While CO₂ is non-toxic at ambient concentrations, its level serves as a proxy for combustion activity and urban metabolic intensity [30]. Concentrations were expressed in ppm.
• Nitrogen Dioxide (NO₂)—A key traffic-related pollutant emitted from high-temperature combustion, particularly in diesel engines; NO₂ is both a primary pollutant and an ozone precursor central to urban photochemistry [31]. Concentrations were expressed in ppm.
• Ozone (O₃)—A secondary pollutant produced via photochemical reactions between NO_x and volatile organic compounds in the presence of sunlight. Beneficial in the stratosphere, O₃ at ground level is a respiratory irritant and oxidative stressor [32,33]. Concentrations were expressed in ppm.
• Volatile Organic Compounds (VOCs)—A diverse class of organic compounds emitted from traffic exhaust, fuel evaporation, solvent use, and vegetation [34]. VOCs act as essential precursors for ozone and secondary organic aerosol [35]. Concentrations were expressed in ppm.

Meteorological parameters were recorded in parallel to provide contextual and explanatory variables for pollutant behavior, including: air temperature (°C), relative humidity (%), and general weather conditions (sunny, partly cloudy, cloudy, rainy) noted at the time of each measurement.

To preserve fidelity to device outputs and enable transparent replication, units were reported in the instruments’ native formats: PM_2.5 and PM₁₀ in mg·m⁻³; NO₂, O₃, CO₂, and VOCs in ppm. The combined pollutant and meteorological datasets enabled meteorology-normalized analyses, where the influence of short-term weather fluctuations was statistically controlled [36], allowing intrinsic site characteristics particularly green infrastructure coverage to be isolated.

2.4. Instrumentation and Calibration

To ensure precision, reliability, and reproducibility, all measurements were conducted using portable, high-accuracy monitoring instruments selected for their proven performance in multi-site, short-duration urban field campaigns. Instrument choice was guided by three key criteria:

• Compatibility with multiple pollutant sensors to reduce logistical complexity.
• Compliance with internationally recognized air quality monitoring standards, EU Air Quality Directive.
• Ease of transport and rapid setup to accommodate frequent daily relocation between sites.

A.

Air Quality Monitoring

Primary air pollutant measurements—including PM_2.5, PM₁₀, CO₂, NO₂, O₃, and VOCs—were performed using the Aeroqual S-Series 500 portable air quality analyzer (Aeroqual Ltd., Auckland, New Zealand) [37] (Figure 3). This instrument features:

• Modular sensor heads allowing rapid swapping between pollutant-specific electrochemical and optical modules without recalibration downtime.
• Laser-based optical particle counting for PM_2.5 and PM₁₀, ensuring accurate mass-concentration estimates in µg·m⁻³.
• Electrochemical gas sensors for NO₂, O₃, and VOCs, providing high sensitivity at low ambient concentrations.
• Non-dispersive infrared (NDIR) detection for CO₂, delivering stable, interference-resistant ppm readings.

B.

Meteorological Monitoring

Meteorological parameters—air temperature (°C), relative humidity (RH%)—were recorded using the TSI VelociCalc 8720 Velocity Meter (TSI Incorporated, Shoreview, MN, USA) [38] (Figure 4). This instrument was chosen for:

• Integrated temperature and humidity sensors to capture environmental variables influencing pollutant dispersion and chemical transformation.
• Ruggedized field housing suitable for outdoor deployment in variable summer conditions.

C.

Calibration Procedures

To minimize systematic bias and maintain data comparability, calibration was carried out daily at the A.U.A. base site using pollutant-free filtered air (zero calibration) and certified span gases. Additionally, post-campaign calibration checks were performed to detect any potential sensor drift over the two-week measurement period. No significant deviations were observed, confirming instrument stability throughout the fieldwork. The combined use of calibrated high-precision analyzers and daily quality control routines ensured that all recorded data met reproducibility requirements, allowing for robust statistical analyses and inter-site comparisons.

To account for instrument uncertainty and environmental influences on sensor response, additional data quality control procedures were implemented. Manufacturer-reported accuracy for the PM_2.5 and PM₁₀ optical module is ±(3–5%) under standard operating conditions, and ±(2–4%) for NO₂, O₃, VOCs, and CO₂ sensor heads. Prior to each sampling session, zero-air calibration was performed to verify baseline stability, and periodic on-site cross-checks were conducted by colocating monitors for short intervals to detect potential sensor drift. Temperature and relative humidity were recorded concurrently to identify conditions under which optical scattering or electrochemical sensor sensitivity could vary. Direct solar exposure was avoided by shading instruments during operation, and equipment was positioned at a standardized breathing-zone height to reduce surface heating effects. No systematic drift was observed across the campaign period, and the combination of calibration control, environmental shielding, and cross-validation ensures that the reported pollutant differences reflect site conditions rather than measurement artifacts.

2.5. Sampling Procedure

At each monitoring location, the air-quality and meteorological instruments were deployed to capture representative conditions of the surrounding air mass. Prior to each session, instruments were allowed to stabilize for 2–3 min. Pollutant data were logged as follows:

• PM_2.5 and PM₁₀: stabilized end-interval values.
• CO₂: averaged across the sampling interval.
• NO₂, O₃, VOCs: continuous logging, averaged post hoc.

Contextual notes (e.g., unusual traffic, dust events) were manually recorded alongside measurements. Field data were transcribed into standardized logs and entered into a structured Excel database. An independent reviewer cross-checked all entries before statistical analysis.

Quality assurance included session repetition in case of anomalies, synchronization of timestamps, and immediate outlier screening after each field day.

2.6. Statistical Analysis

Meteorology-normalized pollutant–vegetation relationships were evaluated using a combination of ANOVA, ANCOVA, and ordinary least-squares regression, following standard procedures in atmospheric environmental data analysis. One- and two-way ANOVA were used to assess the effects of site and time of day, while ANCOVA was applied to account for meteorological covariates (temperature and humidity) to reduce confounding by short-term atmospheric variability. When significant differences were identified (p < 0.05), post hoc pairwise comparisons were conducted using Tukey’s HSD. These statistical procedures allowed for controlled comparison of pollutant concentrations across the green-cover gradient under comparable meteorological conditions.

3. Results

3.1. Statistical Summary of Monitoring Data

3.1.1. Patterns and Relationships Among Pollutants and Meteorological Variables

The monitoring campaign yielded a comprehensive dataset comprising 288 individual measurements collected over a 13-day period (17–29 June 2025), during the early summer season in Athens. Sampling was conducted at four urban locations representing a gradient of green infrastructure coverage and urban morphological conditions. Each site was monitored at four fixed times per day, 08:00 (morning rush hour), 13:00 (midday peak solar radiation), 18:00 (evening rush hour), and 22:00 (post-traffic night period), allowing the dataset to capture both short-term diurnal variability and broader weekday–weekend dynamics. Measurements were accompanied by on-site meteorological data collection, including T_air and RH, to contextualize pollutant fluctuations under varying atmospheric conditions.

Climatic conditions during the study period were characteristic of Mediterranean summer meteorology: mean daily T_air was 30.2 °C, with peaks exceeding 34 °C during midday, and mean RH averaged 36.7%, frequently dropping below 30% during the hottest hours. Such conditions favored atmospheric stability and reduced vertical mixing, enhancing pollutant accumulation under low winds, while simultaneously promoting photochemical activity leading to ozone formation. The seasonal timing is particularly relevant for understanding vegetation influences, as green infrastructure was at peak physiological activity, maximizing pollutant uptake via stomatal conductance. No precipitation events occurred during the monitoring period, and, therefore, rain-driven wet deposition did not influence pollutant concentrations during the campaign. No Saharan or Middle Eastern dust intrusion events were detected during the monitoring period. This was confirmed using daily dust load and aerosol optical depth forecasts from CAMS regional atmospheric composition analyses and SKIRON dust transport model outputs for Greece. Long-range dust transport is a known contributor to elevated PM concentrations in Athens during spring and autumn; however, no such events occurred between 17–29 June 2025. Therefore, the particulate matter levels observed in this study primarily reflect local urban emission sources and microenvironmental dispersion conditions rather than regional dust contributions.

The Pearson correlation matrix (Figure 5) highlights several significant relationships (p < 0.05). O₃ exhibited strong negative correlations with NO₂ (r = −0.61 *) and VOCs (r = −0.76 ***). This aligns with established photochemical dynamics, where NO₂ acts both as a precursor and as a sink for O₃ through titration, particularly under high NO conditions. The negative VOC–O₃ relationship reflects the interplay of traffic emissions: while VOCs contribute to O₃ formation, their co-occurrence with elevated NO₂ in traffic-heavy sites can lead to net ozone suppression through NO scavenging.

Meteorological influences were also evident. T_air showed a weak positive correlation with O₃ (r = 0.16), consistent with the temperature dependence of photochemical reactions and volatilization of precursors. By contrast, RH correlated negatively with O₃ (r = −0.62 **), likely due to increased cloudiness or aerosol water uptake under humid conditions that reduce photolysis rates. RH also showed weak positive associations with PM_2.5 (r = 0.10) and PM₁₀ (r = 0.05), reflecting potential hygroscopic particle growth and reduced dispersion under humid, stable atmospheres.

From an urban morphology perspective, these patterns are consistent with expectations for a Mediterranean megacity like Athens: traffic-dominated corridors exhibited high primary pollutant concentrations, whereas vegetated areas diluted emissions through enhanced dispersion and altered secondary pollutant dynamics, notably supporting higher midday O₃ levels by reducing NO titration.

The correlation analysis confirms the interdependence of pollutants in the urban atmosphere, associated with a mix of primary emissions, secondary chemical transformations, and meteorological modulation. This baseline understanding sets the stage for more granular analysis of spatial contrasts, temporal cycles, and the mitigating role of green infrastructure, as explored in subsequent sections.

3.1.2. Average Pollutant Levels by Site

Marked spatial heterogeneity was evident across the four sites (Figure 6). The greenest location, Syggrou Grove Park (S.G.P.), consistently exhibited the lowest means for primary pollutants. PM_2.5 averaged ≈ 4.20 µg·m⁻³ at S.G.P., compared with ≈11.00–13.00 µg·m⁻³ at the other sites, i.e., roughly two-thirds lower. PM₁₀ showed a comparable pattern, with S.G.P. registering about half the levels observed elsewhere (µg·m⁻³). For CO₂, differences were modest; S.G.P. (645 ppm) was only slightly lower (on the order of 8–9%) than the remaining sites. The largest contrasts arose for NO₂ (ppm), with concentrations at S.G.P. substantially reduced relative to the traffic-dominated corridors (A.N.R., PL.K.S.), on the order of 70–90% lower.

In contrast, O₃ peaked at greener sites, 0.022 ppm at S.G.P. and 0.018 ppm at A.U.A. and was minimal at A.N.R. (0.002 ppm), consistent with NO titration suppressing ozone in high-NO environments despite abundant precursors. VOCs were highest at PL.K.S. (0.255 ppm) and A.N.R. (0.245 ppm) and lowest at S.G.P. (0.035 ppm), about six- to seven-fold lower than at traffic-influenced sites, indicating dominant control by local emissions with vegetation exerting indirect effects via dispersion and microclimatic modulation.

Site-wise differences were supported by omnibus tests (one-way ANOVA, p < 0.05; see Section 3.2.1), with pairwise contrasts evaluated using Tukey’s HSD.

3.1.3. Distribution of Pollutant Levels by Site

The boxplots in Figure 7 illustrate both the central tendency and variability of pollutant concentrations at each site. Boxes represent the interquartile range (IQR; Q1–Q3), center lines mark the median, and whiskers extend to 1.5× IQR from each quartile; observations beyond these limits are plotted as outliers (points). A.N.R. and PL.K.S. display noticeably wider IQRs for PM_2.5, PM₁₀, and NO₂, indicating greater short-term variability, plausibly driven by fluctuating traffic volumes, intermittent local emissions, and changing meteorological conditions. Such variability implies more frequent short-lived—but potentially high—concentration peaks that may be obscured by daily means. In contrast, S.G.P. exhibits the smallest variability for most pollutants, consistent with a more stable microenvironment characterized by extensive vegetation, reduced local emissions, and enhanced dispersion.

3.2. Spatial Statistical Significance & Green Infrastructure Links

3.2.1. Effect of Site Location on Pollutant Concentrations (ANOVA)

A one-way ANOVA across the four sites (df = 3, 284) showed that site was a significant factor for all pollutants (Figure 8). Exact test statistics were:

• PM_2.5: F(3, 284) = 22.51, p = 4.19 × 10⁻¹³
• PM₁₀: F(3, 284) = 24.60, p = 3.52 × 10⁻¹⁴
• NO₂: F(3, 284) = 16.71, p = 5.03 × 10⁻¹⁰
• CO₂: F(3, 284) = 24.50, p = 3.96 × 10⁻¹⁴
• VOCs: F(3, 284) = 42.67, p = 8.6 × 10⁻²³
• O₃: F(3, 284) = 29.14, p = 1.87 × 10⁻¹⁶

Effects were strongest for VOCs and O₃, followed by PM₁₀/CO₂ and PM_2.5, with NO₂ still clearly significant. Directionally, results align with the GI gradient: O₃ is higher at greener locations (consistent with reduced NO titration), whereas primary pollutants (PM_2.5, PM₁₀, NO₂, VOCs) are elevated at traffic-dominated sites.

3.2.2. Post Hoc Tukey HSD Comparisons

Post hoc Tukey HSD tests resolved the specific site pairs responsible for the omnibus significance (Δ denotes mean (Group1) − mean (Group2); negative values indicate lower means in Group1).

• PM_2.5 (µg·m⁻³): S.G.P. was lower than A.N.R. (Δ = −8.9, p < 0.001) and PL.K.S. (Δ = −6.2, p < 0.001); A.U.A. was also lower than A.N.R. (Δ = −6.6, p < 0.001).
• PM₁₀ (µg·m⁻³): S.G.P. was lower than A.N.R. (Δ = −15.5, p < 0.001), A.U.A. (Δ = −11.1, p < 0.001), and PL.K.S. (Δ = −11.6, p < 0.001).
• NO₂ (ppm): S.G.P. was lower than A.N.R. (Δ = −0.024, p < 0.001) and PL.K.S. (Δ = −0.025, p < 0.001); A.N.R. was lower than A.U.A. (Δ = −0.0188, p = 0.001); A.U.A. and S.G.P. did not differ (n.s., p = 0.617).
• O₃ (ppm): S.G.P. exceeded A.N.R. (Δ = +0.0225, p < 0.001); greener sites (S.G.P., A.U.A.) generally exceeded traffic-dominated locations, consistent with reduced NO titration.
• CO₂ (ppm): S.G.P. was lower than A.N.R. (Δ = −80.1, p < 0.001), A.U.A. (Δ = −49.2, p < 0.001), and PL.K.S. (Δ = −72.6, p < 0.001).
• VOCs (ppm): Lowest at S.G.P., differing from A.N.R. (Δ = −0.218, p < 0.001) and PL.K.S. (Δ = −0.230, p < 0.001); the contrast with A.U.A. was not significant (Δ = −0.024, n.s.).
• Day-type (Weekday − Weekend): PM_2.5 (µg·m⁻³) Δ = −2.7, p = 0.042; NO₂ (ppm) Δ = −0.0098, p = 0.035; O₃ (ppm) Δ = −0.0072, p = 0.040; PM₁₀, CO₂, and VOCs: n.s.

Although some day-type contrasts are statistically significant, their magnitudes (e.g., PM_2.5 ≈ −2.7 µg·m⁻³) are small relative to site contrasts (e.g., PM_2.5 reductions of 6–9 µg·m⁻³ at S.G.P.), indicating that spatial factors linked to green infrastructure/land-use dominate practical exposure differences in this dataset (

Table 2. Tukey HSD pairwise comparisons by site and day-type. Δ = mean (Group1) − mean (Group2); 95% CI in brackets; significance: * p < 0.05, *** p < 0.001; n.s. = not significant. Units: PM_2.5/PM₁₀ (µg·m⁻³); CO₂, NO₂, O₃, VOCs (ppm).

Group1	Group2	Pollutant	Factor	Δ (Mean Difference)	95% CI	Sig.
A.N.R.	A.U.A.	PM2.5	Site	−2.3	[−5.4, 0.7]	n.s.
A.N.R.	PL.K.S.	PM2.5	Site	−2.7	[−5.7, 0.4]	n.s.
A.N.R.	S.G.P.	PM2.5	Site	−8.9	[−12.0, −5.9]	***
A.U.A.	PL.K.S.	PM2.5	Site	−0.3	[−3.4, 2.7]	n.s.
A.U.A.	S.G.P.	PM2.5	Site	−6.6	[−9.6, −3.5]	***
PL.K.S.	S.G.P.	PM2.5	Site	−6.2	[−9.3, −3.2]	***
Weekday	Weekend	PM2.5	Day-Type	−2.7	[−5.2, −0.1]	*
A.N.R.	A.U.A.	PM10	Site	−4.4	[−9.5, 0.7]	n.s.
A.N.R.	PL.K.S.	PM10	Site	−3.9	[−9.0, 1.2]	n.s.
A.N.R.	S.G.P.	PM10	Site	−15.5	[−20.6, −10.4]	***
A.U.A.	PL.K.S.	PM10	Site	+0.5	[−4.6, 5.6]	n.s.
A.U.A.	S.G.P.	PM10	Site	−11.1	[−16.2, −6.0]	***
PL.K.S.	S.G.P.	PM10	Site	−11.6	[−16.7, −6.5]	***
Weekday	Weekend	PM10	Day-Type	−1.2	[−5.8, 3.4]	n.s.
A.N.R.	A.U.A.	CO2	Site	−30.9	[−58.5, −3.36]	*
A.N.R.	PL.K.S.	CO2	Site	−7.5	[−35.1, 20.1]	n.s.
A.N.R.	S.G.P.	CO2	Site	−80.1	[−107.6, −52.5]	***
A.U.A.	PL.K.S.	CO2	Site	+23.4	[−4.14, 51.0]	n.s.
A.U.A.	S.G.P.	CO2	Site	−49.2	[−76.7, −21.6]	***
PL.K.S.	S.G.P.	CO2	Site	−72.6	[−100.1, −45.0]	***
Weekday	Weekend	CO2	Day-Type	−20.6	[−44.7, 3.58]	n.s
A.N.R.	A.U.A.	NO2	Site	−0.0188	[−0.0306, −0.0069]	***
A.N.R.	PL.K.S.	NO2	Site	+0.0008	[−0.0110, 0.0127]	n.s
A.N.R.	S.G.P.	NO2	Site	−0.0242	[−0.0360, −0.0123]	***
A.U.A.	PL.K.S.	NO2	Site	+0.0196	[0.0077, 0.0314]	***
A.U.A.	S.G.P.	NO2	Site	−0.0054	[−0.0173, 0.0064]	n.s
PL.K.S.	S.G.P.	NO2	Site	−0.0250	[−0.0368, −0.0132]	***
Weekday	Weekend	NO2	Day-Type	−0.0098	[−0.019, −0.0007]	*
A.N.R.	A.U.A.	O3	Site	+0.0167	[0.0092, 0.0242]	***
A.N.R.	PL.K.S.	O3	Site	+0.0033	[−0.0042, 0.0108]	n.s.
A.N.R.	S.G.P.	O3	Site	+0.0225	[0.0150, 0.0300]	***
A.U.A.	PL.K.S.	O3	Site	−0.0133	[−0.0208, −0.0058]	***
A.U.A.	S.G.P.	O3	Site	+0.0058	[−0.0017, 0.0133]	n.s.
PL.K.S.	S.G.P.	O3	Site	+0.0192	[0.0117, 0.0267]	***
Weekday	Weekend	O3	Day-Type	−0.0072	[−0.014, −0.0004]	*
A.N.R.	A.U.A.	VOCs	Site	−0.193	[−0.264, −0.122]	***
A.N.R.	PL.K.S.	VOCs	Site	+0.0125	[−0.0584, 0.0834]	n.s.
A.N.R.	S.G.P.	VOCs	Site	−0.218	[−0.288, −0.147]	***
A.U.A.	PL.K.S.	VOCs	Site	+0.206	[0.135, 0.277]	***
A.U.A.	S.G.P.	VOCs	Site	−0.0242	[−0.0951, 0.0468]	n.s.
PL.K.S.	S.G.P.	VOCs	Site	−0.230	[−0.301, −0.159]	***
Weekday	Weekend	VOCs	Day-Type	−0.046	[−0.123, 0.030]	n.s.

).

3.2.3. Meteorology-Normalized Mean Concentrations of Air Pollutants by Site

To isolate vegetation effects from meteorology, it estimated an ANCOVA with site as a fixed factor; time of day (08:00, 13:00, 18:00, 22:00) and day type (weekday/weekend) as categorical factors; and air temperature (T_air, °C) and relative humidity (RH,%) as continuous covariates. From this model we derived estimated marginal means (EMMs) per site, evaluated at the sample means of T_air and RH and averaged over the levels of the categorical factors. Figure 9 reports these EMMs using Syggrou Grove Park (S.G.P.; ~95% green cover) as the reference; pairwise differences are expressed as Δ = EMM (site) − EMM (S.G.P.).

Even after normalization, S.G.P. retained the cleanest profile for PM_2.5, PM₁₀, CO₂, NO₂, and VOCs, while O₃ remained higher at greener locations—consistent with reduced NO titration:

• PM_2.5 (µg·m⁻³): S.G.P. 6.90; Δ(A.N.R.) +4.80, Δ(A.U.A.) +5.20, Δ(PL.K.S.) +3.81.
• PM₁₀ (µg·m⁻³): S.G.P. 14.17; Δ(A.N.R.) +8.58, Δ(A.U.A.) +8.75, Δ(PL.K.S.) +7.49.
• CO₂ (ppm): S.G.P. 655.14; Δ(A.N.R.) +42.97, Δ(A.U.A.) +37.08, Δ(PL.K.S.) +50.55.
• NO₂ (ppm): S.G.P. 0.00837; Δ(A.N.R.) +0.00990, Δ(A.U.A.) +0.00143, Δ(PL.K.S.) +0.01644.
• O₃ (ppm): S.G.P. 0.01803; Δ(A.N.R.) −0.01096, Δ(A.U.A.) −0.00315, Δ(PL.K.S.) −0.01217.
• VOCs (ppm): S.G.P. 0.08467; Δ(A.N.R.) +0.09716, Δ(A.U.A.) −0.00079, Δ(PL.K.S.) +0.15664.

These levels were under comparable meteorological conditions, consistent with vegetation-driven mechanisms (deposition, gaseous uptake, enhanced dispersion). The higher O₃ at greener sites persists after normalization, aligning with the expected reduction in NO titration where primary NO_x is lower.

3.2.4. Green Cover–Pollutant Relationships and Quantification of the Green Infrastructure Effect

To isolate vegetation influences across the urban gradient, meteorology-normalized site means (ANCOVA EMMs) were regressed against site-level green cover (%) (

Table 3. Green-cover effect sizes from ANCOVA-normalized regressions (EMMs): slopes per +10 percentage-points of green cover with 95% confidence intervals (R² shown). Negative slopes indicate lower concentrations with higher green cover. Units: PM_2.5/PM₁₀ in µg·m⁻³; CO₂, NO₂, O₃, and VOCs in ppm.

Pollutant	Slope per +10%	95% CI	R2
PM2.5	−0.450	[−0.639, −0.261]	0.508
PM10	−0.840	[−1.20, −0.483]	0.584
NO2	−0.00155	[−0.00220, −0.00090]	0.579
CO2	−4.93	[−7.02, −2.84]	0.688
VOCs	−0.0155	[−0.0221, −0.00898]	0.575
O3	+0.00144	[+0.00098, +0.00190]	0.845

, Figure 10). In the ANCOVA, time of day (08:00, 13:00, 18:00, 22:00) and day type (weekday/weekend) were treated as factors, while air temperature (T_air, °C) and relative humidity (RH, %) were continuous covariates; EMMs were evaluated at the sample means of T_air and RH and averaged over factor levels. The green-cover gradient spanned A.N.R. (7.5%), PL.K.S. (35%), A.U.A. (65%), and S.G.P. (95%).

Across pollutants, slopes were negative for PM_2.5, PM₁₀, NO₂, CO₂, and VOCs, and positive for O₃, indicating progressively cleaner primary-pollutant conditions with increasing green cover. Reporting slopes per +10 percentage points of green cover (mean ± 95% CI; R²):

• PM_2.5 (µg·m⁻³): −0.450 (95% CI −0.639, −0.261), R² = 0.508.
• PM₁₀ (µg·m⁻³): −0.840 (95% CI −1.20, −0.483), R² = 0.584.
• NO₂ (ppm): −0.00155 (95% CI −0.00220, −0.00090), R² = 0.579.
• CO₂ (ppm): −4.93 (95% CI −7.02, −2.84), R² = 0.688.
• VOCs (ppm): −0.0155 (95% CI −0.022, −0.0090), R² = 0.575.
• O₃ (ppm): +0.00144 (95% CI +0.00098, +0.00190), R² = 0.845.

These slopes provide a policy-relevant metric of the GI effect size. For example, moving from 7.5% → 95% green cover (+87.5%, i.e., 8.75 × 10%) corresponds to an expected change of approximately −3.94 µg·m⁻³ (PM_2.5), −7.35 µg·m⁻³ (PM₁₀), −0.0136 ppm (NO₂), −43.1 ppm (CO₂), −0.136 ppm (VOCs), and +0.0126 ppm (O₃). Overall, the results indicate that increments in green cover are associated with material reductions in primary pollutants and CO₂, while O₃ tends to increase modestly at greener sites, consistent with reduced NO titration under photochemically active conditions.

It should be noted that regression slopes and R² values were derived from the full pooled dataset (n = 288) under ANCOVA-based normalization, rather than from site-averaged values, ensuring that the reported trends reflect observation-level variability rather than four-point regression fits.

3.3. Temporal Dynamics & Combined Spatio-Temporal Visualizations

3.3.1. Diurnal Patterns of Pollutant Concentrations by Site

The diurnal profiles of pollutant concentrations across the four monitoring sites exhibit clear temporal signatures shaped by both anthropogenic activity and atmospheric processes (Figure 11). For primary pollutants (PM_2.5, PM₁₀, CO₂, NO₂, VOCs), a bimodal cycle was observed with peaks at 08:00 and 18:00. These peaks were most pronounced at A.N.R. and PL.K.S., where morning concentrations exceeded midday levels by >40% for PM_2.5 and NO₂, consistent with rush-hour emissions and shallow early-morning mixing depths.

At 13:00, primary pollutant concentrations declined markedly, especially at S.G.P., where PM_2.5 reached ≈8 µg·m⁻³ and NO₂ 0.010 ppm, respectively. These midday reductions are consistent with enhanced turbulent mixing under strong solar heating and photolytic NO₂ loss. CO₂ followed a similar but less pronounced pattern, with elevated morning (~760 ppm at A.N.R.) and evening (~745 ppm) values linked to traffic and human activity, and a midday minimum (~695 ppm at S.G.P.) reflecting improved dispersion and reduced emission intensity.

VOCs closely tracked NO₂, with clear peaks at 08:00 and 18:00 and an evening rebound at A.N.R. (~0.29 ppm), consistent with reduced mixing heights after sunset and persistent post-work traffic emissions.

In contrast, O₃ displayed an inverse diurnal cycle, with midday maxima and nocturnal minima. Highest midday peaks occurred at S.G.P. (0.030 ppm) and A.U.A. (0.025 ppm), where lower NO₂ reduced titration losses. At A.N.R., midday O₃ reached only 0.018 ppm, indicative of NO_x-rich conditions suppressing ozone.

By 22:00, primary pollutants remained elevated at traffic-influenced sites, whereas O₃ declined sharply across all locations due to the cessation of photochemical production and continued NO titration. The enclosed street morphology at A.N.R. and PL.K.S. likely limited nighttime dispersion, sustaining higher nocturnal PM, NO₂, and VOCs relative to greener sites.

Time-of-day effects were statistically significant for all pollutants (p < 0.05). Tukey HSD comparisons indicated that differences between the 08:00/18:00 peaks and the 13:00 minima were significant for NO₂, PM_2.5, and PM₁₀ in high-traffic locations, and comparatively muted in green-dominated areas. Overall, these diurnal patterns indicate that green infrastructure not only lowers mean concentrations but also dampens diurnal amplitude, reducing peak exposure during the busiest periods.

3.3.2. Spatio-Temporal and Hourly Variations in Pollutant Concentrations Across Sites

The spatio-temporal heatmaps (Figure 12) illustrate how pollutant levels vary jointly by site and time of day, revealing clear interactions between local emission sources, green infrastructure coverage, and diurnal dynamics. For primary pollutants (PM_2.5, PM₁₀, NO₂, CO₂, VOC), the highest concentrations generally occurred during the morning (08:00) and evening (18:00) periods at the traffic-dominated sites A.N.R. and PL.K.S. These peaks align with rush-hour emission surges and are intensified at locations with limited dispersion capacity due to dense urban morphology and minimal vegetation.

In contrast, S.G.P., with the highest green cover, maintained consistently low concentrations for all primary pollutants throughout the day, while A.U.A. showed intermediate values, reflecting its mixed land-use context and moderate vegetation density. This pattern confirms that green infrastructure helps maintain cleaner air not only during peak traffic periods but also during off-peak hours. For ozone (O₃), the heatmaps display the expected inverse pattern relative to primary pollutants: midday (13:00) concentrations were highest at S.G.P. and A.U.A., reflecting favorable photochemical production under high sunlight and lower NO titration. Conversely, traffic-intensive sites (A.N.R. and PL.K.S.) showed suppressed O₃ during morning and evening peaks due to elevated NO emissions, which drive rapid O₃ depletion.

3.3.3. Weekday–Weekend Contrasts

The comparative analysis of weekday and weekend pollutant concentrations (Figure 13) shows a clear and consistent pattern. For primary pollutants (PM_2.5, NO₂, CO₂, VOCs), weekday averages tended to exceed weekend averages across sites, with the strongest relative differences for NO₂ in traffic-influenced areas—consistent with reduced commuting and freight emissions on weekends. PM_2.5 exhibited noticeable weekday–weekend contrasts (typically ~10–20%), indicating a substantial road-traffic contribution to fine particulate levels in Athens. By contrast, PM₁₀ differences were smaller and not statistically significant overall, reflecting a larger influence of resuspension and coarse sources less tightly coupled to weekday traffic. CO₂ was modestly higher on weekdays (combustion-related emissions), whereas VOCs also tended to be higher at traffic-heavy locations; neither CO₂ nor VOCs showed overall significance.

In contrast, ozone (O₃) was higher on weekends across sites (the “weekend ozone effect”), consistent with NO_x-rich urban atmospheres in which reduced weekend NO emissions lessen O₃ titration. Weekend enhancements in Athens were typically on the order of ~10–15% at greener sites (e.g., S.G.P., A.U.A.), reflecting a shift in photochemical balance under lower traffic.

Tukey HSD post hoc testing (Table 2) indicated that weekday > weekend differences were significant for NO₂ and PM_2.5, while O₃ showed a significant weekend increase. Differences for PM₁₀, CO₂, and VOCs were not significant. These contrasts underscore that short-term emission reductions primarily decrease primary pollutants, while secondary O₃ can increase, highlighting the need for balanced controls on NO_x (and, where relevant, VOCs).

4. Discussion

4.1. Overview of Key Findings

The results of this study provide evidence that vegetation cover is systematically associated with variation in both absolute pollutant concentrations and diurnal exposure dynamics across the examined urban microenvironments in Athens. After normalizing for meteorological state using an ANCOVA framework, consistent negative associations were observed between green cover and primary pollutants (PM_2.5, PM₁₀, NO₂, CO₂, VOCs), while ozone (O₃) exhibited a positive association with green cover, consistent with reduced NO titration under lower-emission, more vegetated conditions. Across the range of 7.5% to 95% vegetation cover, estimated concentration changes per +10% increase in green cover were PM_2.5 −0.450 µg·m⁻³ (R² = 0.508), PM₁₀ −0.840 µg·m⁻³ (R² = 0.584), NO₂ −0.002 ppm (R² = 0.579), CO₂ −4.930 ppm (R² = 0.688), VOCs −0.016 ppm (R² = 0.575), and O₃ +0.001 ppm (R² = 0.845). These regression outputs were derived from the full pooled observation dataset (n = 288), rather than from site-averaged values, ensuring that the reported relationships reflect observation-level variability rather than four-point regression fits.

Diurnal structure further reinforced these relationships. Primary pollutant peaks at 08:00 and 18:00, attributed to commuter emissions and boundary-layer stability, were consistently dampened at sites with greater vegetation cover, while midday mixing (13:00) produced the lowest primary-pollutant levels, most notably at the high-cover site (S.G.P.). O₃ displayed the inverse pattern, with pronounced midday maxima at more vegetated sites under strong photochemical activity.

It is important to note that the ANCOVA normalization applied here was designed to control for meteorological variability (temperature and relative humidity), which strongly governs atmospheric mixing and photochemical reactivity. This procedure enables controlled comparison of pollutant gradients under comparable atmospheric conditions; however, it does not eliminate variation arising from structural urban characteristics such as street geometry, emission density, or building morphology. Accordingly, the associations reported in this section should be interpreted as meteorology-normalized empirical relationships observed within the studied settings, rather than fully isolated causal effects of vegetation alone.

4.2. Mechanistic Interpretation

4.2.1. Physical and Chemical Pathways

The observed relationships are consistent with multiple GI-mediated processes acting in parallel to reduce primary pollutants—most notably particulate matter—and to reshape secondary chemistry:

• Dry deposition and surface capture: Foliage intercepts particles (impaction, interception, sedimentation) and absorbs gases on leaf surfaces. Deposition efficiency increases with leaf area index, surface roughness, and microstructure; species with hairy or waxy leaves—common in Mediterranean flora—exhibit enhanced particulate capture [39,40]. The negative PM_2.5/PM₁₀ slopes with green cover and the damped diurnal amplitudes at greener sites are consistent with strengthened deposition and surface trapping.
• Stomatal uptake (gas exchange): Gaseous pollutants (NO₂, O₃, CO₂) enter through stomata during photosynthetic gas exchange. Uptake rates vary with physiology and environment; high summer photosynthetic activity maximizes CO₂ assimilation and contributes to NO₂ and O₃ removal [41,42]. Midday reductions in NO₂ at greener sites align with periods of higher stomatal conductance.
• Dispersion and microclimate: Vegetation alters near-surface flow by increasing aerodynamic roughness and generating canopy-scale turbulence, which enhances dilution and vertical mixing in open parks and setbacks; concurrently, shading and evapotranspiration lower air and surface temperatures and can increase relative humidity, modifying boundary-layer stability and, indirectly, reaction rates [43]. The consistently lower primary-pollutant means at high-cover sites and the weaker morning/evening peaks indicate more effective dispersion in vegetated environments, and the combination of cooler, more humid microclimates with enhanced daytime mixing is consistent with the observed midday minima in primary pollutants at greener locations.
• Dust resuspension and source buffering: Vegetated ground cover and porous surfaces suppress mechanical resuspension of road dust and provide physical separation from traffic lanes, decreasing local PM₁₀ and PM_2.5 exposure. The stronger PM₁₀ response along the green-cover gradient is compatible with reduced coarse-particle resuspension near vegetated buffers.
• NO titration and the ozone trade-off: Higher O₃ at greener, lower-NO sites reflects reduced NO titration rather than increased precursor supply. In traffic-dense areas, fresh NO rapidly depletes O₃; where NO is lower, midday photochemistry maintains higher background O₃. The weekend O₃ enhancement and the positive O₃–green-cover slope are both consistent with this NO_x-rich regime. These dynamics underline the need to pair GI expansion with NO_x management, so that PM and NO₂ benefits are realized without unintended O₃ increases.

The combined action of deposition/uptake, enhanced dispersion, microclimate cooling, and reduced resuspension provides a coherent mechanistic basis for the measured declines in PM_2.5, PM₁₀, NO₂, CO₂, and VOCs with increasing vegetation cover, while the NO-titration mechanism explains the inverse O₃ response.

4.2.2. Pollutant-Specific Dynamics

• Particulate Matter (PM_2.5, PM₁₀): The reductions observed with increasing vegetation are consistent with leaf-surface capture mechanisms, particularly effective for coarse and fine particles. The greater slope for PM₁₀ than PM_2.5 suggests that coarse particles, with higher deposition velocities, benefit more from interception, whereas PM_2.5 removal also depends on longer-term diffusion processes.
• Nitrogen Dioxide (NO₂): The strongest negative slope across pollutants indicates that NO₂ removal benefits greatly from both stomatal uptake and dilution in vegetated areas. NO₂’s short lifetime makes it particularly sensitive to local emission-removal balances; reduced concentrations in green areas imply efficient mitigation of nearby traffic emissions.
• Volatile Organic Compounds (VOCs): While vegetation can emit biogenic VOCs (BVOCs) [44], the observed negative relationship suggests that, in this context, traffic-related VOCs dominate the ambient VOC budget. In this study, VOCs were measured as total VOCs (TVOC; ppm) using a non-speciated sensor, i.e., the instrument reports aggregate VOC burden rather than individual compounds. Accordingly, the decline in TVOC with increasing green cover indicates that vegetation acts more as a sink than a source under these urban conditions. It is also plausible that prevalent plant species emit comparatively low levels of highly reactive BVOCs, mitigating any enhancement of O₃ formation potential.
• Carbon Dioxide (CO₂): The reductions in CO₂ with green cover reflect both photosynthetic assimilation and improved mixing, with the latter likely dominating during nighttime hours when photosynthesis ceases but dispersion remains more efficient in open, green areas.
• Ozone (O₃): The positive slope reflects a classic NO titration effect, in traffic-dense, low-vegetation areas, NO from combustion scavenges O₃, reducing its concentration. In greener sites, lower NO emissions allow O₃ formed through VOC–NO photochemistry to accumulate. This is reinforced by midday peaks at vegetated sites, coinciding with high temperatures (>30 °C) and maximum photolysis rates.

4.3. Integrating Findings with Existing Literature

The results are consistent with reports from Mediterranean cities such as Barcelona and Rome, where 10–40% reductions in PM_2.5 and NO₂ have been observed in proximity to green areas [45,46]. The higher O₃ at greener sites is likewise in line with findings from Northern Italy, where reduced NO titration outweighed ozone deposition [47]. Taken together, the present outcomes fit the broader Mediterranean evidence that vegetation was associated with lower concentrations of primary pollutants while potentially elevating daytime O₃ under NO_x-rich regimes.

This contribution extends prior work on the subject in three key ways:

• Meteorological control: By applying ANCOVA-based normalization, short-term weather influences were statistically removed, enabling direct estimation of the vegetation (green-cover) effect independent of temperature and humidity fluctuations.
• Urban gradient coverage: Site selection spanned a near-complete vegetation gradient (≈5–100% cover) and diverse urban morphologies, revealing a consistent scaling relationship rather than a simple binary contrast between “green” and “non-green” locations.
• Temporal resolution: Hourly sampling at fixed times captured diurnal structure and identified critical exposure windows (commuting peaks) while quantifying the extent to which vegetation dampens peak amplitudes, a feature often missed in studies with coarser temporal resolution.

These advances position the study within, yet incrementally beyond, existing Mediterranean literature by quantifying a meteorology-controlled, gradient-based GI effect on PM and co-pollutants, and by clarifying the O₃ trade-off that should be addressed through complementary NO_x management.

4.4. Public Health and Urban Planning Implications

Health relevance is clear. Primary pollutant peaks at 08:00 and 18:00 coincide with periods when large numbers of people, children on school routes and workers commuting are outdoors. By lowering peak amplitudes, green infrastructure can disproportionately reduce short-term exposure spikes associated with cardiovascular events and asthma exacerbations.

It is important to note that the observed reductions in CO₂ concentrations at higher-cover sites reflect localized microenvironmental modulation and short-term plant physiological uptake rather than net changes in emission fluxes or ecosystem-scale carbon sequestration. Accordingly, these CO₂ differences should not be interpreted as direct evidence of greenhouse gas mitigation.

Planning implications include:

• Incremental greening is effective. Based on the Table 3 slopes, a +10% increase in green cover is modeled to change concentrations by approximately PM_2.5 −0.00045 mg·m⁻³ (≈−0.450 µg·m⁻³), PM₁₀ −0.00084 mg·m⁻³ (≈−0.840 µg·m⁻³), NO₂ −0.00155 ppm (≈−2.92 µg·m⁻³ at 25 °C, 1 atm), VOCs −0.0155 ppm, and CO₂ −4.93 ppm, with O₃ +0.00144 ppm (≈+2.83 µg·m⁻³). For context, the WHO (2021) [26] annual guidelines are PM_2.5 = 5 µg·m⁻³, PM₁₀ = 15 µg·m⁻³, NO₂ = 10 µg·m⁻³, and O₃ (warm-season, 8-h) = 60 µg·m⁻³; the EU AAQD (2030 revisions) adopts PM_2.5 = 10 µg·m⁻³ and NO₂ = 20 µg·m⁻³ annual limit values. Thus, a +10% greening corresponds to roughly ~9% of the WHO PM_2.5 annual guideline, ~6% of the WHO PM₁₀ guideline, ~29% of the WHO NO₂ annual guideline (and ~15% of the EU NO₂ limit), and ~5% of the WHO O₃ warm-season guideline—indicating that even modest greening can yield population-relevant gains.
• Site-specific design is critical. In narrow street canyons, poorly configured vegetation may impede ventilation. Planting should favor porous, well-spaced canopies, setback vegetated buffers, and continuous ground cover to enhance dispersion while providing deposition surfaces. Preference for lower-BVOC-emitting species is advisable to avoid adding reactive precursors.
• O₃ requires co-controls. Because O₃ increased with green cover (consistent with reduced NO titration), GI expansion should be paired with NO_x mitigation (e.g., low-emission transport corridors, fleet electrification, flow-smoothing signal timing) so that PM/NO₂ gains are realized without unintended O₃ increases.
• Target exposure hotspots. Priority should be given to school routes, bus corridors, transit nodes, and pedestrian approaches to intersections, where diurnal peaks and footfall coincide, maximizing population-level exposure reduction.

The final results indicate that well-designed, incremental greening can lower PM exposure and dampen rush-hour peaks, aligning with the Special Issue focus on particulate-matter reduction, provided that NO_x co-controls accompany GI deployment to manage the ozone trade-off.

4.5. Limitations and Future Directions

The findings of this study should be interpreted in the context of its sampling design. The monitoring campaign spanned two weeks and included four fixed sampling windows per day across four sites, providing mid-duration, multi-site observational evidence suitable for controlled intra-city comparison under normalized meteorological conditions. However, this design does not, by itself, support generalized causal inference beyond the measured urban contexts. The relationships reported between vegetation cover and pollutant concentrations therefore represent context-specific empirical associations within the examined microenvironments, rather than universally transferable scaling laws.

Vegetation cover was used here as an integrated landscape indicator that co-varies with additional urban-form attributes, including street geometry, surface roughness, emission density, and access to background air masses. While the meteorological normalization procedure (ANCOVA) controlled for temperature- and humidity-driven variations in boundary-layer mixing and photochemical reactivity, it did not fully resolve heterogeneity associated with traffic intensity, building height, or street-canyon structure. Future analyses would benefit from incorporating explicit morphometric indices (e.g., sky-view factor, aspect ratio, aerodynamic roughness length) and emission inventories to further separate physical, chemical, and source-dependent contributions to pollutant variability.

The two-week summer period captures vegetation’s peak physiological activity and high photochemical intensity, and may therefore represent upper-bound mitigation effects. Extending measurements across multiple seasons would allow estimation of seasonal differences in deposition rates, stomatal uptake, BVOC emissions, and boundary-layer dynamics. Additionally, while portable low-cost sensors were factory-calibrated, cross-checked prior to deployment, and periodically validated using filtered-air reference checks, some degree of measurement uncertainty remains inherent to low-cost sensing platforms.

Future research would benefit from co-location of short-term campaign measurements with regulatory-grade reference stations (e.g., Greek Ministry of Environment and EEA monitoring networks) to extend seasonal representativeness and contextualize observed intra-urban gradients relative to longer-term climatological baselines. Integration with urban-scale dispersion modeling (e.g., ENVI-met, CFD) and micrometeorological or tracer-based approaches would further support mechanistic attribution and scenario-based planning.

5. Conclusions

A multi-site comparative assessment was undertaken to quantify how variation in vegetation cover shapes particulate matter (PM_2.5 and PM₁₀), co-occurring gaseous pollutants (NO₂, CO₂, VOCs), and ozone across representative urban microenvironments in Athens, Greece. The four monitoring sites were selected to span a continuous vegetation-cover gradient, enabling evaluation across contrasting morphological and emission contexts. Over a two-week warm-season campaign, 288 meteorology-paired observations were collected, and meteorology-normalized statistical models (including ANCOVA and regression) were applied to distinguish vegetation-associated effects from short-term atmospheric variability. The results provide empirical evidence that vegetation cover is associated with systematic modulation of pollutant concentration fields and diurnal exposure patterns in Mediterranean urban settings.

5.1. Key Quantitative Outcomes

Greener sites consistently exhibited lower levels of primary pollutants (PM_2.5, PM₁₀, NO₂, CO₂, VOCs), while O₃ tended to be higher, reflecting reduced NO titration where traffic-derived NO_x emissions were lower. This contrast was most pronounced between Syggrou Grove Park and Athens National Road, illustrating how vegetation interacts with local emission regimes. These findings indicate that increasing vegetation cover can contribute to reduced exposure to primary pollutants, provided that the well-known O₃ trade-off is managed through complementary emission controls.

5.2. Temporal and Spatial Dynamics

Rush-hour peaks were most pronounced at traffic-dominated corridors, while greener areas exhibited flatter diurnal profiles and lower variance, suggesting that vegetation influences both average concentrations and peak exposure intensity. Midday minima at highly vegetated sites align with enhanced mixing and deposition processes, whereas higher midday O₃ at greener locations reflects photochemical conditions with less NO scavenging. These dynamics highlight the importance of targeting greening interventions in high-exposure public-use corridors, in combination with NO_x management, to avoid unintended O₃ increases.

5.3. Implications for Urban Air Quality Management

The scaling observed across the vegetation-cover gradient suggests that measurable improvements in urban air quality are achievable even with incremental greening, particularly in traffic-intensive zones where diurnal peaks drive population exposure. However, the O₃ response emphasizes that green infrastructure should not be regarded as a stand-alone mitigation strategy. Rather, it is most effective when integrated with transport emission reductions, NO_x controls, and context-appropriate canopy design (e.g., avoiding stagnation in canyons, selecting lower-BVOC-emitting species).

5.4. Broader Significance

Evidence from this study shows that green infrastructure is associated with lower primary pollutant concentrations, damped diurnal peaks, and more stable day-to-day air quality cycles in a Mediterranean megacity context. Methodologically, the work contributes a meteorology-normalized, gradient-based, multi-pollutant dataset for Athens, advancing beyond fragmented single-site approaches.

In policy terms, the modeled per +10% increase in green cover represents a meaningful step toward health-based standards: PM_2.5 ≈ −0.450 µg·m⁻³ and PM₁₀ ≈ −0.840 µg·m⁻³ per +10% correspond to ~9% and ~6% of the WHO (2021) [26] annual guidelines (5 and 15 µg·m⁻³, respectively) and to roughly ~4–8% progress toward indicative EU AAQD (2030) limits (e.g., PM_2.5 = 10 µg·m⁻³). For NO₂, the study’s slope (≈−0.0016 ppm, ~−2.9 µg·m⁻³ at 25 °C, 1 atm) equates to ~15% of the EU 2030 annual limit (20 µg·m⁻³) and ~30% of the WHO guideline (10 µg·m⁻³), underscoring that incremental greening can yield population-relevant gains, especially when targeted to traffic-dense corridors. At the same time, the observed increase in O₃ with greening, consistent with reduced NO titration, highlights the need to pair GI deployment with NO_x co-controls so that PM/NO₂ benefits are realized without unintended ozone penalties.

Future Research Directions, extending monitoring across seasons, adding source apportionment to separate traffic and industrial contributions, and quantifying co-benefits (urban cooling, biodiversity, carbon sequestration) will sharpen design guidance and benefit–cost estimates. Collectively, the results position GI as a scalable, cost-effective complement to emissions abatement that can move cities measurably closer to WHO and EU targets while delivering broader ecosystem and climate-resilience co-benefits.