Assessing the Impact of Port Emissions on Urban PM_2.5 Levels at an Eastern Mediterranean Island (Chios, Greece)

Abstract

Air pollution from ship operations can pose a significant challenge for coastal cities, particularly where ports are closely integrated into the urban fabric. This study examines the influence of ship docking on PM_2.5 concentrations in Chios, Greece, a medium size island city where the port directly borders densely populated neighbourhoods. Calibrated PurpleAir sensors were installed at urban and suburban sites to measure PM_2.5, with data analysed alongside ship call records and meteorological observations. An event-based concentration enhancement metric (%ΔC) was estimated to compare PM_2.5 during docking with the preceding 3 h background for 170 ship arrivals in February and August 2022. The results showed that under prevailing northerly winds in August, PM_2.5 at the downwind urban site increased on average by 5.0 µg m⁻³ (48%), whereas winter increments were smaller (6.1%) due to higher background variability. When both seasons and all wind directions were pooled, the urban site exhibited a mean enhancement of 1.7 µg m⁻³ (19%), while impacts at the suburban site remained minor (3%). Median-based uncertainty analysis confirmed robust enhancements under northerly winds only. Wind direction and wind speed were the primary controls on %ΔC, whereas ship engine power and time at berth had limited influence. The results suggest that ship-related PM_2.5 impacts are detectable but remain spatially and temporally limited in coastal urban environments, including medium-sized islands characterised by relatively low shipping activity.

1. Introduction

Maritime transport is an essential component of global trade, yet it represents a significant and growing source of atmospheric pollution in coastal environments. It is estimated that about 50% of global trade by value is transported by sea, with low-income countries and small island nations depending on ports 1.5 to 2 times more than the global average [1]. This highlights the importance of ports as essential infrastructure that supports supply chains. For island economies, ports are even more critical due to their geographic isolation, limited local resources, and fewer available transport options compared with the mainland. In these contexts, island ports are of vital importance not only for handling cargo but also as entry points for people, medical supplies, tourists, and essential services [2].

Despite their socio-economic significance, ports generate significant externalities arising from their continued heavy reliance on fossil-fuel-based operations [3] that generate substantial emissions originating from ships, cargo-handling machinery, drayage, heavy-duty trucks, and terminal operations [4,5]. The impacts are exacerbated when ports are located within urban areas, where limited buffer zones increase the exposure risks for nearby populations. Substantial emission reductions can be achieved through the adoption of cleaner energy systems and intelligent energy-management strategies [3].

IMO’s MARPOL Annex VI aims to reduce air pollution from ships by regulating both nitrogen oxide (NO_x) emissions and the sulphur content of marine fuels, which also influences particulate matter formation. Tobarra et al. [6] in a 5-year study at the Port of Alicante, Spain, report the decrease in the shipping source of PM₁₀ that coincides with the IMO 2020 sulphur regulations. However, while switching to low-sulphur marine fuels reduces aerosol sulphate formation, substantial soot and organic aerosol emissions persist [7,8].

A substantial body of research has examined air pollution in port environments. It was found that pollution comes not only from marine fuel combustion but also from passenger cars, diesel trucks, cranes, and ancillary port equipment. Ship emissions, from auxiliary engines operating at berth, and the main engines during manoeuvring, release substantial amounts of fine particulate matter (PM_2.5), nitrogen oxides (NO_x), sulphur oxides (SO_x), and volatile organic compounds (VOCs), which can adversely affect urban air quality and human health [9,10,11,12]. Port activities contribute disproportionately to coastal pollution episodes due to their proximity to densely populated zones and the continuous operation of ship engines during loading, unloading, and hoteling periods [13].

More specifically, existing studies have primarily focused on long-term observational analyses [5,6], top-down statistical assessments and pollution datasets compiled from multiple sources [9,14], photochemical, dispersion, and receptor modelling approaches [11,15,16,17,18], as well as detailed emission inventories [12]. Measurement campaigns have been conducted either within the port area [10,19,20,21] or across both port facilities and adjacent urban surroundings [5,15,16,17]. Notably, most of this work has focused on large urban ports [5,15,17,18]. Several studies across the Mediterranean [10,18,19,20] have shown that port traffic is associated with increases in urban PM_2.5 and that the magnitude and spatial footprint of these impacts are strongly modulated by local meteorology, particularly wind direction and speed.

However, quantifying these effects remains challenging due to complex coastal atmospheric dynamics, topographical variability, and the limited availability of high-resolution emission and meteorological data. Specifically, there is comparatively little evidence on the short-term, event-scale behaviour of ship plumes, the magnitude of PM_2.5 enhancements during individual docking events, and the extent to which meteorological factors—especially wind direction and seasonal regimes—govern these enhancements.

Chios Island, located in the eastern Aegean Sea, Greece, represents a characteristic example of a port city where substantial ship traffic occurs near residential areas. The port of Chios plays a pivotal role in sustaining the island’s economy, tourism and social connectivity. However, its proximity to densely populated neighbourhoods raises concerns about potential air-quality degradation.

Considering the above knowledge and framework, the main objective of the study is to quantify short-term PM_2.5 concentration changes associated with individual ship docking events at the port of Chios and to determine how these changes vary across seasons, local meteorology and ship’s characteristics. To address this objective, the study pursues the following specific aims:

• Quantify ship-related PM_2.5 increments during docking events.
• Characterize seasonal and spatial variations between urban and suburban monitoring sites.
• Determine the key meteorological and ship related factors that influence plume transport.
• Build a data-driven, scientific understanding of ship plume impacts in insular environments to guide effective mitigation.

The novelty of this study lies in introducing a framework that effectively detects short-term port-related PM_2.5 concentration enhancements in port cities. The framework applies an event-based analysis of urban PM_2.5 changes in conjunction with detailed meteorological information and vessel’s operational and technological characteristics.

The remainder of the manuscript is organised as follows: Section 2 describes the study area, the PM_2.5 monitoring setup, the ship traffic and meteorological datasets, and the event-based analytical framework applied to quantify ship-related concentration enhancements. Section 3 presents the results, including the spatial and seasonal patterns of PM_2.5 changes, uncertainty analysis by wind sector, and correlations with meteorological and ship-related variables. Section 4 discusses the findings in the context of existing literature and underlying physical mechanisms. Finally, Section 5 summarizes the main conclusions of the study and discusses its limitations and future research perspectives.

2. Materials and Methods

2.1. Study Area

Chios is the fifth largest island of Greece, located in the northeastern Aegean Sea (Figure 1). The island’s capital, Chios City, is located by the eastern coast, near the centre of the island. According to the 2021 national census, the population of the island exceeds 50,000, with almost 27,000 residents in Chios City.

Chios has a Mediterranean climate, characterized by hot, dry summers and mild, wet winters. Persistent northerly winds that prevail over the eastern Mediterranean (the Etesians) flow during the summer [22]. They originate as northeasterlies over the northern Aegean and become northerly over the central and southern Aegean [23].

The island has a strong and long-standing maritime tradition and shipping-related employment remains significant, particularly in Chios City, Vrontados, Kardamyla and the nearby island of Oinousses [24]. Agriculture is also important, especially with the cultivation of mastic trees in the southern region supporting 5000 producers [25]. Tourism, centred on cultural heritage, agrotourism and marine activities, also contributes substantially to the local economy.

The Port of Chios (38.3724° N, 26.1382° E) (Figure 1) serves as the island’s main maritime gateway. Like many island ports, it is located adjacent to the urban centre, which results in traffic congestion and increased pollution levels, especially during the scheduled arrival and departure of large Ro-Pax vessels connecting Chios with the major Mediterranean port of Piraeus. These operations typically occur twice daily, between 04:00 and 06:00 (arrival from Piraeus and departure to Mytilene) and 21:00 and 23:00 (arrival from Mytilene and departure to Piraeus), with vessels usually staying at dock for about 45 min.

The port can simultaneously accommodate two 20,000-ton passenger/cargo vessels and one 700-ton passenger ship. Its southern sector includes a marina for fishing boats, sailboats and private yachts. Chios is connected by ferry to mainland ports (Piraeus, Thessaloniki and Kavala) and to several Aegean islands (Lesvos, Ikaria, Samos, Limnos, Mykonos, Patmos and Syros). Smaller vessels provide daily service to the nearby islands of Oinousses and Psara, as well as to Çeşme on the Turkish coast—an important tourist destination. The port handles approximately 3830 vessel calls, 502,500 passengers and 104,500 tonnes of cargo annually [26], underlying its importance for mobility, trade, and tourism given the island’s limited alternative transport options.

Table 1 summarizes the main characteristics of the vessels. The vessels differ substantially in size, engine power, age and time spent at port, which together determine their PM_2.5 emission capacity. All vessels are of Ro-Pax type, except for Ship F, which operates as a Ro-Ro vessel. In terms of age, vessels D and E are 18 years old, vessels A, B, and C range between 35–38 years, and vessel F is 47 years old.

Ro-Pax ships are generally among the older ship categories in the European flagged fleet with an average of 27–28 years [27]. Newer vessels typically achieve higher energy efficiency due to more advanced engine designs and improved combustion technologies [27]. However, age alone is not a definitive indicator of performance or emissions, as a well-structured and properly implemented maintenance programme can ensure compliance with safety and pollution-prevention standards as required by the ISM Code [28].

Table 1. Specifications and operational parameters of the ships examined in this study.

Characteristic	Ship A	Ship B	Ship C	Ship D	Ship E	Ship F
Gross tonnage (GT)	30,694	29,422	15,362	8126	8126	5466
Deadweight (DWT) (tons)	7622	6148	3348	1960	1960	8661
Length overall (m)	192.91	192.50	141.54	141	141	154
Beam (m)	29.40	27.28	23.00	21	21	21.67
Year built	1988	1987	1990	2007	2007	1978
Passengers	2213	2210	1462	1715	1715	12
Cars	~750	~748	274	418	418	-
Lane meters (m)	2400	2105	~900	-	-	1650
Total engine power (MW)	17.5	~17.2	~9.8	31.7	31.7	11.6
Max speed (kt)	~20.5	~22	~21	-	-	-
Number of calls at port (August and February 2022)	57	44	30	12	14	13
Average time at port per call (min)	47	46	37	20	21	60

Sources: [29,30].

Table 1. Specifications and operational parameters of the ships examined in this study.

Characteristic	Ship A	Ship B	Ship C	Ship D	Ship E	Ship F
Gross tonnage (GT)	30,694	29,422	15,362	8126	8126	5466
Deadweight (DWT) (tons)	7622	6148	3348	1960	1960	8661
Length overall (m)	192.91	192.50	141.54	141	141	154
Beam (m)	29.40	27.28	23.00	21	21	21.67
Year built	1988	1987	1990	2007	2007	1978
Passengers	2213	2210	1462	1715	1715	12
Cars	~750	~748	274	418	418	-
Lane meters (m)	2400	2105	~900	-	-	1650
Total engine power (MW)	17.5	~17.2	~9.8	31.7	31.7	11.6
Max speed (kt)	~20.5	~22	~21	-	-	-
Number of calls at port (August and February 2022)	57	44	30	12	14	13
Average time at port per call (min)	47	46	37	20	21	60

Sources: [29,30].

2.2. Materials

For the monitoring of PM_2.5 concentrations, PurpleAir Classic Plus monitors equipped with dual PMS-5003 laser particle counters were used (PurpleAir, Draper, UT, USA). Data were transmitted to the cloud via Wi-Fi. A reference-equivalent analyser (Thermo Fisher Scientific Model 5014i, Thermo Fisher Scientific, Waltham, MA, USA) was used to evaluate and correct sensors’ measurement response [31].

2.3. Data Collection and Analysis

Initially, a set of eight PA-II devices were co-located at the suburban site in Chios, between 27 October and 12 November 2021, to assess their repeatability (precision). The inter-comparison of paired-sensor measurements demonstrated excellent repeatability, with coefficients of determination (R²) and slopes close to unity and intercepts near zero in the linear regression analyses.

Subsequently, to evaluate the accuracy of the monitors, one of these devices was co-located with a reference-equivalent analyser (Thermo Fisher Scientific Model 5014i), at the National Observatory of Athens Thissio supersite monitoring station in Athens from 23 December 2021 to 28 February 2022. The intercomparison produced the following calibration equation for PM_2.5, developed with the inclusion of relative humidity (RH) as an explanatory variable (Equation (1)).

y_{cor_PM2.5} = 11.051 + 0.423 × X_{sens_PM2.5} − 0.117 × RH, R² = 0.810

(1)

Accordingly, the readings from the urban and suburban PurpleAir sensors used in this study were corrected using this calibration relationship.

Two PM_2.5 monitors were deployed to continuously monitor hourly PM_2.5 concentrations [32] (Figure 1). An urban site (38.3657° N, 26.1367° E) was located approximately 0.8 km south of the port within the Chios residential area. A suburban site (38.3895° N, 26.1297° E) was located approximately 2.2 km NNW of the port in a less densely built environment.

Meteorological data (barometric pressure, temperature, dew point, wind speed and wind direction) were obtained from the Hellenic National Meteorological Service (HNMS) station at Omiros Chios Airport (2.7 km straight distance South of the port).

Ship arrival and departure data for vessels with a gross tonnage (GT) of at least 5000 was obtained from the Chios Port Authority for February and August 2022.

For each ship docking event, the mean PM_2.5 concentration during the docking period (C₁) was compared with the average concentration during the preceding 3 h (C₂) in order to determine the short-term impact of port activity (Equations (2) and (3)).

$$∆\mathrm{C}={\mathrm{C}}_1-{\mathrm{C}}_2$$

(2)

$$\%∆\mathrm{C}=\frac{({\mathrm{C}}_1-{\mathrm{C}}_2)}{{\mathrm{C}}_2}\times 100$$

(3)

These indicators represent the short-term absolute and relative increases in PM_2.5 concentrations during the ship’s manoeuvring and docking at the port. They are driven by emissions from main and auxiliary ship engines for onboard electric power generation, and from the associated road traffic within the port area (trucks and cars waiting to load or unload to the ship). As it is not possible to separate the ship and truck components from ambient measurements, ΔC represents the combined port-related enhancement at each receptor site.

2.3.1. Wind Direction

Wind direction values (θ in degrees, converted to radians) were transformed into sine and cosine components. This approach transforms directional information into continuous orthogonal variables that represent the east–west and north–south flow components, avoiding discontinuities at 0°/360° [33] and enabling correlation with PM_2.5 changes (Equations (4) and (5)):

$$\mathrm{wd\_cos}=\cos (\mathrm{\theta }\times \frac{\mathrm{\pi }}{180})$$

(4)

$$\mathrm{wd\_sin}=\sin (\mathrm{\theta }\times \frac{\mathrm{\pi }}{180})$$

(5)

where wd_cos corresponds to the north–south flow component and wid_sin to the east–west component.

2.3.2. Statistical Analysis

Before applying statistical analyses, the normality of the dependent variables was evaluated using both the Kolmogorov–Smirnov (K–S) and Shapiro–Wilk tests. Both tests indicated significant departures from normality (p < 0.001) for the absolute and relative PM_2.5 concentration differences (ΔC and %ΔC) at the urban site.

Given the presence of right-skewed distributions, with outliers typical of episodic emission events and meteorological variability, the study proceeded with resistant non-parametric analyses. Specifically, to explore the relationships between ship-related and meteorological parameters and the observed short-term PM_2.5 variations at the two monitoring stations, the non-parametric Spearman’s rank correlation coefficient (ρ) was used.

Spearman correlations were calculated between the percentage change in average PM_2.5 during ship docking (%ΔC) and a set of potential explanatory variables. Ship-related factors included main engine power (kW) and time at port (minutes). Meteorological variables comprised wind speed (kt), the sine and cosine components of wind direction (wd_sin and wd_cos), air temperature (°C), dew point (°C), and atmospheric pressure (hPa).

Because %ΔC was non-normally distributed, uncertainty was quantified using the median, interquartile range (IQR), and bias-corrected accelerated bootstrapped 95% confidence intervals across northerly, southerly, and other wind sectors.

All the analyses were carried out with IBM SPSS vers. 28.0.1.0 (142).

3. Results

3.1. Assessment of the Influence of Ship Docking Events on PM_2.5 Levels in Chios City

This study aims to assess the extent to which ship dockings and associated vehicle traffic at the Chios port contribute to the city’s fine aerosol pollution of PM_2.5. The study period (February and August 2022) was selected to represent contrasting meteorological conditions, port traffic and emission activity.

February in Chios is dominated by domestic heating emissions [32] and a more dispersed distribution of wind directions compared to August (Figure 2). August is characterized by persistent northerly winds and minimal precipitation, which limits wet removal processes and significantly promotes elevated PM_2.5 levels [34]. Enhanced photochemical production rate of fine particle aerosols such as sulphate and nitrate further contributes to fine particle formation [35]. In August, ship calls nearly doubled, and vehicle flows considerably increased compared to February, reflecting the summer tourism peak.

The study focuses on the impacts of PM_2.5 levels associated with the operation of six large (≥5000 GT) Ro-Pax and Ro-Ro vessels calling at the port of Chios (Table 1). Emissions associated with these large vessels are expected to be detectable above typical emissions from other local sources. A total of 170 ship docking events were documented over the study period: 60 in February 2022 and 110 in August 2022.

In February, ferry activity typically consists of two itineraries per day. The first departs from Mytilene (Lesvos) and arrives in Chios at around 21:00, before departing for Piraeus approximately 45 min later. The second, covering the reverse route, departs from Piraeus and arrives in Chios between 05:00 and 7:00 before continuing to Mytilene after a similar stop. In August, increased tourist flows from the Greek mainland to the islands result in continuous circular itineraries, nearly doubling the number of daily dockings and sometimes altering the usual schedule. During the summer period, ship occupancy for passengers, cars and trucks approaches 100% due to both tourism and the higher quantities of food, fuel, and manufactured goods transported to meet seasonal demand. In contrast, occupancy in February is much lower, particularly for passengers and passenger vehicles.

Ships A and B, as the largest (≈30,000 GT, 17.2–17.5 MW) and the most frequent visitors (59% of the calls, 66% of the total time at port), are expected to be major contributors to port-related emissions. Ships D and E, although newer (built 2007) and probably equipped with more efficient engines, have a high engine power (≈31.7 MW), and they are still expected to produce considerable emissions.

The fuel type is determined by the regulations of the International Maritime Organization and the European Union, which Greece fully implements. In 2022 the ferries in the Greek seas should comply both with the EU Sulphur Directive 2016/802 and the IMO’s MARPOL Annex VI (2020 sulphur limit). This means that they used Marine Gas Oil (MGO, containing ≤ 0.1% sulphur) while docked and Very Low Sulphur Fuel Oil (VLSFO, containing ≤ 0.5% sulphur) while at sea.

The wind roses for February and August reveal distinct seasonal differences in both wind direction and speed over Chios Island (Figure 2). Winds in February exhibit a more dispersed directional pattern, with significant contributions from N, NNE, E, as well as S–SSW sectors. Although northerly winds remain dominant, southerly flows appear frequently, indicating different synoptic scale weather patterns typical of the winter in Mediterranean [36]. Wind speeds span a broader range and include a relatively large proportion of fresh to strong breezes (8–15 kt) and occasional northerly winds of ≥15 kt. In August, N and NNE winds dominate, characteristic of the summer Etesian winds in the North Aegean. Wind speeds are also persistently high in this period, with most winds falling within the 10–15 kt range and frequent occurrences above 12 kt.

Regarding the measured PM_2.5 concentrations, in February 2022 the urban site averaged 16 μg/m³, exceeding the suburban mean of 13 μg/m³; however, a Mann–Whitney U test showed no statistically significant seasonal difference in overall PM_2.5 levels between February and August, as the rank distributions were nearly identical (p = 0.865). In August 2022, means declined to 14 μg/m³ (urban) and remained at 13 μg/m³ (suburban). The boxplots in Figure 3 illustrate hourly PM_2.5 concentrations measured at urban and suburban sites during February and August. A seasonal pattern is visible, with February showing slightly higher medians and greater variability than August at both stations.

The potential influence of ship dockings on the urban and suburban pollution profiles was initially examined by plotting PM_2.5 concentration time series and comparing them with the temporal distribution of ship calls. A recurring pattern was evident in the urban station time series in August: sharp, short-lived increases in PM_2.5 coinciding with the presence of a ship at the port under northerly winds, followed by a rapid return to background levels once the ship departed (Figure 4). Given this recurring behaviour, subsequent quantitative analyses used the percentage change in PM_2.5 during docking (%ΔC), relative to the three-hour pre-arrival average, as the dependent variable (Equations (2) and (3)). This metric helps subtract the background pollution load and provides an estimate of the ship-related increments.

Figure 4 displays a representative PM_2.5 time series from 3–4 August covering a period with three ship docking events: one on the morning (07:10–08:10) (Ship B) and one in the evening of 3 August (21:40–22:35) (Ship B), and a third the following morning (06:05–07:10 of 4 August) (Ship A). The corresponding ΔC values were +8.2, +6.4, and +2.1 μg/m³, reflecting %ΔC increases of 127%, 59%, and 19%, respectively. During this period, the wind direction ranged from 360° to 20°, with speeds decreasing from 16 to 11 kt. As expected, the suburban site, located upwind of the port, did not register any impact. This pattern shown in Figure 4 is a recurrent feature in August, not an isolated event. Of the 110 dockings that month, 58 occurred under northerly winds, when the urban station was downwind. In 56 of these 58 cases (97%), %ΔC was positive, confirming repeated PM_2.5 increases during the ship’s stay. This behaviour was observed across all six vessels included in the study, indicating that the pattern is not ship-specific.

Quantitative analysis (

Table 2. Average change (ΔC) and per cent change (%ΔC) in PM_2.5 concentrations during ship docking events at the urban and suburban stations of Chios for different wind sectors and months (February and August 2022).

Month	Site	ΔC (μg/m3)	% ΔC	Wind Direction
February 2022	Urban	−0.3	−1.7	All directions (60 obs.)
	Suburban	−0.2	−1.7
August 2022	Urban	2.8	28	All directions (110 obs.)
	Suburban	0.3	3.4
February 2022	Urban	0.8	6.1	Northern (23 obs.)
	Suburban	0.3	2.3
August 2022	Urban	5.0	48	Northern (58 obs.)
	Suburban	0.4	4.5
February 2022	Urban	0.7	4.7	Southern (10 obs.)
	Suburban	2.4	27
August 2022	Urban	−0.7	−5.1	Southern (1 obs.)
	Suburban	0.0	−0.3
February & August 2022 combined	Urban	1.7	19	All directions (170 obs.)
	Suburban	0.1	3.0

) confirmed the seasonal differences. When averaged over all wind conditions, port-related PM_2.5 increments (ΔC) in February were negligible at both monitoring sites (urban: −0.3 μg/m³, −1.7%; suburban: −0.2 μg/m³, −1.7%), as winter background variability due to domestic heating and synoptic-scale circulation masked ship signals. By contrast, in August, distinct urban enhancements were observed (mean ΔC = +2.8 μg/m³, +28%), while the suburban site recorded minor changes (+0.3 μg/m³, +3.4%).

When the data are stratified by wind direction, a directional dependence becomes apparent. Under northerly winds, which are aligned with the port-to-urban station axis, the urban station consistently records the highest increases (February: +0.8 μg/m³, +6.1%; August: +5.0 μg/m³, +48%), whereas the suburban site, typically upwind of ship emissions, shows much smaller responses (in February, +0.3 μg/m³, +2.3%; in August, +0.4 μg/m³, +4.5%). Under southerly winds, however, the pattern reversed; in February, the suburban site showed a higher mean increase (+2.4 μg/m³, +27%) compared to the urban station site (+0.7 μg/m³, +4.7%), though this result was primarily driven by one high-pollution episode and a small number of southerly cases (N = 10). In August, only a single southerly event was recorded (urban −0.7 μg/m³, −5.1%), providing too little information for a reliable interpretation.

Finally, when the data from both months and all wind directions are combined, a consistent spatial pattern emerges. The urban site records an average ΔC of +1.7 μg/m³ (+19%), while the suburban area exhibits a negligible mean increase of +0.1 μg/m³ (+3%).

3.2. Uncertainty Analysis of PM_2.5 Enhancements by Wind Sector

The uncertainty analysis confirmed a strong dependence of short-term PM_2.5 enhancements on wind direction at the urban station (

Table 3. Median, IQR, and bootstrapped confidence intervals for %ΔC at the urban site by wind sector.

Wind Sector	N	Median %ΔC	IQR %ΔC	95% CI %ΔC (Bootstrap)
Northerly	81	25.5	35.6	[18.4, 31.5]
Southerly	10	−0.8	23.2	[−7.0, 19.4]
Other	79	0.3	19.1	[−2.5, 4.8]
All sectors	170	9.7	34.0	[5.2, 14.3]
All sectors, excluding southerly	160	10.7	34.0	[5.8, 14.9]

). Under northerly winds, the median %ΔC reached 25.5%, with substantial event-to-event variability (IQR = 35.6), but a relatively narrow bootstrapped 95% confidence interval (18.4–31.5%), reflecting the large number of observations (n = 81) and a well-constrained estimate of the central tendency. In contrast, southerly winds showed a median %ΔC close to zero (−0.8%) with moderate dispersion (IQR = 23.2), but a much wider confidence interval (−7.0 to 19.4%), indicating high uncertainty and limited statistical power due to the small sample size (n = 10). Other wind sectors also exhibited median values near zero with confidence intervals spanning zero. When all wind directions were combined, the median %ΔC was 9.7%, while exclusion of the low-frequency southerly sector yielded a similar median (10.7%) with overlapping confidence intervals, confirming the robustness of the directional conclusions.

3.3. Correlation of Concentration Change with Ship’s Particulars and Meteorology

The next step was to quantify the relationship between %ΔC in the urban and suburban stations and a set of predictors reflecting ship characteristics (main engine power), operational activity (time at port) and meteorological conditions (wind speed and direction, temperature, dew point and relative humidity) (

Table 4. Spearman’s rho correlation coefficient (ρ) (and its significance level) of the urban and suburban PM_2.5 concentrations with various predictors.

	Urban Station		Suburban Station
	Correlation Coef.	Significance	Correlation Coef.	Significance
Power (kW)	0.02	0.76	−0.02	0.84
Time at port (min)	0.02	0.77	−0.10	0.22
Weed speed (kt)	0.33 **	<0.001	0.03	0.74
wd_sin	0.05	0.54	0.03	0.68
wd_cos	0.34 **	<0.001	0.11	0.14
Temperature (°C)	0.34 **	<0.001	0.17 *	0.03
Dew Point (°C)	0.14	0.07	0.17 *	0.03
Pressure (hPa)	−0.16 *	0.04	0.00	1.00

** Significant correlation 0.01 level (2 tailed). * Significant correlation 0.05 level (2 tailed).

).

Spearman’s rank correlation analysis revealed several statistically significant, moderate associations between %ΔC at the urban site and key meteorological parameters. The correlations point to a transport dominated process in the near field of the port. The north–south wind flow component (wd_cos) exhibited a moderate, positive correlation with %ΔC (ρ = 0.34). The east–west (wd_sin) component is uninformative because changes along this axis do not determine plume–receptor alignment. These results align with the findings of Section 3.1. Wind speed also correlates positively with %ΔC (ρ = 0.33) at the urban site which is 0.8 km downwind. Temperature correlates positively (ρ = 0.34, p < 0.001), likely reflecting seasonal clustering, as warmer conditions coincide with August—when northerly winds and increased port activity amplify pollution episodes. Pressure shows a weak negative correlation (ρ = −0.16, p = 0.04), though this effect is minor and mainly reflects its covariance with temperature and seasonal conditions.

Main engine power and time at port, which are parameters that characterise ship technology and operations, showed negligible, non-significant correlations with %ΔC (ρ = 0.02).

At the suburban site, correlations are notably weaker. Only temperature (ρ = 0.17, p = 0.03) and dew point (ρ = 0.17, p = 0.03) show small but significant associations, while all wind-related variables are non-significant. This reflects the greater distance (~2 km) and less favourable alignment between the port and the suburban site: the plume reaches this receptor far less consistently, and only under specific (and relatively uncommon) southerly flow conditions. As a result, meteorology does not exert the same directional control, and the PM_2.5 increments are smaller, more diffuse, and more influenced by background environmental conditions rather than ship activity.

4. Discussion

Regarding the levels of PM_2.5 pollution measured in this study, elevated winter values likely result from increased emissions due to residential heating, reduced atmospheric mixing, and inversion conditions [17,32,37]. In contrast, the lower and more stable concentrations in August reflect enhanced dispersion and dilution under warmer, more turbulent conditions. Urban-site levels were consistently higher than at the suburban site attributable to higher traffic and population density. Numerous outliers, particularly in February, indicate episodic pollution events associated with unfavourable weather conditions or localized emission spikes typical in air quality datasets [32,38].

The results of PM_2.5 enhancements associated with ship manoeuvring and hoteling are detectable, spatially coherent, and strongly governed by wind direction (Table 2). The urban monitoring site located 0.8 km downwind of the port, under prevailing northerlies, which were the most favourable conditions, exhibited the clearest signal with average increases of 5 μg/m³ (48%) in August and 0.8 μg/m³ (6.1%) in February. When observations from both seasons and all wind directions were pooled, the mean enhancement reached 1.7 μg/m³ (19%). By contrast, the suburban site, typically upwind or crosswind relative to the port, showed negligible mean increases (average 0.1 μg/m³; 3.0%), except in a few cases of southerly winds.

To characterize uncertainty and event-scale variability median %ΔC, IQR, and bootstrapped confidence intervals were estimated (Table 3). Northerly winds show a median %ΔC of 25.5% with high variability but a well-constrained confidence interval due to the large sample size. Southerly and other wind sectors exhibit median values near zero with wide confidence intervals reflecting limited statistical power. Excluding southerly winds yields similar median %ΔC values, confirming the robustness of the directional conclusions.

The average 19% increase observed at the Chios urban station under all conditions (both months and all wind directions) falls well within the range reported internationally. Daily average PM_2.5 studies have consistently reported non-negligible port-related contributions, ranging from 5–10% (up to 17% including secondary aerosol formation) in the Bay of Algeciras [13], to 11–15% in Barcelona [15], and up to ~26% in harbour/industrial areas of Xiamen, with lower contributions at urban background sites [16].

To better resolve short-term ship-related PM_2.5 enhancements, several studies have adopted event-based approaches using high-temporal-resolution measurements, which are conceptually aligned with the event-based enhancement method applied in the present study. Contini et al. [19] used 1 min-resolution PM_2.5 data at three sites in the Port of Venice, comparing downwind concentrations during ship passages with downwind concentrations in the absence of ship passages, normalised by the overall mean PM_2.5 concentration, and estimated PM_2.5 enhancements of approximately 1–8% attributable to ship traffic. Applying the same methodology, Donateo et al. [20] analysed 30 min-resolution PM_2.5 measurements at a site located about 50 m from a ferry dock in the Port of Brindisi, estimating an average contribution of 9.3% from ship emissions and associated port traffic activities. Using the same approach at the Port of Shanghai, Wang et al. [21] reported PM_2.5 increases of 27.6% under downwind conditions, decreasing to 5.9% when all wind directions were considered.

In this study Spearman correlations confirmed that plume transport variables, particularly the north–south component of wind direction, and wind speed, were the dominant predictors of PM_2.5 increments (Table 4).

At the urban site, %ΔC shows statistically significant moderate correlation with the north–south wind component (wd_cos, ρ = 0.34), while correlations with the east–west component are negligible. This pattern is consistent with the geometry of the port–city axis and indicates that plume alignment and advection efficiency are the primary controls on short-term concentration changes. Merico et al. [10] reported for the port of Brindisi that shipping contributions to pollutant concentrations, including downwind increases of PM_2.5, have a plume-like behaviour with relatively short and intense peaks modulated by port activities (arrival, departure, and hoteling of ships) and local meteorology. Perez et al. [15] showed that shipping contributions to PM_2.5 at monitoring sites depend strongly on wind sector and source–receptor geometry, with highest contributions when the site is downwind of the port. Xu et al. [16] found that ship-emission tracers were significantly elevated at the port-industrial site compared to the urban site, but that under certain wind sectors, those tracer signals also appeared at the urban site, showing transport from port to city.

The wind speed shows statistically significant moderate correlation with %ΔC (ρ = 0.33). Although higher wind speeds generally promote dilution [5,39], the findings of this study suggest that at ~0.8 km downwind, the increased advection and more consistent transport of the plumes toward the receptor likely outweigh dispersive losses: Stronger, steady northerlies decrease travel time, reduce meandering, and maintain the plume at receptor height via building/ship downwash [40]. Thus, the positive sign of the correlation is appropriate near the source, while at greater distances it would likely switch.

According to the standardized EMEP/EEA [41] methodology ship emissions are estimated based on engine power, load factor, activity duration, and emission factors that depend on fuel type, engine type, and the operational phase of the voyage (cruise, manoeuvring, and hoteling). In this study, ship’s technological and operational parameters (main engine power and time at port) showed negligible, non-significant correlations with PM_2.5 levels. This does not imply they are irrelevant; rather, their effects are conditional on wind alignment and load. Main engine power is probably a poor proxy for instantaneous auxiliary/ manoeuvring load (which is the real driver of emissions).

Mitigation strategies of air pollution from ports include integrating renewable energy sources, smart-grid management, and electrified equipment that can markedly reduce pollution and improve environmental performance [42]. Measures to reduce the impact include shore-power (cold ironing) for vessels with high frequency of use, stricter control of idling trucks during docking periods, optimized ramp logistics and improved scheduling to prevent emission peaks occurring at the same time. Given the dominance of two vessels in the call statistics (ships A and B, Table 1), ship-specific interventions may yield disproportionately large benefits.

The study highlights the suitability of island ports as natural laboratories for studying the impact of ship plumes and demonstrates the practical value of combining calibrated, low-cost sensors with meteorological decomposition and event-based analysis. The methodology is transferable to similar port–city environments and provides a robust foundation for developing effective air-quality management strategies that balance maritime connectivity with public health protection.

5. Conclusions and Recommendations

This study quantified short-term PM_2.5 concentration changes associated with 170 ship docking events at the Port of Chios during February and August 2022, using calibrated low-cost sensors deployed at an urban site (0.8 km south of the port) and a suburban site (2.2 km NNW). An event-based concentration enhancement metric (%ΔC), referenced to a three-hour pre-arrival background, was applied to isolate port-related impacts.

The results demonstrate that ship-related PM_2.5 enhancements at the urban site were strongly dependent on wind direction and season. Under northerly winds, which align the port with the urban receptor, mean PM_2.5 increases reached 5.0 μg m⁻³ (48%) in August and 0.8 μg m⁻³ (6.1%) in February. When both months were combined across all wind directions, the urban site exhibited an average increase of 1.7 μg m⁻³ (19%), whereas the suburban site showed negligible changes (0.1 μg m⁻³, 3%). These spatial contrasts indicate that ship-related PM_2.5 impacts were confined to the near-field, downwind urban area.

Uncertainty analysis based on median statistics further confirmed the dominant role of wind direction. For northerly winds, the median %ΔC at the urban site was 25.5%, with a well-constrained bootstrapped 95% confidence interval (18.4–31.5%), supported by a large sample size (n = 81). In contrast, southerly and other wind sectors exhibited median %ΔC values close to zero with wide confidence intervals, reflecting limited statistical power. Excluding the poorly sampled southerly sector yielded similar median %ΔC values, indicating that the directional conclusions are robust.

Correlation analysis showed that %ΔC at the urban site was moderately and significantly associated with wind speed and the north–south wind component, while no statistically significant relationships were found with ship main engine power or time at port. At the suburban site, correlations with meteorological and operational variables were weak and largely non-significant, consistent with its greater distance from the port and less frequent downwind alignment.

Despite these clear directional and seasonal patterns, several limitations should be acknowledged. Meteorological data were obtained from the Hellenic National Meteorological Service station at Omiros Airport, located 2.7 km south of the port. Although of high quality, these data may not fully capture fine-scale coastal flows, including sea–land breeze transitions and wind channelling within the complex urban waterfront environment.

In addition, the analysis relied on hourly mean PM_2.5 concentrations, which can mask short-lived, high-intensity peaks occurring over a few-minute timescale during ship docking and manoeuvring. As most docking events lasted less than one hour, PM_2.5 concentrations during these periods were often represented by a single hourly value, precluding a statistically meaningful estimation of intra-event variability. Consequently, the derived %ΔC values should be interpreted as conservative estimates of port-related enhancements. Higher temporal resolution measurements would better capture transient peaks and improve source attribution.

Operational variables such as total installed engine power and time at port were also simplified proxies for real-time emissions. These metrics do not account for auxiliary engine load, manoeuvring intensity, maintenance practices, or distinct operational phases (approach, berthing, hoteling, vehicle loading), all of which influence PM_2.5 emissions. Moreover, port-related emissions are affected by associated passenger car and truck traffic, which varies substantially between winter and summer. The limited number of observations under southerly winds, particularly in August, further restricts the reliability of results for these conditions and highlights the need for longer datasets.

Finally, although the PurpleAir sensors were carefully calibrated and validated, low-cost optical particle counters can exhibit reduced performance under high humidity, rapidly changing aerosol composition, or near strong exhaust plumes. While correction equations were applied, residual uncertainties relative to reference-grade instruments remain.

Future research should prioritize in situ meteorological measurements to better resolve local flow structures. High-temporal-resolution measurements of PM_2.5, black carbon, and ultrafine particles would allow individual ship plumes to be identified with greater accuracy. In parallel, ship-specific emission estimates derived from AIS-based power modelling or direct plume sampling would enable differentiation between vessels and operational modes. The application of dispersion modelling is recommended to quantify plume behaviour within the complex coastal setting and to assess the effectiveness of mitigation measures such as shore power. Finally, multi-year monitoring would help characterize interannual variability, track fleet evolution, and evaluate the impacts of forthcoming regulatory measures (e.g., FuelEU Maritime within the Fit-for-55 framework).

Collectively, these advances would strengthen causal inference, improve quantification of ship-related PM_2.5 contributions, and support evidence-based mitigation strategies for island port cities.