1 Introduction

The impact of air pollution on human health is a critical issue, particularly in urban environments. It is projected that by 2050, 70% of the global population will reside in urban areas (United Nations 2018). These areas will host the main sources of air pollution, including transportation, industrial production, and domestic heating. Among these sources, vehicle emissions are often identified as the leading cause of urban air pollution (Di Sabatino et al. 2018; Huang et al. 2020). Numerous studies have highlighted the need for robust modelling approaches to understand and mitigate the effects of vehicular emissions on urban air quality (Kumar et al. 2011; Sokhi et al. 2018). In recent years, advancements in air quality modelling have incorporated data from diverse sources, including satellite observations, ground-based sensors and advanced computational models. These integrated approaches offer more accurate and comprehensive assessments of urban air pollution (Jiang et al. 2021; Liu et al. 2022). Moreover, the development of machine learning techniques has enhanced the ability to predict pollution levels and identify key factors influencing air quality (Zhang et al. 2021).

Considering this, the role of sustainable mobility becomes essential. The European Sustainable Development Strategy, adopted by the European Council in 2006, highlights the importance of ensuring that transport systems meet economic, social, and environmental needs while minimizing adverse impacts (European Council 2006; EEA, 2020). Various air pollution modelling approaches have been developed, including Computational Fluid Dynamics models, Gaussian dispersion models, and hybrid models integrating multiple data sources (Zhang et al. 2013; Lateb et al. 2016).

This study aims to develop functional relationships between urban morphology and the concentration and dispersion of air pollutants from vehicular traffic. These relationships, once integrated into the National Recovery and Resilience Plan (PNRR) - National Center for Sustainable Mobility (MOST) project, will enable real-time assessment of vehicular pollutant impacts in urban areas. This will be achieved within the whole project by integrating data from dispersion models, meteorological reanalyses, and in situ measurements over various time periods. To accomplish the objectives, various Italian cities characterized by differing meteorology, urban morphology, and vehicular traffic patterns will be studied. This paper presents the initial analyses conducted in urban areas within the cities of Lecce and Bari, both located in the Apulia region. Through these analyses, the aim is to enhance the understanding of urban air pollution dynamics and contribute to more effective urban planning and pollution mitigation strategies.

2 Methodology

2.1 ADMS-roads modelling

Lecce has a very compact and irregular city centre, surrounded by compact neighbourhoods built in the last decades. Even Bari has a compact and irregular old town that extends into the sea, while the more recent neighbourhoods have been built according to an orthogonal grid (Esposito et al. 2023). Figure 1 shows the location of the two cities and the study areas, i.e. eight areas in Lecce (Caserma Pica, Duomo, Garigliano, Libertini, Piazzale Rudiae, Porta Napoli, Stadio, Via Merine), and six areas in Bari (Bari vecchia, Carbonara, Caserma Picca, Murat, Parco 2 Giugno, Poggiofranco).

Fig. 1
figure 1

Aerial view of (a) Italy with the locations of the cities of Lecce and Bari, (b) the study areas in Lecce, and (c) the study areas in Bari, including the positions of meteorological stations (purple dots)

Full size image

In each study area, various air pollution dispersion scenarios were created using the ADMS-Roads (Atmospheric Dispersion Modelling System) software (www.cerc.co.uk), a sophisticated tool for modelling atmospheric pollution, specifically designed to evaluate and predict the impact of vehicular traffic emissions within defined areas. This software produces high-resolution maps illustrating the distribution of pollutants. The model accuracy has been confirmed through numerous studies, including comparisons with data from the UK’s Automatic Urban and Rural Network (AURN) and specific validation exercises using standard field, laboratory, and numerical datasets. Moreover, it has been favourably compared with other systems used by the EU and the US EPA (https://www.cerc.co.uk/environmental-software/model-validation.html). ADMS-Roads has been employed with comprehensive inputs including detailed morphology and geometry data, emission sources characterization, and meteorological conditions as detailed below.

In addition to modelling software, pollutant monitoring sensors, meteorological analysis tools and traffic cameras will be employed. Meteorological and traffic data will be used for input into ADMS-Roads and to validate the modelling results. This comprehensive approach will ensure accurate assessment and prediction of air pollution dispersion in the study areas.

2.2 Characterization and selection of the study areas using QGIS

The selection of the study areas (Fig. 1) was primarily carried out using QGIS software, with a focus on the morphological parameter, named planar area index (λp). It represents the percentage ratio of building plan area to total plan area and was calculated as the average value for each area. Additionally, two other morphological parameters were considered: the aspect ratio (AR), which denotes the mean ratio of street canyon height to width, and the mean building height (H), representing the geometric average of building heights within a specified area. Initially, these parameters were computed using QGIS on 100-meter square cells covering the entire cities of Lecce and Bari. Spatially averaged values were then determined for all study areas, each measuring 500 × 500 m. The study areas were selected with increasing λp within each city while maintaining similar values between cities. This approach enables comparison between the cities to evaluate morphology-pollutant concentration trends and assess their consistency across different urban environments. Each city included at least one area from each classification category: sparse canopy (low λp), intermediate canopy (λp around 0.25), compact canopy (λp around 0.40), and very compact (high λp), following the frameworks of Buccolieri et al. (2010) and Oke (2017). Initially, eight areas were investigated in Lecce as a comprehensive test, suggesting that six areas, as done in Bari, are sufficient for validation and for future studies in other Italian cities.

In parallel, the QGIS plugin “UMEP” was used to calculate the roughness length (z0) for each grid cell, using the roughness calculation method developed by MacDonald et al. (1998). The roughness values for the city of Lecce were then used as input in ADMS-Roads to account for the effects of complex terrain. For Bari it was not possible to calculate z0 for each grid cell and it was therefore decided to assign a unique value of z0 to each area studied, estimated based on the morphological similarities to the areas in Lecce.

2.3 Processing of meteorological parameters

Hourly meteorological data (wind speed and direction, air temperature, relative humidity and solar radiation), covering the period from 2019 to 2021, were collected from the CNR-ISAC meteorological station for the city of Lecce, and from the Regional Civil Protection meteorological station (Bari Idrografico) and the ARPA Puglia meteorological station (Carbonara) for the city of Bari. Please note that, for each city, the same meteorological station data were used for all areas due to limited station availability. ADMS-Roads accommodates this by adjusting for varying roughness lengths between the meteorological site and each study area (see the previous subsection).

To create the necessary meteorological input for ADMS-Roads, the collected data were processed using MATLAB software to create four distinct “typical seasons” representative of the whole period 2019–2021. Each season contains hourly data spanning three months (Fall: Sept-Nov, Winter: Dec-Feb, Spring: Mar-May, Summer: Jun-Aug). Wind speed, temperature, relative humidity, and solar radiation were averaged hour by hour, while the wind direction was taken as the prevailing direction. The resulting four hourly meteorological files were used as input in ADMS-Roads which, using its internal meteorological preprocessor, calculated additional data necessary for a detailed characterization of local meteorology and atmospheric turbulence, which is essential for evaluating pollutant dispersion. All the simulations were conducted on an hourly basis.

The “summer season” is dealt with in detail in this work and Fig. 2 presents the wind roses for the this “typical season”. The percentage of missing data was less than 3% for Lecce and less than 1% for Bari.

Fig. 2
figure 2

Wind roses for the “summer season” for the cities of (a) Lecce and (b) Bari

Full size image

2.4 Modelling of urban road transport and pollutant dispersion

The road geometries were mapped using QGIS, where descriptive parameters such as width, length, and average height of the road canyons were calculated and associated with each modelled road segment. Additionally, each road source was assigned numerical values representing vehicle types, their number and average speed per hour. It was assumed that there could be a maximum of 23,250 light vehicles per study area travelling at an average speed of 30 km/h. These parameters - standardised across all areas to ensure consistency - were chosen based on an expected mean daily value of 8,000 vehicles per road. The distribution of vehicles on different roads within each area was based on traffic data provided by Google Traffic. Additionally, the number of vehicles on each road varies according to the time of day, including weekdays, weekends, and holidays, following the “3DayDiurnal” scheme, which shows higher emission values during weekdays and working hours.

To estimate emissions from road transport, the internal ADMS-Roads dataset “UK EFT v9.0 (2VC)” was used, specifying the emission year as 2022 and designating “England (urban)” as the road type. The modelling included the effects of street canyons and traffic-induced turbulence, with the street canyon module based on the Danish Operational Street Pollution Model (OSPM, Hertel and Berkowicz 1990). It is worth noting that, while the ‘UK EFT v9.0 (2VC)’ emission dataset and ‘England (urban)’ road type were employed due to their default availability in ADMS-Roads, the consistency across study areas was maintained. This approach allowed for a robust comparative analysis of urban morphology’s influence on pollutant dispersion. Although the meteorological data spans 2019–2021, the 2022 emissions dataset was employed to ensure the most up-to-date emissions factors, as this study primarily focuses on the impact of urban morphology on pollutant dispersion rather than exact concentration levels. Future improvements will focus on incorporating region-specific emission data and updated meteorological data beyond 2019–2021 to enhance the accuracy of these findings without altering the fundamental insights gained.

The output concentrations for two vehicular pollutants were requested: the hourly average concentration of NO2 (µg/m3) and the daily average concentration of PM10 (µg/m3) for each grid point representing the study areas. Each study area was divided using a 31 × 31 points “source-oriented” grid, which included additional points close to road sources to capture all features of the concentration distribution within the study areas.

3 Results

Tables 1 and 2 provide PM10 concentrations (µg/m³) derived from the averaging of daily average concentrations across grid points, with the objective of obtaining representative values for the “summer season”. These tables also include the spatially averaged values of λp and AR. Figure 3 shows the data from the tables, comparing PM10 concentrations as the spatially averaged value of λp varies in the cities of Lecce and Bari.

Wind speed plays a crucial role in the dispersion of pollutants emitted by vehicles, in fact, as shown in Fig. 3, the city of Lecce, characterized by lower average wind speeds (Fig. 2), tends to have higher PM10 concentrations than Bari for comparable average λp values. However, a different behaviour is observed for the areas characterized by λp values around 60%, in both cities. In fact, in Lecce, PM10 concentrations show a decrease, in contrast with the increasing trend observed up to λp 44.6%, while in Bari, an opposite trend is observed. PM10 concentrations decrease with increasing λp up to 49.8% and then increase for λp around 60%. This may be attributed to the higher density of buildings associated with higher λp values, leading to lower average wind speeds and consequently lower dispersion of pollutants.

Table 1 Average PM10 concentrations per study area for the city of Lecce for the “summer season”, sorted by λp value
Full size table
Table 2 Average PM10 concentrations per study area for the city of Bari for the “summer season”, sorted by λp value
Full size table
Fig. 3
figure 3

Graph of the trend in daily PM10 concentrations for the “summer season” (averaged over the three years of simulation) for the cities of Lecce and Bari vs. the spatially averaged λp of each study area

Full size image

As an example, Fig. 4a, b shows a contour plot of the average PM10 concentration for the “Garigliano” area in Lecce and the “Parco 2 Giugno” area in Bari during the “summer season”, generated using ADMS-Roads and subsequently processed in QGIS. Referring to Figs. 1 and 2, it is evident that the dispersion of pollutants is influenced by wind direction. During the “summer season”, only winds from the north-north-west prevail in Lecce, while winds from the north-north-west and north-east prevail in Bari. The former are large-scale winds occurring in both cities, while the latter are sea breezes recorded only in Bari due to its coastal location. Therefore, in Fig. 4, it can be observed that the dispersion of PM10 increases in the direction opposite to that of the wind source. Specifically, for both Lecce and Bari, this results in a larger area of dispersion south of the road source.

Fig. 4
figure 4

Contour plot of the 3-year mean values of PM10 concentration in “summer season” for (a) the “Garigliano” area in Lecce and (b) the “Parco 2 Giugno” area in Bari

Full size image

4 Conclusions

This study represents an initial evaluation using simplified inputs to assess the impact of urban morphology on PM10 concentrations in Lecce and Bari. Despite the study’s limitations, including the use of simplified fixed emissions data and emission factors based on data from England, the results highlight significant correlations between morphological characteristics and pollutant concentrations. Specifically, the analysis reveals that lower wind speeds in Lecce contribute to higher PM10 concentrations compared to Bari, emphasizing the role of local meteorology in pollutant dispersion. Interestingly, areas with denser building configurations in Lecce exhibited unexpected decreases in PM10 concentrations, potentially due to complex airflow patterns and canyon effects. Conversely, Bari showed varied trends with urban morphology, indicating the complexity of pollutant dispersion dynamics across different urban layouts.

Looking forward, future works will focus on refining model inputs with more detailed data and integrating sensor measurements to enhance accuracy. This iterative approach will enable a deeper understanding of urban air pollution dynamics and support targeted interventions under the PNRR Italian National Center for Sustainable Mobility (MOST) project. By expanding this research to additional urban areas and incorporating real-time data, the aim is to provide actionable insights for policymakers and urban planners to implement effective pollution mitigation strategies.