Modelling of Glass Soiling Due to Air Pollution Exposure at Urban and National Scales: Coimbra (Portugal) Case Study

Abstract

Impacts of air pollution are not limited to human health and ecosystems, but are also important for building materials. The main objective of this study is the quantification and mapping of air pollution effects on the materials, namely the soiling effect of modern glass. An integrated modelling approach was implemented to quantify and analyze the spatial distribution of glass soiling due to exposure to air pollution. The methodology is based on an integrated modelling approach (transportation-emissions-dispersion modelling) applied with high spatial resolution for Coimbra (Portugal) urban area and compared with national scale modelling, showing the important contribution of local pollution sources affecting spatial variability in the soiling effect. Air quality data from CAMS (Copernicus Atmosphere Monitoring Service) were used to quantify the soiling effect at national scale. The results are presented and analyzed in terms of haze. The results obtained at national scale suggest that the average time to reach a 1% haze is 320 days, and this time is reduced to 180 days in the most affected areas. However, urban scale modelling applied with a detailed characterization of local pollution sources in Coimbra provides significantly different results and reveals that in the most affected areas, near road traffic, 1% haze could be reached in approximately 80 days. The methodology proposed and implemented in this study provides relevant information for the maintenance and preservation of building materials and highlights the importance of integrated modelling with high spatial resolution for the assessment of the soiling effect in the built environment.

1. Introduction

The impacts of air pollution are usually associated with human health and ecosystem degradation. However, it is also one of the deterioration factors that affect construction materials, including historic buildings and monuments [1,2,3,4]. While buildings naturally deteriorate due to environmental factors like sunlight, wind, humidity, and rain, air pollution accelerates the natural deterioration processes of materials, reducing their esthetic value and lifespan [3,5,6]. Additionally, air pollution leads to corrosion (caused by acidifying compounds), biodegradation, wear, and colour fading (mainly caused by O₃), as well as soiling effects, resulting in the change in optical properties and darkening of surfaces exposed to atmospheric pollution.

The built environment, and particularly urban areas, face significant challenges from air pollution, driven by contributions from local sources, such as road traffic, biomass burning, and emissions from the construction sector, but also due to long-range transport of air pollution from distant sources [1,3,7]. The pollutants can interact with surfaces in the built environment, resulting in undesirable alterations and leaving traces of these interactions [8], while some alterations, such as soiling, are largely reversible as the accumulated particles can typically be removed through cleaning. Deterioration, on the other hand, is an irreversible process where air pollutants chemically or physically alter the material itself, leading to structural damage, corrosion, or surface wear [9].

Particulate matter (PM), sulphur dioxide (SO₂), and nitrogen dioxide (NO₂) are recognized as the main air pollutants responsible for the most visible and detectable degradation of building materials [10,11]. Given that urban areas host the bulk of the built environment, including heritage structures, the potential for damage due to soiling is significantly heightened compared to rural areas [6].

The soiling of man-made surfaces is probably the most visible damage evidence linked to air pollution impacts [6]. The extent of soiling is influenced by multiple factors, encompassing the material’s properties (such as porosity, colour, and surface texture), atmospheric aerosol characteristics (including size distribution, concentration, and aerosol type), the surface’s exposure conditions (wall inclinations, architectural ledges or saliences, sheltering from rainfall, etc.), and the prevailing climate conditions [6,11]. Not only buildings are affected by soiling, but all exposed surfaces are, including vehicles and photovoltaic modules [6,7]. Thus, the efficiency of solar panels is significantly compromised by soiling. As dust settles on solar panels, reducing irradiance and restricting solar flux, power loss becomes inevitable [12].

Understanding the interplay and relative impact of various degradation processes on the overall surface condition is challenging since soiling is often associated with other processes such as surface wear and corrosion [7]. For example, soiling rates can increase when associated with other degradation processes that occur in glass, such as weathering due to attack by airborne particles, mainly by K and Ca compounds, as well as chlorides (NaCl and KCl), causing depths and cracks in the glass surfaces so that more particle matter is encouraged to deposit on these surfaces, increasing soiling ratios [13]. Despite this complexity, soiling frequently emerges as a primary rationale for cleaning, particularly for windows [7].

The repercussions of soiling are not uniformly distributed, being more pronounced in areas adjacent to heavy traffic routes. Exhaust emissions from road traffic due to incomplete combustion of fossil fuels and dark particles from tyre and brake wear abrasion are identified as substantial contributors to PM pollution [14,15] and soiling on the facades along the main urban roads [16].

To quantify soiling effects, several metrics could be used. Thus, for opaque surfaces (stone, concrete, plastic, wood), changes in surface gloss, lightness, and any other factors directly tied to surface reflectance serve as useful indicators for evaluating visual degradation caused by soiling. However, for transparent materials, transmittance measurements are favoured over reflectance [17], with soiling typically expressed as haze (Equation (1)), a measure of the ratio between diffused light transmission (Td) and direct transmission (TL) in the initial light direction [6,18].

$$\mathrm{Haze}\%={\displaystyle \frac{\mathrm{T}\mathrm{d}}{\mathrm{T}\mathrm{L}}}\times 100$$

(1)

A haze exceeding 1% is perceptible to the naked human eye [6,17], indicating a significant decline in esthetic appeal and necessitating increased cleaning and maintenance efforts [7]. In urban areas, where the problem is typically more severe, the frequency and therefore the cost of cleaning operations required to restore the original visual appearance of buildings can be high [17,18].

Understanding the relationship between air pollution and material degradation is essential to figuring out the magnitude of the problem, identifying the most affected areas, and establishing strategies for the sustainable maintenance and preservation of external materials [1]. Traditionally, the quantification and assessment of material degradation primarily involve two approaches: (i) conducting measurement campaigns [16,17,18,19,20], which require great effort and are based on test sites, providing data at selected locations, or (ii) employing computational modelling that takes into account dose–response functions and enables the spatial distribution of data [6,7,21,22,23].

Dose–response functions (DRF) provide information on the soiling rates of materials in response to different levels or loads of pollutants, together with meteorological and air pollution parameters [21,24], and are important features in studying the corrosion and soiling of construction materials, especially in the context of exposure to environmental pollutants in urban areas. Computational modelling based on dose–response functions allows for a better understanding of soiling problems at different scales and facilitates the replicability of the methodology [7,25].

This paper focuses on the study of soiling of transparent materials, and the main objective is the implementation of a modelling approach for quantification and mapping of the soiling effects in modern glass due to exposure to air pollution, emphasizing both national (Portugal) and urban scales (Coimbra). This research provides information for establishing conservation strategies and proactive measures to mitigate the effects air pollution in the built environment by mapping soiling patterns and evaluating factors contributing to its magnitude.

2. Methodology

2.1. Atmospheric Soiling Quantification

The methodology adopted for assessing atmospheric soiling is based on modelling and mapping the pollutants concentrations and soiling at multi-scale following the main recommendations of the International Co-operative Programme on Effects on Materials (ICP Materials) [24]. The methodological framework implemented in this study includes four steps: (i) estimation of pollutant concentrations; (ii) quantification of atmospheric soiling; (iii) spatial distribution at national and urban scales; and (iv) a comparison between the results (Figure 1).

To implement the methodology, atmospheric pollution data for mainland Portugal were obtained from the European Copernicus Atmosphere Monitoring Service (CAMS). For the analysis at urban scale, an integrated modelling approach (transportation–emissions–air pollution dispersion) was implemented for Coimbra, allowing for the assessment of local sources of pollution with high spatial resolution. The methodology applied on the national scale will be detailed in Section 2.2 and for the urban scale in Section 2.3.

The soiling effect for transparent materials was estimated according to a multi-pollutant dose–response function derived from a multilinear regression equation (Equation (2)), employed based on the work of Lombardo et al. [17], for a chemically inert, modern silica–soda–lime glass. The results are expressed as haze H (%) which is influenced by concentrations of SO₂, NO₂, and PM₁₀ (µg m⁻³) and the exposure time t (days):

$$\mathrm{H}={\displaystyle \frac{0.2529\left[{\mathrm{S}\mathrm{O}}_2\right]+0.1080\left[{\mathrm{N}\mathrm{O}}_2\right]+0.1473\left[{\mathrm{P}\mathrm{M}}_{10}\right]}{1+{\left(\frac{382}{\mathrm{t}}\right)}^{1.86}}}$$

(2)

This dose–response function was derived from field research programmes conducted under the UNECE ICP Materials and based on data gathered from various field campaigns in different environments for sheltered conditions [24].

2.2. National Scale Air Quality Modelling

Mainland Portugal, considered in this study, is located at the southwestern tip of the Iberian Peninsula, bordered north and east by Spain and to the west and south by the Atlantic Ocean (Figure 2). It has a temperate climate with rainy winters and dry summers [26]. Demographically, it has two large metropolitan areas: the Lisbon Metropolitan Area, with approximately 2.8 million inhabitants [27], divided into two sub-regions, Greater Lisbon and the Setubal Peninsula, and the Porto Metropolitan Area, with approximately 1.7 million inhabitants [27].

As mentioned previously, air pollution data at the national scale were obtained from CAMS, and were processed to obtain annual average concentrations considering reanalysis data from the multi-model ENSEMBLE approach [28,29]. CAMS reanalysis data combine model simulations with observations from multiple sources to account for uncertainties and improve data reliability, creating a more accurate representation of atmospheric conditions [29,30].

The ENSEMBLE products have a spatial grid resolution of 0.1° × 0.1° (~10 km × 10 km) and are calculated for different vertical layers, ranging from the surface (at the monitoring station level) to 5000 m in altitude. Meteorological data from the European Centre for Medium-Range Weather Forecasts (ECMWF) drive the chemical transport models, maintaining the same spatial and temporal resolution used by CAMS [28,29].

The analysis and soiling quantification involved extracting CAMS-validated reanalysis data for surface-level concentrations of PM₁₀, NO₂, and SO₂ in NetCDF format. As the files originally extracted from CAMS provide hourly concentrations, the data were processed in Geographic Information System (GIS) to obtain annual average concentrations for each pollutant.

2.3. Urban Scale Air Quality Modelling

Coimbra urban area (Figure 2) is a part of Municipality and Region with the same name, located in the centre of mainland Portugal. The Municipality of Coimbra had a resident population around 141,000 inhabitants in 2021 and average population density of around about 440 inhab/km² [27]. Coimbra is distinguished by its historical and architectural heritage; its central area is classified as a UNESCO World Heritage site. It is also home to one of the oldest universities in Europe, the University of Coimbra, known as a university city, although about 25% of the resident population is elderly. Local air pollution in Coimbra is mainly related with road traffic emissions, commercial activities, and residential biomass burning [14,15,31,32,33,34,35,36].

An integrated modelling chain was established to assess air quality at the urban scale for Coimbra. This approach involved three interconnected models: (i) macroscopic transportation model, (ii) emission model, and (iii) air pollution dispersion model. This integrated approach has previously been applied in other studies to assess atmospheric pollution on an urban scale, with different research objectives, and for different pollutants [14,31,32].

(i)

Macroscopic transportation model

The VISUM 13 software [37], a versatile traffic modelling tool, was employed for multimodal analysis. This model integrates various transportation modes into a reliable network model [31] and is one of the most used to provide traffic volume estimation [33]. The VISUM transportation model implemented in this study is based on the four-step approach and includes trip generation, trip distribution, modal split, and traffic assignment. Detailed characterization of the road network and data on origin-destination (O-D), trip duration, route taken, and average speed are available for the city-scale simulations. The outputs from the transportation model on traffic volume and average speed for each road segment are used as inputs to the emission model to quantify road traffic emissions with high temporal and spatial resolution.

(ii)

Emission model

The Traffic Emission and Energy Consumption Model (QTraffic) [33] was applied at urban scale to estimate emissions induced by vehicles considering exhaust and non-exhaust processes based on the updated European guidelines for emission factors [34,38]. The QTraffic model is designed for line sources, being able to estimate emissions by considering emission factors as a function of average speed and vehicle class (based on engine age, type, capacity, vehicle weight, fuel type, and emission reduction technology) [14]. Consequently, the required input data for the emissions model have been compiled according to statistical information on fleet composition, spatial data on road network, traffic volume, and vehicle speed provided by VISUM. Traffic emission data calculated at this step were important inputs to simulate air pollution dispersion at urban scale [14].

(iii)

Air quality model

The advanced Gaussian dispersion model adapted for line sources, known as ADMS-Roads [39], was employed to obtain the spatial distribution of hourly pollutant concentrations in Coimbra’s urban area, utilizing a 30 m spatial grid resolution in combination with the street canyon module [40] to consider the influence of buildings along the street on airflow movement and its effects on the dispersion of pollutants. In addition to the local emission sources explicitly considered in the model, background pollution levels were characterized based on regional modelling provided by CAMS. The hourly outputs provided by ADMS were processed to obtain annual average concentrations for PM₁₀, NO₂, and SO₂, considering a selected year (2018).

To effectively communicate results, maps were created that highlight the spatial distribution of pollutant concentrations. This comprehensive methodology allowed for a detailed assessment of urban air quality.

3. Results and Discussions

The results of air quality modelling, conducted at national and urban scales, along with the presentation of the atmospheric soiling maps for mainland Portugal and Coimbra urban area are presented and discussed in this section.

3.1. Air Quality

Spatial variations in the annual mean concentrations for PM₁₀, NO₂, and SO₂ in 2018 for Portugal are presented in Figure 3. These maps are obtained by processing CAMS data with 0.1° × 0.1° grid resolution. The maps reveal distinct patterns for each analyzed pollutant.

NO₂ concentrations (Figure 3b) exhibit a strong correlation with road traffic emissions, showcasing higher pollution levels in urban areas, with greater prevalence in the most populous municipalities in Portugal, namely Lisbon, Porto/Vila Nova de Gaia, and Setubal Peninsula. Conversely, SO₂ concentration (Figure 3c) peaks are in industrial zones, reflecting their association with industrial activities. With emphasis on the location where the Sines refinery is installed in the south of the Setubal district, where the highest values are observed. The spatial variation of annual mean PM₁₀ concentrations (Figure 3a) presents a smoother pattern influenced by diverse pollution sources and the long-range transport of mineral dust. Overall, densely populated coastal zones emerge as the most significantly affected areas for all pollutants.

While the CAMS data presented in Figure 3 offer valuable insights on air pollution at national scale, their applicability on the urban scale is limited due to the size of grid cells (0.1° × 0.1°) when compared with the results obtained at higher spatial resolution, with grid cells of 30 m × 30 m presented in Figure 4 for Coimbra urban area, where emissions arising from road transport are very important.

It is observed that the annual mean PM₁₀ concentrations for Coimbra derived from urban-scale modelling and mapping reach values of 40 µg/m³, in contrast to the 17 µg/m³ provided from the national-scale approach (Figure 3a). A similar spatial pattern is observed for NO₂ (Figure 4b), reaching concentrations of 105 µg/m³, significantly above the background levels. This emphasizes the importance of employing high-resolution modelling approaches when assessing air quality at the city level, as they reveal a pattern with high spatial gradients in pollutant concentrations associated with local sources. The spatial distribution of air pollution close to main roads supports the substantial contribution of road traffic emissions. Moreover, meteorological conditions and urban morphology are also important factors that contribute to the spatial heterogeneity of pollution levels in the urban environment [31].

Although there are no specific reference values for the exposure of construction material to atmospheric pollution, the Ambient Air Quality Directive (2008/50/EC) has established air quality standards for the key air pollutants in the European Union (EU) [41]. The upcoming revision of the Ambient Air Quality Directive aims to align these standards more closely with the World Health Organization guidelines [42] by setting interim EU air quality targets for 2030 and proposing 20 µg/m³ as a new limit value for annual average concentrations of PM₁₀, NO₂, and SO₂ [43]. This revision is part of the EU’s broader strategy to achieve zero air pollution at the latest by 2050 [44].

The analysis of the modelling results reveals no exceedance of the limit values at national scale (Figure 3). However, the urban scale modelling with higher spatial resolution emphasizes the concentrations of PM₁₀ and NO₂ above the established limits in several locations. While such analyses allow for the primary identification of critical areas and conjecture regarding the main sources of pollution, modelling on a national scale can lead to the smoothing of results, hiding important local variations in the distribution of air pollutants.

3.2. Glass Soiling

The spatial variability of glass soiling for mainland Portugal is analyzed in terms of the number of days required to reach a 1% haze, utilizing a multipollutant dose–response function for the soiling of transparent material, as previously described (Equation (2)). The resulting soiling map (Figure 5) unveils significant variations in the exposure time needed to attain a 1% haze and illustrates its spatial distribution across Portugal.

Modelling results based on air pollution data for 2018 indicate that, on a national scale, the average number of days to reach a 1% haze (H = 1%) on glass surfaces is 320 days, while in the most affected areas, H = 1% is reached in just 180 days (

Table 1. Soiling results for mainland Portugal.

Parameters	Description	Mean	STD	Min	Max	P95
t [days]	National domain (H = 1%)	320	63	184	533	427
H [%]	National domain (t = 365 days)	1.2	0.3	0.7	2.4	1.8

). The average loss of transparency at the national domain over a one-year exposure to air pollution is H = 1.2%, with the most affected areas exhibiting a maximum haze of 2.4% during the same period.

Coastal areas exhibit the highest soiling potentials, with a 1% haze achieved in approximately six months. This indicates a heightened necessity for more frequent cleaning operations in these regions, thereby incurring increased maintenance costs.

The soiling map for Coimbra city, produced by applying the same dose–response function but using high spatial resolution of air quality data, can be seen in Figure 6a. The results indicate an average of 185 days to reach a 1% haze (H = 1%) on glass surfaces, with the most affected areas reaching H = 1% in just 84 days (see

Table 2. Soiling results for Coimbra urban area.

Parameters	Description	Mean	STD	Min	Max	P95
t [days] to reach H = 1%	Results from National scale modelling	233	-	223	238	-
	Results from Urban scale Modelling	185	19	84	206	202
H [%] reached after the exposure time of t = 365 days	Results from National scale modelling	1.7	-	1.6	1.8	-
	Results from Urban scale Modelling	2.4	0.5	2.0	8.5	3.4

).

For one-year of exposure to air pollution, the mean loss of transparency for the urban domain is H = 2.4%, with the most affected areas experiencing a maximum haze of 8.5% during the same period. These results are significantly different from the haze estimations based on air quality data, with gross spatial resolution provided by CAMS for Coimbra urban area (223–238 days for reach H = 1% and the maximum haze was 1.8%), as presented in Table 2 (due to the limited number of grid cells overlapping the urban area, STD and P95% from national scale modelling are not presented).

The comparative analysis of the results (Table 2) was obtained for the Coimbra region through modelling at different scales, a national scale and an urban scale, highlighting the differences in values and the importance of carrying out atmospheric soiling analyses with a higher spatial resolution to account for the local sources of pollution and assess potential impacts on glass surfaces.

4. Conclusions

This study investigated the soiling of modern glass due to exposure to air pollution based on a modelling approach. This methodology involved the estimation of pollutant concentrations, the quantification of atmospheric soiling, and the analysis of spatial distribution of soiling at national (Portugal) and urban (Coimbra) scales.

The results show significant spatial variability in soiling effects across different scales, highlighting the importance of spatial resolution in environmental modelling. Thus, national scale modelling allows for the identification of general trends and spatial patterns, providing valuable insights for national-level pollution control policies. However, when the spatial resolution is refined to an urban scale, as in the Coimbra case study, the results underscore the limitations of coarse-scale models in capturing the spatial variability in local pollution. In terms of soiling, on a national scale, the average time to reach a 1% haze on glass surfaces is approximately 320 days, but this decreases to 180 days in the most polluted areas in coastal zone. However, the urban-scale modelling provides a more detailed view, showing that near the road traffic sources, 1% haze can occur in just 80 days. This suggests that urban areas with high traffic emissions are more susceptible to soiling, emphasizing the need for more frequent maintenance of glass surfaces in such environments. These findings reveal the importance of using high-resolution models to accurately assess air quality and its effects in urban environments, where local pollution sources and urban morphology play an important role.

The obtained results emphasize the relevance of employing a modelling approach to quantify the impacts of air pollution on building materials and highlight the need to implement air pollution mitigation strategies. The urban-scale analysis might be a valuable tool for estimating average maintenance intervals and assessing damage costs attributable to air pollution. The findings enhance our understanding of the current issues arising from atmospheric pollution and help to identify sites requiring special attention, typically located in areas with significant anthropogenic activity. This study serves as a foundation for further research and site-specific analyses, highlighting the need for additional measures and policies to reduce atmospheric pollution in cities to prevent future damage to outdoor materials.

This study is a pilot effort, and some limitations could be highlighted. Thus, in addition to road transport emissions, other local sources of air pollution, such as biomass burning and emissions from the construction sector, could be important when assessing glass soiling and should be considered. Expanding the analysis to include other materials beyond modern glass is also relevant; however, this would require supplementary data, such as information on precipitation acidity. While this was beyond the scope of the current study, it will be addressed in future research.

As a next step, microscale modelling could be explored to investigate the link between air pollution, the built environment, and mitigation strategies. Thus, green infrastructure and nature-based solutions hold significant promise as a strategic recourse for addressing air pollution issues and consequently protecting outdoor materials. However, the non-homogeneous distribution of pollution sources and receptors in cities, along with the spatial and temporal variability in dispersion conditions and the challenge of obtaining air quality data with sufficient spatial resolution make it difficult to quantify the contribution of green solutions in improving urban air quality. In this context, modelling approaches can provide a valuable framework for the quantitative evaluation and comparison of different mitigation strategies.