Assessing the Impact of Vehicle Speed Limits and Fleet Composition on Air Quality Near a School

Traffic is a major source of urban air pollution that affects health, especially among children. As lower speed limits are commonly applied near schools in many cities, and different governments have different policies on vehicle fleet composition, this research estimated how different speed limits and fleet emissions affect air quality near a primary school. Based on data of traffic, weather, and background air quality records in Dublin from 2013, traffic, emission, and dispersion models were developed to assess the impact of different speed limits and fleet composition changes against current conditions. Outside the school, hypothetical speed limit changes from 30 km/h to 50 km/h could reduce the concentration of NO2 and PM10 by 3% and 2%; shifts in the fleet from diesel to petrol vehicles could reduce these pollutants by 4% and 3% but would increase the traffic-induced concentrations of CO and Benzene by 63% and 35%. These changes had significantly larger impacts on air quality on streets with higher pollutant concentrations. Findings suggest that both road safety and air quality should be considered when determining speed limits. Furthermore, fleet composition has different impacts on different pollutants and there are no clear benefits associated with incentivising either diesel or petrol engine vehicles.


Introduction
Research has shown significant associations between traffic pollution and adverse human health effects related to lung, heart, psychological, and other body systems (e.g., increasing lung cancer, heart disease, dementia, and other health problems) [1][2][3][4][5][6][7][8][9][10][11]. Air pollution is a major environmental problem that causes 6.4 million premature deaths worldwide per year, which is 72% out of 9 million deaths from all types of pollution [12]. Children in particular have been noted to be at high risk of pollution-related disease. Even very low-level exposures to air pollutants during windows of developmental vulnerability can result in disease, disability, and death in childhood and in later life [13]. Traffic related air pollution is not only linked to worsening existing health problems in children, but is also connected with the development of diseases, such as asthma [14]. Traffic related air pollution is harmful to the development of children's nervous system, causing neurodegeneration, neuro-inflammation, and problems connected to cognition [15][16][17]. Many other

Model Development
The modelling chain comprised a traffic model and an emission and dispersion model (See Figure 2). The size of the modelled area was approximately 40 hectares. A primary school was located at the centre of the modelling area. This area was chosen to represent the vicinity of the school, and the traffic and air quality conditions around the school. Distance-decay studies on traffic air pollutant concentrations have found that motorways have impacts on concentrations of NOX and PM typically up to 200m away from the roadside [6]. Roads in the city centre generally have lower traffic volumes than motorways, and a street canyon topography is often present, which concentrates the air pollution and prevents it from dispersing freely. Therefore, roads in the city centre have a smaller area of influence compared to motorways. In addition, many pupils of the school live within the modelled area, and thus its air quality is crucial for their health.
The traffic model was developed in VISUM [32] and was used to assess the traffic volume and vehicle speed on the road network. The traffic condition outputs from the traffic model were used as an important input to the emission and dispersion model, which was developed using the Operational Street Pollution Model (OSPM) [33]. The methodology for emission calculations and emission factors for different types of vehicles that were applied in OSPM were based on the European Emission Model COPERT4 [34]. The study focused on hourly pollutant concentration predictions across a full year, and the implementation of a traffic assignment model (VISUM) and emission calculation method based on hourly average speed (COPERT4) complied with the objective of the study. The details of the development of the traffic model, and the emission and dispersion model, are explained in the following subsections.

Model Development
The modelling chain comprised a traffic model and an emission and dispersion model (See Figure 2). The size of the modelled area was approximately 40 hectares. A primary school was located at the centre of the modelling area. This area was chosen to represent the vicinity of the school, and the traffic and air quality conditions around the school. Distance-decay studies on traffic air pollutant concentrations have found that motorways have impacts on concentrations of NO X and PM typically up to 200m away from the roadside [6]. Roads in the city centre generally have lower traffic volumes than motorways, and a street canyon topography is often present, which concentrates the air pollution and prevents it from dispersing freely. Therefore, roads in the city centre have a smaller area of influence compared to motorways. In addition, many pupils of the school live within the modelled area, and thus its air quality is crucial for their health.
The traffic model was developed in VISUM [32] and was used to assess the traffic volume and vehicle speed on the road network. The traffic condition outputs from the traffic model were used as an important input to the emission and dispersion model, which was developed using the Operational Street Pollution Model (OSPM) [33]. The methodology for emission calculations and emission factors for different types of vehicles that were applied in OSPM were based on the European Emission Model COPERT4 [34]. The study focused on hourly pollutant concentration predictions across a full year, and the implementation of a traffic assignment model (VISUM) and emission calculation method based on hourly average speed (COPERT4) complied with the objective of the study. The details of the development of the traffic model, and the emission and dispersion model, are explained in the following subsections.

Traffic Model
The flow chart of the traffic model is illustrated as part of the modelling chain in Figure 2. The traffic model estimated traffic volume and traffic speed based on speed limit and traffic condition inputs on each road in the network. VISUM was chosen as the traffic model because of the flexibility to adjust traffic volume based on traffic count, and the ability to use Python script to control the modelling process.
The road network was developed in VISUM to represent the roads of the modelling area shown in Figure 1. The original travel demand was presented as an OD matrix for the modelled area and was derived from a traffic model for the city, developed by Tang et al. [25]. This traffic model for the city was obtained by extracting Dublin city from the National Transport Model (NTpM) of Ireland and was calibrated with the annual average traffic count records of 2013 for roads in Dublin.
The original OD matrix was then calibrated with traffic count data throughout the year 2013 to obtain a more precise travel demand. The traffic count data were derived from hourly traffic counter data from Dublin City Council (DCC). The traffic conditions for weekdays during ten months of 2013-excluding July and August-were modelled to represent the days when the traffic pollution in the modelled area were of most relevance to pupils in the school (pupils did not attend school during weekends or in the summer holiday months of July and August). The traffic model was calibrated using one week of hourly traffic data from every two-month period in 2013. Analysis was conducted using a Python script because of the large amount of data involved in hourly travel data across a full year. The traffic data involved in the calibration were chosen to be representative of various traffic conditions throughout the school year (i.e., they included all weekdays, rainy and sunny days, spring, autumn, and winter periods).
During the calibration process, the original OD matrix was assigned to roads to obtain traffic volumes using a user-equilibrium assignment approach. Then the assigned volumes were compared to traffic counts and adjustments were made to OD matrices to obtain more accurate results. This step of the assignment of the OD matrix to roads, the comparison between the assigned volumes and traffic counts, and the adjustment of the OD matrix was repeated several times to obtain an accurate

Traffic Model
The flow chart of the traffic model is illustrated as part of the modelling chain in Figure 2. The traffic model estimated traffic volume and traffic speed based on speed limit and traffic condition inputs on each road in the network. VISUM was chosen as the traffic model because of the flexibility to adjust traffic volume based on traffic count, and the ability to use Python script to control the modelling process.
The road network was developed in VISUM to represent the roads of the modelling area shown in Figure 1. The original travel demand was presented as an OD matrix for the modelled area and was derived from a traffic model for the city, developed by Tang et al. [25]. This traffic model for the city was obtained by extracting Dublin city from the National Transport Model (NTpM) of Ireland and was calibrated with the annual average traffic count records of 2013 for roads in Dublin.
The original OD matrix was then calibrated with traffic count data throughout the year 2013 to obtain a more precise travel demand. The traffic count data were derived from hourly traffic counter data from Dublin City Council (DCC). The traffic conditions for weekdays during ten months of 2013-excluding July and August-were modelled to represent the days when the traffic pollution in the modelled area were of most relevance to pupils in the school (pupils did not attend school during weekends or in the summer holiday months of July and August). The traffic model was calibrated using one week of hourly traffic data from every two-month period in 2013. Analysis was conducted using a Python script because of the large amount of data involved in hourly travel data across a full year. The traffic data involved in the calibration were chosen to be representative of various traffic conditions throughout the school year (i.e., they included all weekdays, rainy and sunny days, spring, autumn, and winter periods).
During the calibration process, the original OD matrix was assigned to roads to obtain traffic volumes using a user-equilibrium assignment approach. Then the assigned volumes were compared to traffic counts and adjustments were made to OD matrices to obtain more accurate results. This step of the assignment of the OD matrix to roads, the comparison between the assigned volumes and traffic counts, and the adjustment of the OD matrix was repeated several times to obtain an accurate matrix to reflect the traffic condition for each hour corresponding to traffic count data. In this process, traffic count data had to be set as an input to VISUM for each hour at several roads. A second Python script was established to conduct this iterative process automatically. A validation process of the traffic volume was performed after the calibration process. Hourly traffic count data for 600 h of 37 roads were derived from DCC, among which 35 were chosen for calibration and the remaining 2 were reserved for validation. Roads that were calibrated and validated are indicated in Figure 1 with red and cyan colours, respectively.
The Geoff Havers (GEH) statistic was chosen to evaluate the validation. A guidance regulated in the Design Manual for Roads and Bridges (DMRB) in the United Kingdom was chosen as the criteria for the validation [36]. The calculation of GEH and the criteria is summarized in Table 1. Table 1. Validation criteria (Adapted from U.K. DMRB, [36]). In GEH calculation formula, M is the hourly traffic volume from the traffic model and C is the real-world hourly traffic count.

Measure
Criteria Guideline During the assignment process, the traffic volume on each road was determined not only by the OD matrices but also influenced by the relationship between traffic volume and speed, which was represented by a volume-delay function (VDF). A logistic form of the VDF was adopted as it was found to be more accurate when applied to the prediction of speed measurements in this study, as shown in Equation (1): where sat = q c·q max , t cur represents travel time modelled, t 0 represents the free flow travel time, q represents the traffic volume, and q max represents the road capacity. The values of the parameters in Equation (1) are listed below: The comparison between the modelled and recorded traffic average speeds and the result of traffic volume validation are presented in Section 3.1.

Dispersion and Emission Model
The flow chart of the emission and dispersion model is also shown in Figure 2. The traffic emissions were calculated in OSPM based on the predicted traffic flow in streets (vehicles/hour), the traffic speed, and the emission factors at certain speeds for particular types of vehicles (g/vehicle/km). Only the exhaust emissions were considered in this study.
The fleet composition was derived from traffic count data and the Irish national fleet composition. Fleet proportions for each category (e.g., cars, buses, and trucks) were derived from the count data (across 2013) in Dublin city centre. Within each category, the percentages for each vehicle sub-category (e.g., petrol > 2.0 l with PC Euro 3 and Diesel > 2.0 l with PC Euro 4) were assumed to be in proportion with the Irish national fleet composition [37], following the same methodology applied by [25]. Summarized fleet data are presented in Table 2. Further details of the fleet data are given in Table A1 in the Appendix A. A 1% composition of benzene in petrol was assumed, in line with the EU regulation [38]. Information on traffic conditions, which included the traffic volume and speed for each hour of the year and fleet composition information, was developed using a third Python script to gather information from the traffic model, and then formatted and inputted into OSPM.
Besides emissions, other inputs to OSPM that influenced pollutant concentrations included road and building geometry information, weather information and background pollutant concentration. These data were obtained from Google Maps, the Irish meteorological service, and the air quality monitoring network of the Environmental Protection Agency of Ireland, respectively. A summary of these elements is shown in Table 3. In order to facilitate a holistic assessment of traffic emissions, the major harmful pollutants that originate from traffic were included, namely NO X , PM 10 , PM 2.5 , CO, and Benzene. Regarding NO X , a special focus was given to the modelling concentrations of NO 2 (in the Modelling results section), as NO 2 is a major pollutant concerning public health. The modelled concentrations consisted of a background concentration component and a component arising from the local traffic in each road. Whilst the impacts of speed limit and fleet composition changes on the concentrations of NO X , PM 10 , PM 2.5 , CO, and Benzene on each road were estimated, background concentrations were only available for PM 10 and NO X . Therefore, while total concentrations of NO X and PM 10 were modelled, the traffic-induced concentrations of PM 2.5 , CO, and Benzene were modelled.
In OSPM, concentrations of traffic emissions were calculated using a combination of a plume model for the direct contribution of traffic pollution, and a box model for the recirculating part of the pollutants in the street, taking street canyon geometry into account. The receptors (the points of which pollutant concentrations were estimated) were set to be on the building face in the centre of each street section at a height of 1.2 m. These street sections, with lengths of 80 m to 120 m, constituted the streets shown in Figure 1. The reason that the receptor height was set to 1.2 m was because the age of children in the school in the modelling area was in the range of 4 to 11, and the average height for children in this age range is around 1.2 m in Ireland [39]. The receptor height of the street on which the air quality monitoring site was located was set to 2m, the same height as the as the monitor.

Scenarios
The air quality conditions estimated from the traffic counts in 2013 near the school were considered as the baseline scenario. The air quality of hypothetical speed limit changes and fleet composition changes were evaluated as alternative scenarios and compared to the baseline. The baseline scenario and six hypothetical scenarios are summarized in Table 4.  Table 2 with the scenario number corresponding to each scenario number in this table.
At present, diesel cars account for only 3% of total passenger vehicles in the United States and less than 1% in China, whereas this is about 50% in Europe, and it is 45% in Ireland [37,40]. Compared to petrol cars, diesel cars have better fuel economy than petrol-powered cars, thus their emission of CO 2 may be lower. However, diesel remains a major source of harmful pollutants (e.g., ozone-forming gases, including NO X and PM) [40]. Therefore, in line with several recent proposals to reduce the prevalence of this source of air pollution in European cities [41], this paper compares the impact on air quality outside a school in an urban setting of converting diesel cars to comparable petrol cars.
Also, Dublin City Council reduced the speed limit within the modelled area from 50 km/h to 30 km/h in 2009 because of road safety considerations within the city centre. This paper, therefore, also assesses the impact of the speed change by modelling the impact of two hypothetical speed limits on air quality.

Volumes and Speeds
The average modelled traffic volumes and speeds within the modelled area are summarised in Figure 3, showing the average hourly traffic data and average speed for all road segments within the baseline scenario (speed limits of 30 km/h) during different hours of the day. The average modelled speed at AM peak hour (8 to 9 a.m.) and PM peak hour (5 to 6 p.m.) of 12 km/h and 21 km/h was very similar to the measured average traffic speed in the city centre of 13 km/h and 19 km/h [42]. During off-peak hours, the average modelled speed returned towards the relevant speed limit.

Result of Validation for Traffic Volumes
A summary of validation results, using the GEH statistics approach, is presented in Table 5. The validation satisfied the DMRB criteria shown in Table 1, resulting in a robust model which represented the travel demand after being calibrated using the recorded volumes.

Dispersion Model Validation
Studies elsewhere have compared and validated local-scale emission and dispersion models (including OSPM) with in-situ measurements and have found a good fit with this modelling approach, especially for annual average concentrations [35]. Figure 4 shows the results of modelled NOX and PM10 daily average concentrations compared against the observed concentrations in Dublin city centre. The modelled concentrations were acquired from the particular road on the network where the monitoring site was located, and also on the side of that road where the monitor was located.
The modelled concentrations fitted the observed concentrations, with an R 2 = 0.85 and 0.82 for NOX and PM10, respectively. The modelled PM10 concentrations had good accuracy, with a slope of 1.07. However, the modelled NOX data underestimated the measured concentration, where the slope was 0.69. Previous investigations have found similar underestimations of NOX using the OPSM model [43].

Result of Validation for Traffic Volumes
A summary of validation results, using the GEH statistics approach, is presented in Table 5. The validation satisfied the DMRB criteria shown in Table 1, resulting in a robust model which represented the travel demand after being calibrated using the recorded volumes.

Dispersion Model Validation
Studies elsewhere have compared and validated local-scale emission and dispersion models (including OSPM) with in-situ measurements and have found a good fit with this modelling approach, especially for annual average concentrations [35]. Figure 4 shows the results of modelled NO X and PM 10 daily average concentrations compared against the observed concentrations in Dublin city centre. The modelled concentrations were acquired from the particular road on the network where the monitoring site was located, and also on the side of that road where the monitor was located.

The Effect of Speed Limit Changes
The modelled impacts of speed limits on the total concentrations of NO2 and PM10 in each street section in the modelled area are summarised in Figure 5a,b. Each individual data point in Figure 5 represents the concentration on a specific road link for the speed limit change scenarios compared to The modelled concentrations fitted the observed concentrations, with an R 2 = 0.85 and 0.82 for NO X and PM 10 , respectively. The modelled PM 10 concentrations had good accuracy, with a slope of 1.07. However, the modelled NO X data underestimated the measured concentration, where the slope was 0.69. Previous investigations have found similar underestimations of NO X using the OPSM model [43].

The Effect of Speed Limit Changes
The modelled impacts of speed limits on the total concentrations of NO 2 and PM 10 in each street section in the modelled area are summarised in Figure 5a,b. Each individual data point in Figure 5 represents the concentration on a specific road link for the speed limit change scenarios compared to the baseline scenario. The x-axis shows the baseline scenario concentration, and the y-axis shows the concentrations for two speed limit scenarios. A 1:1 relationship between the baseline scenario concentration and the speed limit change scenario is indicated with a grey diagonal line. A light red dot and a dark red dot highlight two speed limit change scenarios for the road on which the school is located. Traffic volumes, and consequently air pollution concentrations, on this street were less than for the majority of road links in the rest of modelled area. The EU air quality standards for annual NO 2 and PM 10 concentrations are also indicated with brown lines. The impact of speed limits on the traffic-induced concentrations of PM 2.5 , CO, and Benzene are summarised in Figure A1a

The Effect of Fleet Composition
The impact of fleet composition on the total concentrations of NO2 and PM10 in each street is shown in Figure 6a,b, whilst the impact of fleet composition on the traffic-induced concentrations of CO and Benzene is shown in Figure 6c,d. Traffic-induced concentrations were estimated instead of total concentration of CO and Benzene because of the absence of background concentration information for these gases. The impact of fleet composition on the traffic induced concentrations of PM2.5 is shown in Figure A2 in the Appendix A.
In Figure 6, the x-axis shows the baseline scenario concentration, and the y-axis shows the concentrations for four fleet composition change scenarios. Again, the grey diagonal line indicates the 1:1 relationship between the baseline scenario concentration and the fleet composition change scenario. Brown lines show the EU standards for annual NO2 and PM10 concentration. Figure 6a,b show declines in NO2 and PM10 concentration for the modelled fleet composition change scenarios. This demonstrates that increasing the proportion of petrol vehicles can reduce the NO2 and PM10 by up to 23% and up to 19%, depending on the street in question. This hypothetical strategy can reduce the traffic-induced NO2 and PM10 by up to 35% and up to 28%, respectively. For similar reasons as outlined above, the general trend for NO2 and PM10 concentrations was that they could be reduced by a greater amount for the streets that originally had higher concentrations.
Among all these scenarios, including speed limit change scenarios, replacing all diesel cars and vans with petrol cars and vans had the greatest predicted impact on NO2 and PM10 concentration. The high original NO2 and PM10 concentration of several streets, which exceeded the EU limit value,  Figure 5 shows a decline in both NO 2 and PM 10 concentration for scenarios with hypothetical increases of speed limits from 30 km/h to 40 km/h and 50 km/h. The alteration of the speed limit from 30 km/h to 50 km/h could reduce NO 2 by up to 18%, depending on the original concentration in the street in the baseline scenario. This action could also reduce PM 10 by up to 15%, also depending on the original concentration. The general trend shows that NO 2 and PM 10 concentrations could be reduced further for the streets that originally had higher NO 2 and PM 10 concentration. Regarding the traffic-induced concentration, alteration of the speed limit from 30 km/h to 50 km/h could reduce these concentrations of NO 2 and PM 10 by 21% to 30% and 14% to 22%, respectively, depending on the street in question. As background concentration for different streets was assumed to be the same, the higher the NO 2 and PM 10 concentration was on the street, the higher the pollutant concentration that was generated by traffic, and therefore, the speed limit change strategy that influences traffic conditions would have a larger effect on the pollutant concentration.
From Figure 5, for both NO 2 and PM 10 , we can see that there were several streets where the original concentration was beyond the EU limit value, and this was predicted to reduce to below the limit as a result of the proposed speed changes. Therefore, the hypothetical speed limit changes modelled here would allow the pollutant concentrations on more streets to adhere to the EU air quality standards.
As the school was located in a place with low traffic volumes and was thus less polluted, the effect of the speed change on air quality was relatively small compared to other streets with heavier pollution. Comparing the speed limit change impact in front of the school with a busier street (yellow dots shown in Figure 5), the predicted concentrations of NO 2 and PM 10 could only be reduced by a maximum of 3% and 2%, respectively, on the street in front of the school. This corresponded to a 22% and 15% reduction of traffic-induced NO 2 and PM 10 . These figures could be reduced by 15% and 12%, respectively, on the street with higher original pollutant concentration (corresponding to 27% and 19% reduction of traffic-induced NO 2 and PM 10 ).

The Effect of Fleet Composition
The impact of fleet composition on the total concentrations of NO 2 and PM 10 in each street is shown in Figure 6a,b, whilst the impact of fleet composition on the traffic-induced concentrations of CO and Benzene is shown in Figure 6c,d. Traffic-induced concentrations were estimated instead of total concentration of CO and Benzene because of the absence of background concentration information for these gases. The impact of fleet composition on the traffic induced concentrations of PM 2.5 is shown in Figure A2 in the Appendix A.   Table 6. The impact of fleet composition changes on the concentration of NO2 and PM10 comparison between the street in front of the school (indicated by red dots in Figure 6) and a street with high original pollutant concentration (indicated by yellow dots in Figure 6).  In Figure 6, the x-axis shows the baseline scenario concentration, and the y-axis shows the concentrations for four fleet composition change scenarios. Again, the grey diagonal line indicates the 1:1 relationship between the baseline scenario concentration and the fleet composition change scenario. Brown lines show the EU standards for annual NO 2 and PM 10 concentration. Figure 6a,b show declines in NO 2 and PM 10 concentration for the modelled fleet composition change scenarios. This demonstrates that increasing the proportion of petrol vehicles can reduce the NO 2 and PM 10 by up to 23% and up to 19%, depending on the street in question. This hypothetical strategy can reduce the traffic-induced NO 2 and PM 10 by up to 35% and up to 28%, respectively. For similar reasons as outlined above, the general trend for NO 2 and PM 10 concentrations was that they could be reduced by a greater amount for the streets that originally had higher concentrations.

Scenarios
Among all these scenarios, including speed limit change scenarios, replacing all diesel cars and vans with petrol cars and vans had the greatest predicted impact on NO 2 and PM 10 concentration. The high original NO 2 and PM 10 concentration of several streets, which exceeded the EU limit value, as shown in Figure 6a,b, could be reduced to below the limit value with the shifting of diesel vehicles to petrol vehicles.
The effect of fleet composition change on the street in front of the school was relatively small compared to other streets with heavier pollution ( Table 6). Regarding the reduction of the concentration of NO 2 and PM 10 on the street in front of school, the scenario of replacing all diesel cars and vans with petrol vehicles was the most efficient, reducing the NO 2 and PM 10 pollution concentrations by 4% and 3%, respectively. When compared to the 3% and 2% reductions found for the scenario of increasing the speed limit to 50 km/h, both interventions produced similar levels of impact. Table 6. The impact of fleet composition changes on the concentration of NO 2 and PM 10 comparison between the street in front of the school (indicated by red dots in Figure 6) and a street with high original pollutant concentration (indicated by yellow dots in Figure 6).

A Street with High Original Concentration
Original concentration NO 2 22 µg/m 3 45 µg/m 3 PM 10 14 µg/m 3 35 µg/m 3 Concentration with 30% diesel changing to petrol cars (change in traffic-induced concentration) Concentration with 60% diesel changing to petrol cars (change in traffic-induced concentration) Concentration with 100% diesel changing to petrol cars (change in traffic-induced concentration) Concentration with 100% diesel changing to petrol cars and vans (change in traffic-induced concentration) Regarding the traffic-induced concentrations of CO and Benzene, Figure 6c,d depict that, in general, increasing the proportion of petrol cars and vans would increase these concentrations appreciably. Thus, in contrast to NO 2 and PM 10 , increasing the proportion of petrol vehicles would exacerbate the CO and Benzene pollution. By changing all the diesel cars and vans to petrol cars and vans, the traffic-induced concentration of CO would be increased by 65% at maximum. This strategy would also increase traffic-induced Benzene concentration by 36% at maximum. The scenario of changing diesel vehicles to petrol vehicles could increase the traffic-induced CO and Benzene concentration at the school by 63% and 35%, respectively.

Discussion
This investigation showed that speed limits and fleet composition changes could have notable impacts on the air quality. These scenarios could potentially affect the air quality at the location of the school in question here, albeit to a lesser extent than they would affect the air quality along the more polluted streets away from the school.
Regarding speed limits, the hypothetical increase of speed limits from 30 km/h to 50 km/h was predicted to reduce the concentrations of all the pollutants examined here (i.e. NO 2 , PM, CO, and Benzene). Ghafghazi & Hatzopoulou [28] found that speed bumps that are used to reduce speeds could lead to an increase in both NO 2 and NO X concentrations in urban areas. Int Panis et al. [44] estimated that a speed limit change from 50 km/h to 30 km/h would increase NO X and PM emissions by 2% to 5% and 3% to 8%, respectively, using the COPERT methodology. Carsten et al. [45] found that PM and NO X would increase by 1% and 0.7% when the number of vehicles conforming to a speed limit of 30 mph was increased. These results are in line with the results of the present study. Some studies have also found a decrease in traffic emissions with speed limit changes from 50 km/h to 30 km/h in Belgium, using microsimulation [44,46]. To date there have been few ex-post studies that can support the conclusions made by the different analyses of speed limit changes, and clear-cut conclusions about the impact of these strategies on air quality do not exist [23].
The results of this study also showed that for the streets that experienced more severe air pollution, the changing of speed limits would have more impact on traffic-related air pollution. For example, EU limit values could be achieved in some streets where the original pollutant concentration exceeded the limit. Regarding the speed limit decrease from 50 km/h to 30 km/h introduced in 2009, this would not have affected the children significantly in terms of the air quality health impacts as the school within this study area is located along a less-busy, less-polluted street. However, this study does also indicate that, for a busy and heavily polluted area, a strategy of decreasing the speed limit could potentially lead to a reduction in air quality, with consequent implications for health.
In assessing the potential risks of vehicle speed to human health, it is necessary that all risks are taken into account. In this respect, it is perhaps worth pointing out that air pollution is responsible for a considerable amount of premature deaths worldwide. In Ireland in 2014, 193 people were killed in road traffic accidents [47], while in the same year, air pollution was associated with 1510 premature deaths [48]. However, transport is only one of many contributors to this figure. Therefore, it does not necessarily suggest that the influence of air pollution on human health outweighs that of traffic accidents. On the other hand, in terms of the relationship of speed limit and car accidents, a review suggests that "studies of the effectiveness of school zone limits have generally found poor driver compliance, particularly when the limits are set very low, and no relationship between pedestrian crashes and the special limits" [49].
In this study, measured traffic count data and speed records were used in the calibration and validation of the model. This enabled the traffic strategies being assessed to be based on the traffic conditions of Dublin in 2013. The emission calculation method used in the study is based on average speeds, which, although complying with the objective and time-scale scope of the study, restricted the evaluation of the impacts of acceleration and deceleration on traffic emission. The driving cycles that are adopted to develop the COPERT emission factors include the condition of driving in real urban roads, which infers that these EFs incorporate the emission related to acceleration and deceleration to some extent. However, changes that may be brought about in the amount of acceleration and deceleration occurring as a result of changes in speed limits are not fully captured using this approach. In order to quantify the instantaneous impacts of speed limit changes on emissions in more detail in future, a traffic micro-simulation would be required. However, such micro-simulations are often limited in their size of study area and length of time scale which can be covered, limiting the results of such investigations in contrast to the current approach. The current approach was based on the average speed in every hour of 2013 for the entirety of Dublin city, and as such was focused on the longer-term impacts of these traffic strategies over a large area.
The health impact of the traffic emissions resulting from reducing speed limits should be taken into consideration, especially for schools located on busy roads. In contrast, outside an urban setting, road transport may contribute significantly less than other sources of air pollution to total exposure, and in such circumstances a speed limit reduction would be especially beneficial.
As petrol and diesel vehicles are still dominant in the fleet, and will be for some time, and there is no alternative that can replace these two types of vehicles worldwide in the short-term, it is worthwhile to evaluate these two types of vehicles in terms of their impact on air quality. Also, the emission change estimation of the fleet conversion to hybrid or electric is worthy of research in the future in line with the development and roll-out of this technology.
For different air pollutants, the effects of the changes of speed limit and the fleet composition were different. Increasing the proportion of petrol vehicles could reduce the on-street concentration of NO 2 and PM, but this strategy would lead to a rise in the concentration of CO and Benzene. Thus, although the most significant modelling result for reducing the pollution of NO 2 and PM was obtained by replacing all diesel cars and vans with petrol cars and vans, when considering the pollution of CO, benzene, and other pollutants that were not covered by our paper, this strategy appears inconclusive, with positive and negative impacts. Policy makers should take as many pollutants as possible into account in order to make a full assessment of the advantages and disadvantages of using petrol and diesel cars. The results of this study highlight that the incentivisation of one fossil fuel type over another has advantages and disadvantages. The negative impacts of dieselisation of the European vehicle fleet are well publicized [50]. However, a similar incentivisation of petrol engines would have resulted in problems of a different nature. Determining which policy is effective in tackling traffic-related health impacts should involve a health impact assessment to quantify the impact of increases and decreases in different pollutants with different levels.
Similar to the effect of changing speed limits, for the streets that experienced more severe air pollution, the strategies of increasing the proportion of petrol cars were more effective in dealing with NO 2 and PM pollution. The improvement in air quality of implementing different traffic management strategies depends on how much the traffic contributes to the air pollution. Therefore, cities with severe traffic pollution will receive larger benefits when applying these traffic air pollution reduction strategies than those found in Dublin, where air quality is relatively good.

Conclusions
This study analysed several strategies and related factors that could affect air pollution in a city centre near a school. Real traffic count data were applied and several vehicle fleet and speed limit scenarios were modelled. The results showed that both strategies can influence the traffic-induced air pollutant concentrations significantly. Changing diesel cars and vans to petrol cars and vans had both advantages and disadvantages; this strategy would lead to benefits in terms of reducing NO 2 and PM pollution, but would increase the pollution from CO and Benzene. Thus, it is overly simplistic to provide incentives for using either petrol or diesel vehicles. Policy makers should seek to strike an appropriate balance of both in the fleet at present. Looking to the future, strategies should be implemented that are aimed at phasing out both types of vehicles and replacing these with e.g. electric vehicles or hybrid vehicles.
Decreasing speed limits near a school, although justified in terms of road safety, was shown to have potentially negative impacts in terms of air quality, especially for streets that are heavily polluted; thus, these impacts should be considered and balanced with the benefits to safety brought about by reducing the speed limit.