Vehicular PM Emissions and Urban Public Health Sustainability: A Probabilistic Analysis for Dhaka City

Particulate Matter (PM) pollution is generally considered as a prime indicator of urban air quality and is linked to human health hazards. As vehicles are a vital component of an urban setting, the risks of particulate pollution need to be assessed. An emission modelling is essential for that, and thus stochastic modelling approach involving Monte Carlo simulation technique was applied, aiming to reduce the uncertainty in emission modelling. The risks scenarios for the emissions were generated for 2019 (present state) and 2024 (future), integrating the probability of emissions and the associated AQI (Air Quality Index). Despite the vehicles being a minor source of PM in Dhaka (compared to the contribution from other sources), about one-third of the city is found under high risk due to the exhaust particulate pollution; having the potentiality to cover more than 60% of the city in the coming years, affecting the urban public health sustainability. However, the extent of implementation of planning and management strategies can revert the scenarios for the city, which can plausibly reduce the risk from 80% to 50%, or even to a no-risk state.


Introduction
Dhaka City of Bangladesh is one of the most densely populated cities in the world (ranked within the top 15), with a population density of approximately 28,000 per sq. km as per 2016 [1]. At the same time, in terms of particulate pollution, it has been identified continuously as one of the most polluted cities [2]. Therefore, extensive population exposure to that particulate pollution is apparent, risking the overall urban public health standings of the city. Particulate Matter (PM), both coarse (PM 10 -particles diameter ≤ 10 micrometres) and fine (PM 2.5 -particles diameter ≤ 2.5 micrometres) particles, are mainly responsible for respiratory health problems [3][4][5][6][7]. The typical effects of PM exposure include asthma, bronchitis, reduced lung function, cardiovascular disease, and other chronic effects associated with premature mortality [8][9][10][11][12][13][14] when exposed for a longer period. Exposure to higher ambient PM 2.5 is considered to be one of the major causes of premature deaths (about 7.6% globally), of which 59% is in the east and south Asia for the population being at risk [15]. Concerning the mortality associated with the PM exposure, an approximately 2-5% increase in mortality might result due to 1 µg/m 3 increase in PM 2.5 concentration [16,17]. PM is also found to be associated with high blood pressure to a great extent, especially among the elderly [14,18]. Epidemiological studies conducted in US have demonstrated a consistent association between short-term PM 2.5 exposure and blood pressure level among the elderly [19]. A higher level of exposure to PM 2.5 emissions from diesel engines was also identified as a cause of lung cancer [19], which might be sourced from diesel vehicles in the city. While the sustainability demands to protect the future from over-exploitation at present [20], BRT) might improve the congestion scenario in future, but that still needs to be assessed concerning the exposure of huge inhabitants to the resulting pollution. There are research that demonstrates how the changed lifestyle changes the pollution scenarios, for instance, the impact of reduced movement during COVID-19 pandemic reduced PM emissions by approximately 28% in Malaysia [69], about 75% in Morocco [70], and about 21% in China [71]. However, these reductions are not solely from reduced vehicular emissions, somewhat reduced industrial emissions also had a part. Again, concerning the PM emissions from vehicles, research indicates that improved vehicle technology and higher speed might reduce PM emissions by 50% [47,52,72], considering other factors being involved in such reductions. Therefore, a modelling study that considers such factors can be useful to predict the contribution of vehicular emissions and what extent that needs to be managed. However, there has been not much research explicitly conducted on accounting the particulate pollution from vehicular sources in Dhaka City, especially the parametric study that can analyse the 'what-if' scenarios. Therefore, Dhaka City was selected for this research to assess the sustainability of both the typical and strategic scenarios, which might be useful in representing the scenarios for developing countries as well.
Sustainability 2020, 12, x FOR PEER REVIEW 3 of 18 but that still needs to be assessed concerning the exposure of huge inhabitants to the resulting pollution. There are research that demonstrates how the changed lifestyle changes the pollution scenarios, for instance, the impact of reduced movement during COVID-19 pandemic reduced PM emissions by approximately 28% in Malaysia [69], about 75% in Morocco [70], and about 21% in China [71]. However, these reductions are not solely from reduced vehicular emissions, somewhat reduced industrial emissions also had a part. Again, concerning the PM emissions from vehicles, research indicates that improved vehicle technology and higher speed might reduce PM emissions by 50% [47,52,72], considering other factors being involved in such reductions. Therefore, a modelling study that considers such factors can be useful to predict the contribution of vehicular emissions and what extent that needs to be managed. However, there has been not much research explicitly conducted on accounting the particulate pollution from vehicular sources in Dhaka City, especially the parametric study that can analyse the 'what-if' scenarios. Therefore, Dhaka City was selected for this research to assess the sustainability of both the typical and strategic scenarios, which might be useful in representing the scenarios for developing countries as well. Emissions of different pollutants are attributed to various processes. Likewise, PM emissions from vehicle exhaust depend mainly on the fuels (gasoline and diesel mostly) [47,51,56,[72][73][74] and engine technology [47,52,64,75], primarily when an engine is operated in a low-temperature condition [8,35]. Although the particulate matter is categorized in PM2.5 or PM10, however, this research considers the emission of PM2.5, as the coarse fraction, present in an insignificant amount in vehicle exhausts [64]. Other than direct combustion emissions, loose materials on the road surface might result in indirect pollution associated with vehicle movement [28,51,53,[76][77][78][79], which was not considered in this research for emissions modelling.
While a monitoring network of the city is capable of showcasing an overall PM pollution scenario, a modelling study can address the responsible factors (e.g., traffic volume, engine technology, fuel types) for vehicular emissions only, and thus can facilitate further planning for attaining sustainability. The number of samples representing a scenario is always a crucial factor, which is often compromised in the modelling studies due to a lack of data. Therefore, adopting a stochastic modelling approach (be generating a series of random scenarios) to obtain average emissions, along with analysing the probability of getting that average emission, might improve the performance of the model. The intensity of the emissions concentrations and its associated probability of occurrence can signify the risks associated with the pollution, and thus can be ranked accordingly. The spatial analysis of risk severity thus was demonstrated in this research denoting the sustainability scenarios.
The emissions modelling, along with other factors, relies on vehicle attributes and their operation patterns at the road links [44]. To reduce the uncertainty of these parameters induced from limited surveys, implementing Monte Carlo simulation technique [80,81] would randomly generate Emissions of different pollutants are attributed to various processes. Likewise, PM emissions from vehicle exhaust depend mainly on the fuels (gasoline and diesel mostly) [47,51,56,[72][73][74] and engine technology [47,52,64,75], primarily when an engine is operated in a low-temperature condition [8,35]. Although the particulate matter is categorized in PM 2.5 or PM 10 , however, this research considers the emission of PM 2.5 , as the coarse fraction, present in an insignificant amount in vehicle exhausts [64]. Other than direct combustion emissions, loose materials on the road surface might result in indirect pollution associated with vehicle movement [28,51,53,[76][77][78][79], which was not considered in this research for emissions modelling.
While a monitoring network of the city is capable of showcasing an overall PM pollution scenario, a modelling study can address the responsible factors (e.g., traffic volume, engine technology, fuel types) for vehicular emissions only, and thus can facilitate further planning for attaining sustainability. The number of samples representing a scenario is always a crucial factor, which is often compromised in the modelling studies due to a lack of data. Therefore, adopting a stochastic modelling approach (be generating a series of random scenarios) to obtain average emissions, along with analysing the probability of getting that average emission, might improve the performance of the model. The intensity of the emissions concentrations and its associated probability of occurrence can signify the risks associated with the pollution, and thus can be ranked accordingly. The spatial analysis of risk severity thus was demonstrated in this research denoting the sustainability scenarios.
The emissions modelling, along with other factors, relies on vehicle attributes and their operation patterns at the road links [44]. To reduce the uncertainty of these parameters induced from limited surveys, implementing Monte Carlo simulation technique [80,81] would randomly generate thousands of data considering the range provided by the field survey (fleet characteristics/volume and speed profile for both peak and off-peak periods). The same research group conducted similar research earlier, focusing on the probabilistic risk assessment for vehicular NOx emissions [82] and could demonstrate the risk scenarios associated with that. Therefore, this research adopted the same methodology for PM 2.5 , aiming to assess the probabilistic air quality risks resulting from vehicular sources, which in turn might trade-off the urban health sustainability of the city.

Methodology
The risks assessment was conducted following the 3 major steps: (a) conducting a probabilistic PM 2.5 emissions inventory that generates emission concentration and the probability of occurrence, (b) risks analysis considering emission intensity and probability, and (c) spatial analysis to demonstrate the scenarios and hot zones.
Dhaka, the capital city of Bangladesh, has a quite widespread road network with several major road links supporting mobility ( Figure 2). As the city is extended more in north-south directions, the north-south-north flow of vehicles along the arterial links supporting the commercial hotspots at the north, south, and middle (shaded areas in Figure 2) parts, would be significant for pollution levels in the city. The flow and characteristics of vehicles differ significantly on the functionality of the zones, and thus the pollution concentration would vary spatially. thousands of data considering the range provided by the field survey (fleet characteristics/volume and speed profile for both peak and off-peak periods). The same research group conducted similar research earlier, focusing on the probabilistic risk assessment for vehicular NOx emissions [82] and could demonstrate the risk scenarios associated with that. Therefore, this research adopted the same methodology for PM2.5, aiming to assess the probabilistic air quality risks resulting from vehicular sources, which in turn might trade-off the urban health sustainability of the city.

Methodology
The risks assessment was conducted following the 3 major steps: (a) conducting a probabilistic PM2.5 emissions inventory that generates emission concentration and the probability of occurrence, (b) risks analysis considering emission intensity and probability, and (c) spatial analysis to demonstrate the scenarios and hot zones.
Dhaka, the capital city of Bangladesh, has a quite widespread road network with several major road links supporting mobility ( Figure 2). As the city is extended more in north-south directions, the north-south-north flow of vehicles along the arterial links supporting the commercial hotspots at the north, south, and middle (shaded areas in Figure 2) parts, would be significant for pollution levels in the city. The flow and characteristics of vehicles differ significantly on the functionality of the zones, and thus the pollution concentration would vary spatially.

Probabilistic PM2.5 Emissions Assessment
The standard vehicular PM2.5 emissions calculation methodology provided by the European Environment Agency (EEA) [64] was adopted in this research, which was also implemented on Dhaka City earlier by Iqbal et al. [44]. The standard method is provided in equation 1 below.
Emissions, E ® PM2.5 (g/km/day) = ∑ where, r = road link; i = vehicle types (e.g., private vehicles, bus, truck, motorcycles, etc.); j = fuel types (petrol, diesel, Compressed Natural Gas-CNG); AADT = average annual daily traffic at a road link r for a vehicle type i with fuel type j; and EFij = PM2.5 emissions factor for each vehicle category (g/km). While the emission factor depends predominantly on vehicle technology and fuel being used in the vehicles, the fuel consumption rate by different vehicles is thus important to scale the emission rate. The Tire-2 PM2.5 emissions factors (g/km) provided by EEA [64] was scaled along with the Tire-3 fuel consumption (g/km) factor for each vehicle category, and fuel consumption-based emissions

Probabilistic PM 2.5 Emissions Assessment
The standard vehicular PM 2.5 emissions calculation methodology provided by the European Environment Agency (EEA) [64] was adopted in this research, which was also implemented on Dhaka City earlier by Iqbal et al. [44]. The standard method is provided in Equation (1) below.
Emissions, E (r) where, r = road link; i = vehicle types (e.g., private vehicles, bus, truck, motorcycles, etc.); j = fuel types (petrol, diesel, Compressed Natural Gas-CNG); AADT = average annual daily traffic at a road link r for a vehicle type i with fuel type j; and EF ij = PM 2.5 emissions factor for each vehicle category (g/km). While the emission factor depends predominantly on vehicle technology and fuel being used in the vehicles, the fuel consumption rate by different vehicles is thus important to scale the emission rate. The Tire-2 PM 2.5 emissions factors (g/km) provided by EEA [64] was scaled along with the Tire-3 fuel consumption (g/km) factor for each vehicle category, and fuel consumption-based emissions factor was calculated [44], as is shown in Table 1. Therefore, the scaled emission factor could accommodate the variations in fuel consumptions depending on the vehicle speeds, and thus PM 2.5 emissions for each road link were calculated. The vehicle specific fuel consumption rate for Dhaka City was acquired from earlier research carried out by Iqbal et al. [82].

Modelling Data
The total registered traffic data for each year was collected from the Bangladesh Road Transport Authority [61], which predicted the vehicle-specific growth patterns for the city (Figure 2). In case of future scenario analysis, the growth trend was considered to predict the vehicle numbers for the scenario.
The distribution of vehicles along the road networks determine the spatial distribution of emissions generated by the vehicles. The dispersion of traffic volume at the major arterial road links, accounted in Iqbal et al. [44], was extrapolated in this research according to the changed vehicle numbers for analysing the emissions. However, plausibly, the limited number of samples collected during the traffic survey might lead to an uncertain estimation, which was controlled in this research using the Monte Carlo simulation technique (a stochastic modelling approach). The stochastic modelling tested by Iqbal et al. [82] for the same traffic dataset determined that the results were steady for more than 10,000 iterations and, thus, in this research, 20,000 random traffic volume were generated within the range obtained in the survey or extrapolated for a year. The simulation was conducted separately for peak (generally 9 am to 11 am, 1 pm to 2.30 pm, and 5 pm to 8 pm) and off-peak periods (other times) [44] valuing the diurnal variation of the traffic volume. The mid-day peak traffic period (1 to 2.30 pm) is induced due to the school traffic (private vehicles); while the morning peak is a bit late than usual in Dhaka city as most of the offices, businesses, and tertiary educational institutions generally starts operation at 10 am [44].
Likewise, the vehicle speed profile was also varied randomly for stochastic modelling, considering the base speed profile data provided by Iqbal et al. [44]. The vehicle speed profile for any subsequent scenarios was adjusted based on the relationship developed by Iqbal and Allan [83], as stated in Equation (2).
where v = estimated vehicle speed (km/h); Ce = car equivalent of vehicles running on the road established by Iqbal and Allan for Dhaka City [83].

Estimating On-Road PM 2.5 Concentrations
The US EPA [84] recommended model for line source dispersion of inert pollutants in urban areas, CALINE4 [85] was used in this research for estimating PM 2.5 concentrations at each road links (data nodes portrayed in Figure 2): The required meteorological data [82] were obtained from the Bangladesh Bureau of Meteorology [86]. CALINE4 was simulated for the dry season of the year (November-March), as the rate of deposition is lower during this period [32]. Thus, having a higher concentration scenario would allow planning for the worst-case scenarios. The meteorological data are provided in Figure 3, showing an east-west movement of the wind (Figure 3b) stimulating the east-west dispersion of pollutants. The model was run for December assuming to be applicable for dry months. The average concentration of peak and off-peak periods was estimated separately by running the model for 1 h for each period. The 1-h average is considered representative as the stochastic modelling generated the average for each period by considering the highest and lowest values. The model was run without considering any background PM 2.5 concentration so that it can predict the concentration resulted from vehicular sources only. The US EPA [84] recommended model for line source dispersion of inert pollutants in urban areas, CALINE4 [85] was used in this research for estimating PM2.5 concentrations at each road links (data nodes portrayed in Figure 2): o The required meteorological data [82] were obtained from the Bangladesh Bureau of Meteorology [86]. CALINE4 was simulated for the dry season of the year (November-March), as the rate of deposition is lower during this period [32]. Thus, having a higher concentration scenario would allow planning for the worst-case scenarios. The meteorological data are provided in Figure 3, showing an east-west movement of the wind (Figure 3b) stimulating the east-west dispersion of pollutants. The model was run for December assuming to be applicable for dry months. o The average concentration of peak and off-peak periods was estimated separately by running the model for 1 h for each period. The 1-h average is considered representative as the stochastic modelling generated the average for each period by considering the highest and lowest values.  [86]). o The model was run without considering any background PM2.5 concentration so that it can predict the concentration resulted from vehicular sources only.

Probability of Emissions/Risks
To conceive the probability function, it is important to know the distribution of data. The traffic volume data on different road links were used to assess the normality of the data, which demonstrated a normal distribution pattern for many cases, while the skewed distribution was also evident ( Figure 4).

Probability of Emissions/Risks
To conceive the probability function, it is important to know the distribution of data. The traffic volume data on different road links were used to assess the normality of the data, which demonstrated a normal distribution pattern for many cases, while the skewed distribution was also evident (Figure 4).  Despite not having concrete normality of data, the normal distribution function was conceived for the probabilistic analysis of PM2.5 emissions for simplicity (equation 3). At the same time, the risk probability was assessed concerning the probability of emissions at the standard level (equation 4). As a summary of the whole process, a flow diagram to evaluate the probability of exceeding threshold emissions is shown in Figure 5.

Risk Analysis
The intensity of risk is associated with the severity of the risk and the corresponding probability of obtaining that severity: Despite not having concrete normality of data, the normal distribution function was conceived for the probabilistic analysis of PM 2.5 emissions for simplicity (Equation (3)). At the same time, the risk probability was assessed concerning the probability of emissions at the standard level (Equation (4)). As a summary of the whole process, a flow diagram to evaluate the probability of exceeding threshold emissions is shown in Figure 5.
where x = PM 2.5 emissions; µ = mean and σ = standard deviation of PM 2.5 emissions.  Despite not having concrete normality of data, the normal distribution function was conceived for the probabilistic analysis of PM2.5 emissions for simplicity (equation 3). At the same time, the risk probability was assessed concerning the probability of emissions at the standard level (equation 4). As a summary of the whole process, a flow diagram to evaluate the probability of exceeding threshold emissions is shown in Figure 5.
In addition, Figure 5. A schematic diagram for analysing the probabilistic risk of PM2.5 emissions above the threshold level.

Risk Analysis
The intensity of risk is associated with the severity of the risk and the corresponding probability of obtaining that severity:

Risk Analysis
The intensity of risk is associated with the severity of the risk and the corresponding probability of obtaining that severity:

•
The severity was classified in this research considering the magnitude of emissions compared to the threshold (standard) level, which is 65 µg/m 3 (hourly average) for PM 2.5 as per Bangladesh standard [58]. • The probability of exceeding the threshold level was classified based on the percentage of probability. A general rule of scaling the probability practised in project management [87,88] was adopted for this research, which was in line with the previous research conducted on NOx probability analysis [82]; viz., probability ≥ 80% = Strong (scale 5), 60-80% = High (scale 4), 40-60% = Medium (scale 3), 10-40% = Low (scale 2), and < 10% = improbable (scale 1).

•
The risk index matrix was generated (Table 2) as a product of severity and probability of risks and thus was categorized into six indices, viz., very extreme, extreme, acute, major, moderate and minor [82].

Spatial Distribution of Emissions/Risks
The emission concentrations and the severity of the risks associated with PM 2.5 emissions from vehicular sources were assessed through modelling for distributed representative nodes of the city, as shown in Figure 2. Depending on the values on the scattered nodes, the spatial distribution of concentration or risk severity was assessed by using interpolation analysis (IDW) of ArcMap (ArcGIS 10.5).

Emissions and Their Probability
As outlined in methodology, the normal distribution and the cumulative distribution of PM 2.5 emissions for each road link (46 road links) was obtained; the result for 2019 is portrayed in Figure 6. The distribution result shows that in the majority of cases higher probability would be around mean emissions, as is indicated by the peakedness of the curves (Figure 6a). As per the cumulative distribution of the emissions, a significant number of the roads (around 35%) would be susceptible to a higher probability of exceeding the threshold level (Figure 6b).

Risk Severity from Particulate Pollution
The risks of being exposed to an unhealthy air quality depend on the concentration prevailing, and therefore the PM2.5 concentration contributed from the vehicular sources was estimated. By interpolating the concentrations estimated at 46 points (locations) in the city, a spatial distribution map of PM2.5 concentrations (1-hr average) for the city was generated for both peak and off-peak periods (Figure 7). The magnitudes of concentrations were categorized according to the concentrations corresponding to AQI levels as per Bangladesh national standard (Figure 7). Though the spatial analysis provides an approximate estimation of concentration, nevertheless, it can demonstrate the air quality hotspots of the city. While considering only the emissions from vehicles (Figure 7), inhabitants of around 10% area in Dhaka City during peak period become

Risk Severity from Particulate Pollution
The risks of being exposed to an unhealthy air quality depend on the concentration prevailing, and therefore the PM 2.5 concentration contributed from the vehicular sources was estimated. By interpolating the concentrations estimated at 46 points (locations) in the city, a spatial distribution map of PM 2.5 concentrations (1-hr average) for the city was generated for both peak and off-peak periods (Figure 7). The magnitudes of concentrations were categorized according to the concentrations corresponding to AQI levels as per Bangladesh national standard (Figure 7).

Risk Severity from Particulate Pollution
The risks of being exposed to an unhealthy air quality depend on the concentration prevailing, and therefore the PM2.5 concentration contributed from the vehicular sources was estimated. By interpolating the concentrations estimated at 46 points (locations) in the city, a spatial distribution map of PM2.5 concentrations (1-hr average) for the city was generated for both peak and off-peak periods (Figure 7). The magnitudes of concentrations were categorized according to the concentrations corresponding to AQI levels as per Bangladesh national standard (Figure 7). Though the spatial analysis provides an approximate estimation of concentration, nevertheless, it can demonstrate the air quality hotspots of the city. While considering only the emissions from vehicles (Figure 7), inhabitants of around 10% area in Dhaka City during peak period become Though the spatial analysis provides an approximate estimation of concentration, nevertheless, it can demonstrate the air quality hotspots of the city. While considering only the emissions from vehicles ( Figure 7), inhabitants of around 10% area in Dhaka City during peak period become susceptible to a PM 2.5 concentration that is unhealthy; for around 40% of areas, the exposure level is found unhealthy for sensitive groups. The exposure map for the off-peak period is indicative of cleaner air; however, about 50% of city areas is found approaching the threshold level.
The pollution concentration level corresponds to an AQI value, which considers threshold emission levels as per the standards. AirVisuals [2] provides continuous AQI levels, based on US EPA standards, for different cities. To assess the validity of the estimations, it was important to compare the modelling result with the available AQI results provided by AirVisuals. Therefore, a total PM 2.5 concentration was estimated, considering that the vehicular sources are responsible for 14% of total concentration.
The assumption was made based on the source apportionment study conducted for Dhaka City by Begum and Hopke [27]. Figure 8 shows a schematic diagram of the comparison, which shows a close match between modelling and monitoring data, although the modelling slightly overestimated the AQI value (248 for modelling vs. 221 for monitoring). Considering the factors that might affect the modelling result, for example, the relative contribution of vehicles to overall pollution, and by considering the proximity or modelling and monitoring results, the model was considered acceptable for further analysis. susceptible to a PM2.5 concentration that is unhealthy; for around 40% of areas, the exposure level is found unhealthy for sensitive groups. The exposure map for the off-peak period is indicative of cleaner air; however, about 50% of city areas is found approaching the threshold level. The pollution concentration level corresponds to an AQI value, which considers threshold emission levels as per the standards. AirVisuals [2] provides continuous AQI levels, based on US EPA standards, for different cities. To assess the validity of the estimations, it was important to compare the modelling result with the available AQI results provided by AirVisuals. Therefore, a total PM2.5 concentration was estimated, considering that the vehicular sources are responsible for 14% of total concentration. The assumption was made based on the source apportionment study conducted for Dhaka City by Begum and Hopke [27]. Figure 8 shows a schematic diagram of the comparison, which shows a close match between modelling and monitoring data, although the modelling slightly overestimated the AQI value (248 for modelling vs. 221 for monitoring). Considering the factors that might affect the modelling result, for example, the relative contribution of vehicles to overall pollution, and by considering the proximity or modelling and monitoring results, the model was considered acceptable for further analysis. However, the Bangladesh national standard for PM2.5 is different from the WHO standard or the US EPA standard applied for globally demonstrated AQI. The threshold values for the same AQI level are higher as per Bangladesh standard for PM2.5 than the US EPA standards (Figure 8) (65 μg/m 3 vs. 35 μg/m 3 for AQI = 100). As the research was conducted for Dhaka, hence the Bangladesh national standard for PM2.5 was considered for risk severity assessment of the estimated concentration. The probability of estimated concentration was considered along with the concentration magnitude to calculate the risk intensity (as per the risk index matrix demonstrated in Section 2.3) for 2019 ( Figure  9) and subsequent scenarios. However, the Bangladesh national standard for PM 2.5 is different from the WHO standard or the US EPA standard applied for globally demonstrated AQI. The threshold values for the same AQI level are higher as per Bangladesh standard for PM 2.5 than the US EPA standards (Figure 8) (65 µg/m 3 vs. 35 µg/m 3 for AQI = 100). As the research was conducted for Dhaka, hence the Bangladesh national standard for PM 2.5 was considered for risk severity assessment of the estimated concentration. The probability of estimated concentration was considered along with the concentration magnitude to calculate the risk intensity (as per the risk index matrix demonstrated in Section 2.3) for 2019 ( Figure 9) and subsequent scenarios. Complementing the exposure magnitude mapped in Figure 7, a significant portion of Dhaka City (about 35% area) is found to be in the acute risk zone during peak time. While the majority of the areas remaining within the safe limit during peak time, the off-peak period has not exhibited any risk induced from the vehicular PM2.5 emissions. This is due to the lower emissions level during offpeak period supported by low traffic congestion, and the lower probability of exceeding the threshold emission level during this time. In general, the data points accommodating higher commercial vehicles alongside higher congestion exhibited higher risks. Thus, the travellers/residents in the congested commercial roads/zones are susceptible to a considerable level of health risk only for vehicular emissions.
Inadequate vehicle standard policies (viz, allowing old and unfit vehicles to operate) that support a higher level of emissions from vehicles, and the unplanned growth of vehicles stimulating traffic congestion on roads, would increase the risks over the years. Considering the prevailing growth rate of vehicles supporting the existing level of congestion and the emissions standard (a 'business-as-usual' case), a risk scenario was assessed for 2024 ( Figure 10). Although there might be reasons that can increase or decrease the magnitude of the risks, those were not considered for the scenario. Complementing the exposure magnitude mapped in Figure 7, a significant portion of Dhaka City (about 35% area) is found to be in the acute risk zone during peak time. While the majority of the areas remaining within the safe limit during peak time, the off-peak period has not exhibited any risk induced from the vehicular PM 2.5 emissions. This is due to the lower emissions level during off-peak period supported by low traffic congestion, and the lower probability of exceeding the threshold emission level during this time. In general, the data points accommodating higher commercial vehicles alongside higher congestion exhibited higher risks. Thus, the travellers/residents in the congested commercial roads/zones are susceptible to a considerable level of health risk only for vehicular emissions.
Inadequate vehicle standard policies (viz, allowing old and unfit vehicles to operate) that support a higher level of emissions from vehicles, and the unplanned growth of vehicles stimulating traffic congestion on roads, would increase the risks over the years. Considering the prevailing growth rate of vehicles supporting the existing level of congestion and the emissions standard (a 'business-as-usual' case), a risk scenario was assessed for 2024 ( Figure 10). Although there might be reasons that can increase or decrease the magnitude of the risks, those were not considered for the scenario.
As the emissions are expected to increase in a 'business-as-usual' 2024 scenario, the risk projection shows that the extremely risky zones would increase to about 40% compared to a current 15%, while the inhabitants of about two-thirds of the city would undergo major health effects during peak time ( Figure 10). For the off-peak period, a sign of moderate risks would appear. The relatively safer zones in the peak period are correlated to areas where vehicles with cleaner fuel operate mainly or the areas with lower congestion. This scenario analysis reveals the fact again that, unless adopting appropriate strategies to control emission rates, the vehicular sources would be capable solely for hazardous air quality in the city.
As Bangladesh authority has planned for improving the vehicle emissions standard from 2019 (euro 1 to euro 3, and euro 2 to euro 4; details is provided in Table 1) [89], the 2024 'business-as-usual' scenario was modified as per the new vehicle emissions standard. The new emission standard is applicable for vehicles registered from 2019 onwards, thus a portion of vehicles would still run on old emissions standard in 2024, which was also incorporated into the scenario (Figure 11). The ratio of vehicles distributed among standards was adopted from Iqbal et al. [82], anticipating that around 50% of the commercial vehicles (more pollution generating category, Table 1) would fall under new vehicle emission standards. The growth rate of CNG vehicles would not affect the PM 2.5 pollution scenario (as CNG vehicles have zero PM emission); however, the gasoline and diesel vehicles would be important for such a case. It is acknowledged that the scenarios assessed here are imaginary and might differ from the actual cases; however, this is capable of predicting a closer risk scenario for such a case.
vehicles alongside higher congestion exhibited higher risks. Thus, the travellers/residents in the congested commercial roads/zones are susceptible to a considerable level of health risk only for vehicular emissions.
Inadequate vehicle standard policies (viz, allowing old and unfit vehicles to operate) that support a higher level of emissions from vehicles, and the unplanned growth of vehicles stimulating traffic congestion on roads, would increase the risks over the years. Considering the prevailing growth rate of vehicles supporting the existing level of congestion and the emissions standard (a 'business-as-usual' case), a risk scenario was assessed for 2024 ( Figure 10). Although there might be reasons that can increase or decrease the magnitude of the risks, those were not considered for the scenario. Figure 10. The risk intensity map of PM 2.5 emissions from vehicular sources for peak and off-peak periods for the 2024 scenario: "business-as-usual". Figure 10. The risk intensity map of PM2.5 emissions from vehicular sources for peak and off-peak periods for the 2024 scenario: "business-as-usual".
As the emissions are expected to increase in a 'business-as-usual' 2024 scenario, the risk projection shows that the extremely risky zones would increase to about 40% compared to a current 15%, while the inhabitants of about two-thirds of the city would undergo major health effects during peak time ( Figure 10). For the off-peak period, a sign of moderate risks would appear. The relatively safer zones in the peak period are correlated to areas where vehicles with cleaner fuel operate mainly or the areas with lower congestion. This scenario analysis reveals the fact again that, unless adopting appropriate strategies to control emission rates, the vehicular sources would be capable solely for hazardous air quality in the city.
As Bangladesh authority has planned for improving the vehicle emissions standard from 2019 (euro 1 to euro 3, and euro 2 to euro 4; details is provided in Table 1) [89], the 2024 'business-as-usual' scenario was modified as per the new vehicle emissions standard. The new emission standard is applicable for vehicles registered from 2019 onwards, thus a portion of vehicles would still run on old emissions standard in 2024, which was also incorporated into the scenario (Figure 11). The ratio of vehicles distributed among standards was adopted from Iqbal et al. [82], anticipating that around 50% of the commercial vehicles (more pollution generating category, Table 1) would fall under new vehicle emission standards. The growth rate of CNG vehicles would not affect the PM2.5 pollution scenario (as CNG vehicles have zero PM emission); however, the gasoline and diesel vehicles would be important for such a case. It is acknowledged that the scenarios assessed here are imaginary and might differ from the actual cases; however, this is capable of predicting a closer risk scenario for such a case.
The risk assessment for the modified 2024 'business-as-usual' scenario reveals that the risk intensity would reduce in such a case. The extremely risky zones would reduce significantly (from the 2024 'business-as-usual' case) though will not be eliminated, however, moderate risk intensity would dominate the city during the peak periods ( Figure 11a). The off-peak period risks would be insignificant considering the result obtained in 2024 'business-as-usual' scenario and thus was not portrayed here. This scenario analysis reveals the fact that applying the vehicle emission standard comprehensively might be influential to bring the risk level down, but it might not be enough for eliminating the risk.  The risk assessment for the modified 2024 'business-as-usual' scenario reveals that the risk intensity would reduce in such a case. The extremely risky zones would reduce significantly (from the 2024 'business-as-usual' case) though will not be eliminated, however, moderate risk intensity would dominate the city during the peak periods (Figure 11a). The off-peak period risks would be insignificant considering the result obtained in 2024 'business-as-usual' scenario and thus was not portrayed here. This scenario analysis reveals the fact that applying the vehicle emission standard comprehensively might be influential to bring the risk level down, but it might not be enough for eliminating the risk.
To accommodate the growing demand for mobility and to enhance the performances of public transports in Dhaka City, new MRT and BRT corridors are underway, expected to be operational from 2024. Thus, this research intended to assess the probable scenario for that time, anticipating modal shift of travellers from private vehicles to public transport, especially for travelling to major commercial areas. To assess the scenario, the 2024 modified 'business-as-usual' scenario was enhanced further by reducing the expected private vehicle count by 50% for travelling to work. This would, in turn, increase the vehicle average speed, complementing the relationship provided in eq 2 ( Figure 11b). The rationale of this hypothetical scenario is assessed previously by Iqbal et al. [66] and was also implemented by Iqbal et al. in NOx scenario assessment [82]. While acknowledging that the scenario might be very gross, it might provide a guideline for future management approaches through this 'what-if' scenario in achieving a sustainable urban health standing.
The risks scenario revealed through this hypothetical scenario predicts no risks induced due to vehicular PM 2.5 emissions (Figure 11b), highlighting the significance of adopting comprehensive strategies. While the hypothesis for this scenario is ambitious, it does not suppress the fact that such uncompromising strategies might control vehicular sources to become solely capable of unhealthy air quality. Otherwise, incorporating the PM 2.5 concentration contributed by other sources, the air quality would be fatal for the city dwellers.
Supporting the current vehicle growth without banning the unfit old vehicles might lead to a situation in five years (by 2024) when eliminating other sources (like banning conventional brick burning, or relocating the industries out of the city) around the city would not be enough to improve the air quality of Dhaka ( Figure 10). Thus, strict improvement and implementation of the standard for emission sources, along with decentralization of facilities, might be influential approaches that need to be thought of now for the future urban health sustainability of the city.

Conclusions
Particulate pollution (especially the fine fraction) is a global issue and is especially important for developing countries where the emission sources remain massively uncontrolled. While there are existing research works that have assessed the particulate pollution levels along with the contributing sources, this research intended to dig further in terms of the contribution from vehicular sources by modelling what-if scenarios. A probabilistic approach supported by stochastic modelling was incorporated with the conventional emission modelling approach to assess the urban health risks, which is believed to produce results with lesser uncertainty.
The analysis pointed to the fact that the uncontrolled emissions from older vehicles in highly congested roads are susceptible to an extremely risky urban health standing. At current practise, about 35% of the city falls under the acute risk zone (very high probability of unhealth air quality) during the day only for vehicular PM emissions. If the current practise persists, about 60% of the city would fall under acute risk zone in five years (by 2024). The risk projection shows that more than 40% of the city would fall under the extreme-risk zone (strong probability of very unhealthy air quality) in 2024, compared to current 15%. If the improved vehicle emission standard (as planned by the government) is implemented now, it would be able to revert the risk but still 30% of the city would fall under major-risk zone (high probability of unhealthy air for sensitive groups). An integrated approach of better traffic management to reduce the congestion along with the emission standard might revert the risk completely, and the city has the potentiality to adopt that considering the future public transport planning to support the mobility. Although it is difficult to implement aggressive policies in the city considering the continuous socio-economic progression, more aggressive strategies at a later stage would be more socio-economically unacceptable. However, the risks mentioned above is only resulting from the vehicular sources, and if the other sources prevail as it is, the air quality would be fatal if appropriate measures are not taken now. This is been indicated by the current trend of air quality in Dhaka, positioning the city among the toppers of the most polluted cities. Although the researchers estimated the air quality based on vehicular sources only, it indicated a critical urban health environment prevailing in Dhaka City. This creates the opportunity for further research, where the probabilistic contribution of all the sources can be incorporated, and thus the management approaches can be illustrated for each source separately.

Conflicts of Interest:
The authors declare no conflicts of interest.