ing fleet renewal, street closures, reduced demand, reduced internal car trips, and reduced through traffic. Several traffic scenarios were modeled in the traffic ...
Traffic Emissions and Air Quality Near Roads in Dense Urban Neighborhood Using Microscopic Simulation for Evaluating Effects of Vehicle Fleet, Travel Demand, and Road Network Changes Ahsan Alam, Golnaz Ghafghazi, and Marianne Hatzopoulou Emission inventories play a key role in the process of selection of suitable policies that can achieve meaningful reductions in traffic emissions. Numerous models that estimate and predict the amount of pollutants at macroscopic, mesoscopic, and microscopic levels exist (2–4). In response to the advances in traffic microsimulation, the use of microscopic emission models is gaining popularity. Microscale models consider instantaneous vehicle speeds and thus account for second-by-second speed profiles, including acceleration, deceleration, idling, and cruising. Because of the computational challenges associated with microsimulation of traffic and emissions, many of the studies that have been performed have focused on corridor-level analyses and thus have largely ignored impacts on the network beyond the direct vicinity of an intervention (5–7). The main objective of the study described here was to capture the effects of changes in vehicle fleet, road network, and travel demand on a networkwide level. For this purpose, a traffic simulation model developed for a dense urban neighborhood comprising more than 8,000 links was extended to have the capability to simulate emissions on the basis of instantaneous vehicle speeds at the link level. The emissions of GHGs, oxides of nitrogen (NOx), and carbon monoxide (CO) were modeled; and the effects of local and regional policy scenarios on total emissions occurring in the neighborhood, as well as on the spatial distribution of emissions across the network, were evaluated. The emissions resulting from the use of an average speed approach and those obtained by the use of second-by-second vehicle speeds were also explicitly compared. Finally, the levels of nitrogen dioxide (NO2) near roadways were also estimated by use of a dis persion model to assess the effects of changes in emissions on local air quality.
A traffic simulation was developed for a dense neighborhood in the city of Montreal, Quebec, Canada (8,656 links), and was linked with a regional traffic assignment model, which was used to determine the travel demand originating in and destined for the study area. With a version of the U.S. Environmental Protection Agency’s Motor Vehicle Emissions Simulator model fit with local data, traffic emissions for each link were simulated by the use of instantaneous speed profiles. Emissions of greenhouse gases (GHG), oxides of nitrogen (NOx), and carbon monoxide (CO) were modeled under a range of regional and local policies, including fleet renewal, street closures, reduced demand, reduced internal car trips, and reduced through traffic. Several traffic scenarios were modeled in the traffic assignment and simulation models to represent these policies. Because of the high congestion levels in the neighborhood under base case conditions, limited networkwide reductions in emissions were observed, except in the scenario that aimed to reduce through traffic (29% reduction in GHG emissions compared with that in the base case scenario). Significant changes in the spatial patterns of emissions were detected. Average and instantaneous speed-based estimates were also compared; the average speed mode tended to overestimate total emissions as network speeds decreased. Finally, dispersion modeling was conducted along selected corridors to evaluate the effects of different scenarios on air quality. The study found significant increases in air pollution as a result of the street closure scenario and significant decreases with the reduced through traffic scenario.
The impact of transportation on greenhouse gas (GHG) emissions is a topic that needs little introduction. Besides the potential impact on climate change, transportation emissions have been associated with a range of health effects, including respiratory and cardiovascular outcomes as well as premature mortality (1). Worldwide, an increasing portion of the population in metropolitan areas lives, works, or travels along busy streets, therefore raising concerns over air quality and exposure near roadways. As a result, planners and policy makers are under growing pressure both to target the reduction of traffic emissions from a climate change perspective and to improve urban air quality.
Materials and Methods The study area encompassed the Plateau-Mont-Royal Borough (the Plateau) in Montreal, Quebec, Canada, a central neighborhood characterized by a high population density (12,476 individuals per square kilometer) and low levels of private auto use among its residents (34% private auto, 32% public transit, and 34% walking and cycling). However, because of the increasing numbers of suburban commuters trying to avoid congested highways, the borough experiences a high volume of through traffic because of its strategic location between the northern suburbs of the city and the downtown core. This through traffic mostly consists of passenger cars that do not have any origin or destination in the borough; rather, these passenger
Department of Civil Engineering and Applied Mechanics, McGill University, Macdonald Engineering Building, 817 Sherbrooke Street West, Montreal, Quebec H3A 2K6, Canada. Corresponding author: M. Hatzopoulou, marianne.hatzopoulou@ mcgill.ca. Transportation Research Record: Journal of the Transportation Research Board, No. 2427, Transportation Research Board of the National Academies, Washington, D.C., 2014, pp. 83–92. DOI: 10.3141/2427-09 83
84
cars use the neighborhood road network to reach a destination outside the borough. The high traffic volumes and low network speeds during peak periods, coupled with the high density of the buildings in the proximity of busy streets, have led to significant levels of air pollutants near the roadways across the neighborhood. Concerns over GHGs and air pollutant emissions among the members of the local council and borough residents have spurred the development of an integrated tool for analysis of potential interventions by analysis of traffic and emissions. The study methodology encompasses four main elements: (a) traffic simulation, (b) modeling of emissions, (c) modeling of air quality, and (d) analysis of various policies and scenarios. A regional traffic assignment model was first developed and used to derive travel demand inputs into a traffic simulation model developed for the entire neighborhood. By use of second-by-second link-specific speed profiles, instantaneous emissions were estimated and then aggregated at the hourly level. Emissions during the period from 7 to 8 a.m. were used as a base case for comparison of scenarios, although the model was run from 5 to 8 a.m. to allow a warm-up period.
Traffic Simulation A regional traffic assignment model was developed for the Montreal metropolitan area by use of the PTV VISUM platform. This model is based on the origin–destination survey data collected by the Agence Metropolitaine de Transport (8). This survey is conducted every 5 years and is the primary source of information on travel habits in Montreal. The regional traffic assignment model consists of 1,552 traffic analysis zones and 127,217 links. It was run in stochastic user equilibrium mode for the period from 5 to 8 a.m. An hourly path file containing all possible paths between each origin and destination as well as the number of trips associated with each path was extracted after every assignment. For the time period from 7 to 8 a.m., the path file included 250,000 paths. Four matrices that formed the main input into the traffic simulation were then extracted from the path file: (a) trips that start and end outside the borough but pass through the borough, generating what is referred to as through traffic (for these trips, the entry and exit links were recorded); (b) trips generated outside the Plateau and destined to the borough (the entry link was recorded for these trips); (c) trips generated in the borough and destined to an outside zone (the exit link was recorded for these trips); and (d) trips generated in and destined for locations within the Plateau, or intratraffic. A traffic simulation for the Plateau neighborhood was developed by use of the PTV VISSIM platform. Model development involved a large data collection effort. The base network was constructed by use of a combination of orthophotographs, topographic maps, cartographic maps, and field visits. Bike lanes were included in the network, as were crosswalks at locations where pedestrians could conflict with turning vehicles. In addition, the 43 transit bus lines and 361 bus stops were introduced. For each of the 576 intersections, the signal phases of the traffic lights were collected and input into the model. Turning restrictions were gathered through the use of Google Maps Street View and were then confirmed through field visits. Speed limits for arteries, local roads, and school zones were incorporated; and areas with reduced speeds were allocated to every turn to account for changes in speed. In total, the VISSIM network has 8,656 links and 5,987 resolved conflict areas.
Transportation Research Record 2427
VISSIM was operated in the dynamic traffic assignment mode; each link on the periphery acted as an abstract parking lot in which vehicles were generated to enter the network and in which vehicles were removed from the network. VISSIM was run in the multirun mode, with the volume initially being set at 10% of the total, and then the volume was increased by 10% for each of the next nine iterations until it eventually reached the convergence criterion. In the base scenario, convergence was reached after 32 iterations. After the convergence, VISSIM was run one additional time to generate second-by-second speed profiles for every link in the network. The second-by-second speed profile obtained for a link could be considered the speed profile for an average car traveling on the link. The speed profile provides the total number of vehicles as well as the average speed of all vehicles traveling on the link for every second of the traffic simulation. The second-by-second speed profile was extracted from VISSIM and was processed for use in the emissions model.
Modeling of Emissions Instantaneous emissions were estimated by use of the U.S. Environmental Protection Agency’s Motor Vehicle Emission Simulator (MOVES; version 2010b) fitted with data expressing local conditions (9). MOVES was used at the project scale to estimate emissions at the link level. MOVES requires information on the road network, vehicle types, meteorology, and fuel specifications. The road network input includes link length, link grade, hourly traffic volume, traffic composition, and link speed. Inclusion of the latter is the most challenging because it entails the extraction of instantaneous vehicle speeds on every link. When traffic is simulated for 1 h, 3,600 instantaneous speeds would, ideally, be input into MOVES for every link. However, because of the size of the network (more than 8,000 links), the input of these data would be computationally burdensome. For this purpose, the researchers decided to input instantaneous link speeds for 360 s, or four traffic signal light cycles (90-s cycles are predominant in the borough), because it was observed that segment emissions stabilize after 210 s, which means that the fluctuation in driving patterns at each intersection becomes minimal. Other MOVES inputs included fleet composition and vehicle age information, which were obtained from the provincial motor vehicle registry. The vehicle fleet circulating within the borough was assumed to be representative of the vehicle fleet in the Montreal region (average age of 8 years, 69% passenger cars, 31% passen ger trucks). Emissions were estimated for all passenger cars and passenger trucks (which comprise SUVs and minivans), which are gasoline fueled. Because of the absence of heavy trucks in the regional assignment model, only household vehicles were included in the traffic simulation as well as in the emissions estimation. The authors recognize this to be a limitation; however, during the peak period from 7 to 8 a.m., commuter travel is the predominant type of travel within the borough and truck travel is minimal. On the basis of the data collected in 2012, the ratio of trucks to total traffic was 0.06; this proportion is based on traffic counts conducted along main arterials in the Plateau. Emissions were estimated for GHGs, NOx, and CO. Although GHGs are interesting in light of increasing concerns over global warming, NOx and CO are considered the best markers of traffic since their levels are the highest in environments near roadways.
Alam, Ghafghazi, and Hatzopoulou
85
Analysis of Various Scenarios For the evaluation of the effects of the different strategies on the emissions of GHGs, NOx, and CO, five scenarios were simulated: Scenario 1 (fleet renewal). This scenario simulates a regionwide fleet renewal strategy in which 50% of the vehicles between the ages of 15 and 30 years are replaced by vehicles between the ages of 0 and 5 years through incentives programs. Scenario 2 (street closures). This scenario simulates populating with pedestrians a subarea characterized by a significant number of schools. The area is bound by four main arterials (not affected) and includes six local streets that are converted to use by pedestrians only. Scenario 3 (reduced internal traffic). This scenario simulates a 50% reduction in car trips originating in and destined for the borough. These trips are replaced by walking and cycling trips by use of the existing pedestrian and cycling infrastructure. Scenario 4 (reduced demand). This scenario attempts to capture the effect of a reduced number of car trips destined for the borough. In light of recent aggressive parking policies that the local council has adopted, a portion of trips destined for the borough is expected to shift from cars to public transit. A reduction of inbound trips of 50% was assumed. Scenario 5 (reduced through traffic). This scenario simulates a reduction of through traffic by 50%. In the Plateau Borough, 50% of the total traffic is through traffic. A reduction in through traffic can be achieved through congestion pricing.
Analysis of Air Quality of Selected Corridors For further examination of the environmental effects of the proposed transportation scenarios, the levels of NO2 along selected corridors that have experienced a significant increase or decrease in NOx emissions as a result of changes in the spatial distribution of emissions were estimated. Two scenarios that were suspected to lead to significant changes in the spatial distribution of emissions across the network were selected. These included the street closure scenario (Scenario 2) and the scenario leading to a reduction of through traffic (Scenario 5). Figure 1 illustrates the study area, the pedestrianized area, and the three corridors selected for further analysis of air quality. Each corridor consists of two links, and dispersion modeling was conducted for each individual link. In the case of the street closure scenario, the corridors Papineau (the main arterial running north and south of the Plateau) and Christophe Colomb (a smaller road serving local traffic) were selected for dispersion modeling, whereas in the scenario for the reduction of through traffic, Papineau and Fabre (a residential road in a school neighborhood) were also selected. In the first step, NO2 levels were computed for each link along the selected corridors in the base case scenario, and background levels were accounted for. In the second step, the new emissions simulated with MOVES were used to compute NO2 levels after implementation of the scenario. Increases or decreases in NO2 before and after implementation of the scenario were tested for statistical significance. The Operational Street Pollution Model (OSPM), developed in Denmark, was used for dispersion modeling. OSPM calculates the concentrations of exhaust gases from road traffic by using a combination of a plume model and a box model to assess the recirculation of pollutants along the street canyon. It considers the effects of both
FIGURE 1 Study area (red link within pedestrian zone was selected for dispersion modeling under reduction of through traffic scenario).
ambient and traffic turbulence caused by moving vehicles. Three reactions are simulated for modeling NO2 when NOx emissions are provided as input: NO + O3 → NO 2 + O 2
(1)
NO 2 + hv → NO + O
(2)
O + O 2 → O3
(3)
where NO = nitric oxide, NO2 = nitrogen dioxide, O = atomic oxygen, O2 = oxygen, O3 = ozone, and hv = ultraviolet sunlight. OSPM requires data on the street configuration; these data include the average height of buildings and the width and orientation of the street. These data were gathered through field visits for every link simulated. In addition, NOx emissions estimated by the use of MOVES were input as NO and NO2, with the directly emitted NO2/ NOx fraction being 0.09. Hourly meteorological data, such as wind speed, wind direction, global radiation, and temperature for the month of October 2012, were collected from Environment Canada (10). A total of 31 days of meteorological data were used for dispersion modeling; the concentration along each segment before and after each scenario reflects a weighted average of the 31 days of meteorology. This concentration is therefore a monthly average for typical October conditions. Wind speed and direction are the two most critical meteorological factors for dispersion modeling along roadways. An increased wind speed is known to enhance dilution and therefore reduce air pollution levels, whereas the direction of the wind with respect to the roadway causes a differential between air pollution
86
Transportation Research Record 2427
North
Wind Speed >30 km/h 20–30 km/h 10–20 km/h 0–20 km/h 5%
10%
15%
20%
FIGURE 2 Wind speeds and directions characterizing conditions in October.
levels on both sides of the roadway. At the extreme, when the wind is perfectly orthogonal to the road, the leeward side experiences significantly higher levels than the windward side. Figure 2 presents the wind rose that summarizes all wind conditions that were modeled to determine the monthly NO2 average along the segments.
Results and Discussion of Results Traffic and Emission Model Validation The results of the traffic simulation in the Plateau were validated against traffic counts conducted within the same period. Counts from 162 intersections were used to validate the model. R2 values for the period from 6 to 7 a.m. and the period from 7 to 8 a.m. were 0.58 and 0.72, respectively. In the microscopic simulation, a large share of the error was contributed by roads with small volumes. The differences between the observed traffic counts and predicted traffic counts indicate that the accuracy for streets with heavy traffic is better than that for streets with lighter traffic (104% for roads with 0 to 500 vehicles per hour and 24% for roads with more than 1,000 vehicles per hour). By use of the simulated traffic counts, emissions of CO, NOx, and CO2 equivalents (CO2-eq) were estimated in MOVES. Although it is difficult to validate traffic emissions unless a portable emission monitoring system is available, the authors validated selected drive cycles extracted from the traffic simulation output with drive cycles collected via GPSs installed on floating cars driven in the neighborhood. The drive cycles (which consist of second-by-second speeds)
are first extracted for the links of interest and used to calculate the vehicle-specific power (VSP) and operating mode (opmode). Within MOVES, opmode refers to a combination of speed and VSP for each second. VSP represents the tractive power exerted by a vehicle to move itself and its passengers. It is a function of instantaneous speed, acceleration, vehicle weight, and road grade (11). A low opmode indicates a low speed and a low VSP, whereas a high opmode indicates a high speed and VSP. Each opmode is associated with an emission rate (grams per hour) as a function of the VSP and a number of other variables (e.g., vehicle type, model year, meteorology, fuel). On the basis of the onboard GPS data collected, VSP was calculated on a link basis (outside MOVES). In the next step, the amount of time spent in each opmode over the entire link was calculated and an opmode distribution was developed. The link-specific opmode distribution provides the amount of time that the vehicles spend under different opmode categories. Finally, a cumulative opmode distribution was developed over the entire link. This cumulative opmode distribution was compared with the cumulative opmode distribution obtained with data simulated in VISSIM, which were also processed in the same way. Figure 3 illustrates the total emissions calculated for six links: two of them along the Chambord Corridor, two of them along the Fabre Corridor, one along the Garnier Corridor, and one along the Marquette Corridor. All of the selected corridors are present within the Plateau Borough, and GPS data from three floating cars were collected for all of the corridors. Figure 3 presents the total GHG emissions (in CO2-eq) for the six links under three cumulative opmode distributions derived from GPS traces and three VISSIM simulations
Alam, Ghafghazi, and Hatzopoulou
87
700
Total CO2 emissions (g)
600 500 400 300 200 100 0
Simulated 1
Simulated 2
Simulated 3 GPS 1 Speed
GPS 2
GPS 3
FIGURE 3 Total CO 2-eq emissions derived by use of speeds collected by GPS (from three floating cars) and simulated speeds in VISSIM (under three random seeds).
under different random seeds. A slight underestimate was observed when the speeds from the GPS data were compared with the simulated speeds; however, a statistical analysis at the link level showed no statistically significant differences (at the 95% confidence level) between emissions derived from observed and simulated drive cycles. Finally, the effect of various random seeds in the traffic simulation on the resulting total emissions in the network was evaluated. For this purpose, base case emissions of NOx were estimated by use of the methodology presented above after the traffic simulation was run under five random seeds. The total NOx emissions under the five random seeds are illustrated below. The results indicate that the effect of random seed is minimal, as shown by the relatively small standard deviation of the results: • Run 1 = 16,676 g, • Run 2 = 16,949 g, • Run 3 = 16,815 g, • Run 4 = 16,846 g, • Run 5 = 16,883 g, • Mean = 16,834 g, and • Standard deviation = 101.4.
Total Network Emissions In the base case scenario, a total of 12.8 tons of GHG emissions (in CO2-eq) during the 7 to 8 a.m. peak period was estimated. On average, the networkwide emission factor was 437.14 g per vehicle mile traveled (VMT), which indicates relatively low speeds and significant levels of acceleration, deceleration, and idling. Table 1 presents total network-level emissions of CO2-eq, CO, and NOx in the base case as well as in the five scenarios. It also shows average network speeds and hourly traffic volumes for each scenario. The fleet renewal scenario, which assumes that 50% of old vehicles are replaced by new vehicles, does not achieve a noteworthy reduction in GHG emissions but does achieve significant reductions in NOx emissions. These reductions are associated with improvements in emission reduction technologies, such as exhaust gas recirculation systems, that provide improvements in emissions greater than those achieved through improved fuel efficiency. For a better observation of the effect of fleet renewal on GHG emissions, the same fleet renewal scenario was simulated for the year 2025, and sufficient vehicle fleet turnover was allowed. In the 2025 scenario, a 17.72% reduction of GHG emissions was achieved.
TABLE 1 Summary of Networkwide Emissions Under the Simulated Scenarios Statistic, by Scenario Characteristic
Base Case
Fleet Renewal
CO2–eq (kg) CO2–eq (g/VMT) CO (kg) CO (g/VMT) NOx (kg) NOx (g/VMT) VMT Average network speed (mph) Traffic volume (vph)
12,889.13 437.14 240.77 8.17 16.69 0.56 29,485 15.75 488,830
12,867 436.41 215.35 7.30 13.75 0.47 29,485 15.75 488,830
Note: vph = number of vehicles per hour.
Street Closures 12,762.48 449.70 235.96 8.31 16.26 0.57 28,383 15.69 473,501
Reduced Internal Traffic
Reduced Demand
Reduced Through Traffic
12,748.29 436.69 237.49 8.13 16.41 0.56 29,193 15.96 482,505
11,672.96 433.54 219.01 8.13 15.12 0.56 26,925 16.09 447,466
9,163.53 416.43 174.80 7.94 11.96 0.54 22,005 16.82 372,480
88
Although the street closure scenario achieved a minimal reduction in total GHG emissions compared with those in the base case (0.98%), it led to an increase in the networkwide emission factor. This increase was attributed to increased congestion around the street closure area. In this scenario, a decrease in the traffic volume on the network, which tended to decrease total emissions, was observed; however, it was offset by an increasing emission factor because of lower driving speeds. A reduction in the number of trips within the borough achieved small reductions in the emissions of GHGs and other pollutants (approximately 1%), a finding that points to the fact that trips within the borough are by far not a major contributor to network emissions. This is because of their shorter distances and their use of paths that are not necessarily on the main arterials but, rather, that are on the less congested local roads. A reduction in the number of car trips destined for the Plateau achieves a more significant decrease in emissions than the fleet renewal, street closure, and reduced internal traffic scenarios. This decrease was due to a lower overall VMT and a higher average network speed. Finally, a reduction of through traffic achieves the best emissions reductions: a 29% reduction in CO2-eq emissions, a 27% reduction in CO emissions, and a 28% reduction in NOx emissions. This result occurs because through traffic encompasses a significant portion of the traffic in the neighborhood and uses the main north–south arterials, which incur significant levels of congestion.
Spatial Distribution of NOx Emissions Although a comparison of network-level emissions provides insight into the overall effectiveness of different scenarios, examination of the spatial distribution of emissions gains and losses on individual links is important to identify drastic changes that could lead to potentially significant effects on air quality near roadways. The focus here is on NOx emissions, in light of their potential effect on NO2 levels near roadways. Figure 4a presents the hourly link emissions of NOx normalized by link length in the base case scenario. Figure 4a shows (in red) that the main arterials, which are also the most congested links on the network, have emissions that reach up to more than 1,000 g/mi. As expected, the fleet renewal scenario does not alter the spatial distribution of emissions on the network; emissions on all links decrease by 15% to 20%. The change in emissions from the levels in the base case resulting from the street closure scenario is illustrated in Figure 4b. A significant increase in emissions is observed on roads immediately adjacent to the pedestrian zone. Because the street closures affect links directed in the north–south direction, most north–south arterials suffer from an increase in emissions. The reduction of internal traffic and reduced demand for trips destined to the Plateau are not associated with significant spatial trends; emissions on most links are reduced. Finally, the scenario that focuses on reductions in through traffic, the results for which are depicted in Figure 4c, leads to a reduction in emissions on most north–south links, especially on the main arterials linking the northern parts of the city and the downtown (located south of the Plateau). The changes in the levels of pollutants are not the same across the different scenarios involving changes to policy. For instance, in the fleet renewal scenario in the current year, the reduction in GHG emissions is only 0.17%, whereas for NOx it is 17.6%. For the same fleet scenario in the year 2025, however, the reduction in GHG emissions is 17.7%, whereas the reduction in NOx emissions is 82%. Again, a 50% reduction in through traffic yields almost the same amount of
Transportation Research Record 2427
reduction for GHG emissions (28.9%) and NOx emissions (28.3%). The results indicate that the different policies have different potentials to reduce the levels of emissions of different pollutants. It is therefore necessary to target the pollutants to be reduced before various policies are implemented.
Comparison of Instantaneous and Average Speed Emissions In parallel with the simulation of emissions through the use of instantaneous speeds, network emissions were estimated at the link level through the use of only the average speed of the link. Comparative results are presented in Table 2. The results show that the difference between the instantaneous speed- and average speed-based methods is not consistent and varies across scenarios and traffic conditions. When MOVES is operated in the average speed mode, it consistently overestimates emissions, with a larger gap occurring when network speeds are lower. This effect is further illustrated in Figure 5a, which plots the difference in CO2-eq emissions estimated by use of the instantaneous and average speeds for every link on the network. As the average speed of a link increases, the two estimates of emissions get closer, whereas at lower speeds the difference is larger. An examination of the embedded default operating mode distributions within MOVES explains this observation. In fact, at lower average speeds, MOVES overestimates the amount of time that a vehicle spends in idle and at low instantaneous speeds compared with the amount of time observed under the typical local conditions in Montreal. This finding points to the importance of the development of local-level operating mode distributions and replacement of the default distributions within MOVES if it is to be used at an average speed mode. When instantaneous speeds are directly input within MOVES, this problem does not arise, as the model generates opmode distributions from the input data. Finally, Figure 5b illustrates average speed and instantaneous speed emissions at different traffic loadings within the network, further illustrating the effect of network speed on the difference between the two emissions estimates.
Dispersion Modeling Along Selected Corridors Three corridors were selected for analysis of air quality: Christophe Colomb, Fabre, and Papineau. The last corridor is a major arterial road running in the north–south direction of the Plateau, and it experiences a significant amount of traffic during the morning peak period. In the base case scenario, Papineau experiences significantly higher NOx emissions than the other two roads. Figure 6 presents the NOx emissions along two segments of each roadway in the base case; the results are presented as averages for the two traffic directions. Stationary air quality monitoring stations located close to Papineau have recorded an average NOx level of 14.6 parts per billion (ppb) on the basis of data collected in the fall of 2005, whereas OSPM simulated an average of 14.1 ppb (minimum = 10.5 ppb and maximum = 25.7 ppb). Moreover, Figure 6 presents average NO2 levels (in parts per billion) for each segment on the basis of the 31 meteorological scenarios; it also presents the maximum and minimum estimated NO2 concentrations among the 31 simulations. The average NO2 level estimated for each segment is closer to the minimum than to the maximum. This finding is due to the meteorological conditions adopted
Alam, Ghafghazi, and Hatzopoulou
89
0.42–60.00 60.01–120.0 120.1–240.0 240.1–1,395 0
0.25
0.5
1 mile
(a)
–90% to –13% –13% to +12%
–100% to –12% –12% to +12% +12% to +74,435%
+12% to +1,054% 0
0.25
0.5
1 mile
0
(b)
0.25
0.5
1 mile
(c)
FIGURE 4 Spatial distributions of link emissions: (a) amount of NO x g/mi in base case scenario and percentage of change in NO x emissions compared with those in base case in (b) street closure scenario and (c) reduced through traffic scenario.
90
Transportation Research Record 2427
TABLE 2 Comparisons of Average Speed- and Instantaneous Speed-Based Network Emissions Emissions (kg), by Scenario Pollutant, Speed-Based Method
Base Case Emissions (kg)
CO2, instantaneous CO2, average CO, instantaneous CO, average NOx, instantaneous NOx, average
Fleet Renewal
Street Closures
Reduced Internal Traffic
12,889 17,328 240.769 254.353
12,867 17,292 215.348 219.894
12,762 17,145 235.964 250.242
12,748 17,154 237.487 252.245
11,673 15,868 219.011 232.823
9,163 11,929 174.796 179.508
16.684 19.819
13.749 16.159
16.260 19.273
16.410 19.709
15.115 18.143
11.963 14.608
50,000
20,000
40,000
16,000
Reduced Demand
Reduced Through Traffic
Difference between average and instantaneous speed estimates (kg)
CO2-eq (kg)
CO2-eq (g)
30,000 20,000
Average speed emissions (kg) Instantaneous speed emissions (kg)
12,000
8,000
10,000 4,000 0 0
5
10
15
20
25
30
35
-10,000
0 350,000
380,000
410,000
Average link speed (mph)
440,000
470,000
Traffic volume
(a)
(b)
FIGURE 5 Difference between emissions estimated with average and instantaneous speeds: (a) as a function of average link speed and (b) as a function of total traffic volume for network.
30 120 25 100
15
60
0
avg. NO2 min. NO2 max. NO2 NOx
Papineau 2
0 Papineau 1
20
Fabre 2
5
Fabre 1
40
Christophe Colomb 2
10
NOx (g)
80
Christophe Colomb 1
NO2 (ppb)
20
Corridor Segment FIGURE 6 NO x emissions and associated NO 2 concentrations in base case scenario (avg. 5 average; min. 5 minimum; max. 5 maximum).
500,000
Alam, Ghafghazi, and Hatzopoulou
91
for the month of October, in which relatively strong winds that favored mixing were emphasized. In fact, the simulated minima occur much more often than the maxima, thus pushing the average closer to the minimum. Although NOx emissions vary significantly across the different segments (because of different traffic conditions), average NO2 levels are less disperse. The correlation between average NO2 (in parts per billion) and total NOx (in grams) emissions is 0.7. This finding indicates the importance of the inclusion of variables affecting dispersion, such as segment orientation with respect to wind, the height of buildings, and the width of the roadway. Although Papineau experiences significantly higher emissions, the average NO2 levels near the road are similar to the ones estimated for Fabre, which is a local street with lower emissions, although it has a narrower configuration with little or no spacing between buildings. This base case analysis illustrates the importance of dispersion modeling as an additional dimension to the evaluation of changes in traffic patterns because changes in emissions do not necessarily translate into equal changes in air pollution levels. Two scenarios were selected for further exploration through dispersion modeling: the street closure scenario and the scenario involving a reduction of through traffic. In the street closure scenario, two segments on Papineau and two segments on Christophe Colomb were selected for analysis of air quality. In the scenario involving a reduction in through traffic, Papineau and Fabre were selected. In the street closure scenario (Table 3), a statistically significant (at the .01 level) increase in the mean NO2 level from 13.7 to 14.6 ppb along Segment 1 of Papineau and from 14.5 to 15.8 ppb along Segment 2 of Papineau is observed. Deterioration in the maximum NO2 levels experienced is also observed, with the maximum being 28.6 ppb. Although these levels remain below ambient air quality standards, in reviewing a recent epidemiological study conducted in Montreal, the authors observed an increase in the risk of breast cancer of 31% with a 5-ppb increase in ambient NO2 levels (which are already below the standards) (12). The second corridor, Christophe Colomb, also experienced an increase in the mean NO2 level. The increase along Christophe Colomb is smaller than that observed along Papineau (a busier street), although the difference is statistically significant (at the .01 level). This finding indicates that although the street closure scenario contributes to a small decrease in networkwide emissions, it also leads to an increase in emissions around the area populated with pedestrians, TABLE 3 NO x Emissions and NO 2 Concentrations Along Selected Roads in Base Case and Street Closure Scenarios Value, by Christophe Colomb Segment
Value, by Papineau Segment
TABLE 4 NO x Emissions and NO 2 Concentrations Along Selected Roads in Base Case and Reduction of Through Traffic Scenarios Value, by Papineau Segment
Value, by Fabre Segment
Scenario
Variable
1
2
1
2
Base case
VKT NOx EF (g/VKT) Total NOx (g) Avg. NO2 (ppb) Max. NO2 (ppb) VKT NOx EF (g/VKT) Total NOx (g) Avg. NO2 (ppb) Max. NO2 (ppb)
360 0.34 124.8 13.7 23.0 161 0.33 53.5 11.5 20.5
148 0.42 62.3 14.5 25.7 77 0.48 37.5 13.5 25.2
48 0.37 17.9 12.7 27.5 28 0.30 8.6 11.9 22.2
109 0.39 42.4 12.5 22.0 63 0.33 20.9 11.9 19.1
50% less through traffic
and thus leads to a deterioration of air quality. The deterioration is stronger along the larger corridor. In the reduction of through traffic scenario (Table 4), the mean NO2 levels along both corridors experience a statistically significant decrease (at the .01 level). The decrease is more important along Papineau (≥1 ppb), whereas it is smaller along Fabre (