On-Road Bus Emission Comparison for Diverse Locations and Fuel Types in Real-World Operation Conditions

: Urban buses have energy and environmental impacts because they are mostly equipped with heavy-duty diesel engines, having higher emission factors and pollution levels. This study proposed a mean distribution deviation (MDD) method to identify bus pollutant emissions including CO, CO 2 , HC, and NO X at road sections, intersections, and bus stops for di ﬀ erent fuel types; and explore the changes in emissions for di ﬀ erent locations in the road sections, bus stops, and intersection inﬂuence areas. Bus speed, acceleration, and emissions data were collected from four fuel types in China. For di ﬀ erent locations and fuel types, the di ﬀ erences in emissions were all statistically signiﬁcant. MDD values for di ﬀ erent locations indicated that there were more obvious di ﬀ erences in emissions between road sections and intersections. In addition, heat maps were applied in this study to better understand changes in bus emissions for di ﬀ erent locations in the bus stop inﬂuence areas, intersection inﬂuence areas, and road sections.


Introduction
Emissions from motor vehicles are a major source of urban air pollutant emissions, such as nitric oxides, carbon monoxide, hydrocarbons, and carbon dioxide [1][2][3][4][5]. For instance, relevant research studies indicated that about 45% of the pollutant emissions in the United States were a direct consequence of vehicle operations [6]. In urban road areas, compared with private motor vehicles, public transportation plays an important role in resident trips and sustainable transportation [7][8][9][10][11]. By optimizing the public transit infrastructure layout and dispatching schemes, the sharing rate of public transit is increasing significantly in many cities. In the U.S., buses have taken around a 20% share of transit market in recent years [12]. As for China, public transport also accounted for around half of daily trips in 2018 in many first-tier cities such as Beijing, Shanghai, and Chongqing [13]. Although emissions and energy consumption per passenger mile of public transportation vehicles are significantly lower than those of other vehicle types [14], exhaust fumes from public transportation are still one of the most relevant environmental problems [15][16][17]. Most urban transit buses are equipped with heavy-duty diesel engines, which have higher emission factors and pollution levels [18][19][20][21]. To alleviate this shortcoming, there is an increasing number of buses that are switching to use new energy power. The U.S. Federal Transit Administration (FTA) has provided several grants for the purchase of new energy buses [22]. China has 340,000 new energy buses, which account for around half of the total number of buses [23]. Because public transportation is being extensively used and

T-Test and Mean Distribution Deviation Methods
To explore the changes in emissions at different locations in the bus stop influence area, intersection influence area, and road section, t-test and mean distribution deviation methods were used: (1) Testing differences in speed, acceleration, and emission values: A t-test was conducted to test the significance of the differences between bus stop influence areas, intersection influence areas, and road sections.
(2) Mean distribution deviation (MDD): This value was calculated by taking the difference in cumulative emission distributions for the two different samples. Higher mean distribution deviation values indicate that there are more obvious differences between the two samples [42]. It can be expressed as follows: where P i,s 1 and P i,s 2 are the proportions of sample s 1 and s 2 at or below the maximum emissions of bin i, respectively. In this study, the absolute value of the difference between the P i,s 1 and P i,s 2 was calculated for all the bins. The MDD values were calculated for each bin and then averaged by dividing the total by the number of bins, i. For this study, i was set to 10.

Data Collection and Analysis
The research team collected the field data for four bus lines in Zhenjiang, China, and line features were displayed in Figure 1. As seen in Figure 1 and Table 1, these lines were similar in terms of traffic volume, number of lanes, degree of roads, length of lines, number of stops, and line features. Such selected routes consist of major transit corridors, which connect the relatively prosperous districts in the city. In addition, they also penetrate into dwelling districts, filling in the gaps of the backbone of urban bus. The data collection was conducted from 11 to 15 April 2016 under good weather conditions. The field data consisted of two parts: (1) bus vehicle characteristic data include fuel type, vehicle speed, and acceleration; and (2) emission situation data include bus pollutant emissions of carbon monoxide (CO), carbon dioxide (CO 2 ), hydrocarbon (HC), and nitric oxides (NO X ). Among them, bus pollutant emissions were measured by an AUTOplus automotive gas emission analyzer unit [43]. The AUTOplus automotive gas emission analyzer unit was employed to measure the second-by-second gaseous exhaust emissions from vehicle exhaust pipes including CO, HC, NO X , and CO 2 . Standard calibration gases were used to verify the accuracy of the system before each individual test, and to set the target pollutants to zero. With the help of an automatic vehicle location system, we also collected bus speed and acceleration values every second.   Table 2 presents the vehicle and emission characteristics for GEHE, CNG, EURO-V, and EURO-IV buses. The speed was calculated by measuring the elapsed time to travel a specific distance (typically approximately 4.5 m). From Table 2, for different fuel types, the performance of pollutant emissions varied. Diesel engine buses complying with EURO-IV and EURO-V emission standards were prone to have higher CO and NO X emissions. On the contrary, CO and NO X levels for compressed natural gas buses were lower, but CO 2 and HC levels were higher. By contrast, gas-electric hybrid electric buses performed best, with lowest emission values. In addition, for four fuel types, there was no obvious difference in speed and acceleration, excluding the potential influences of confounding factors on the emission results.

Results and Discussion
A bus stop influence area includes three sections: bus entry area, alighting and boarding areas for passengers, and exit area. In general, a stopped bus will slow from its running speed about 50 m before the bus stop, and accelerate back to its running speed about 30 m after the stop [44,45]. Thus, in this study, such a range of areas was regarded as the bus stop influence area, as shown in Figure 2a. An intersection influence area was divided into an inside area (D1) and intersection-related area (D2), as shown in Figure 2b. The existing research studies generally designated the road section ranging from a 150 ft (45 m) to a 250 ft (76 m) radius from the center of the intersection as the intersection influence area [46][47][48][49][50]. The bus emissions and time proportions at road sections, intersection, and bus stop influence areas are shown in Figure 3. The figure shows the total emissions of CO, CO 2 , HC, and NO X produced in and around bus stops, which accounted for 24%, 20%, 24%, and 17%, respectively, while intersections shared about 30% of all the emissions during the whole trip. Although the road sections accounted for higher proportions of emissions than other locations (with proportions of 46% for CO, 49% for CO 2 , 45% for HC, and 51% for NO X , respectively), the bus spent much more time on road sections. The findings showed that the emissions at bus stop and intersection influence areas led to a high exposure level of pollution in a short time. In order to show the differences better, the research team added results of t-tests for the differences in speed, acceleration, and emissions between different fuel types of buses. As seen in Table 3, there was no significant difference in emissions between bus stop influence areas and intersection influence areas, with p-values exceeding 0.050 (varying between 0.087 and 0.357). However, between bus stop influence areas and road segments, or between intersection influence areas and road segments, p-values (<0.050) showed that differences were statistically significant. This may be due to the fact that for intersection and bus stop influence areas, vehicles may conduct a complete stop to wait for traffic signals and service boarding and alighting passengers. Thus, the whole process included cruising, deceleration, idling, acceleration, and cruising modes [20]. In addition, results of t-tests (p-values < 0.050) displaying the differences between EURO-IV, EURO-V, CNG, and GEHE buses were all statistically significant. The findings showed significant effects of fuel types on bus emissions. A t-test was utilized to judge if there was a significant difference between different fuel types and locations. In order to further explore the degree of difference, mean distribution deviation (MDD) was used in this study-the higher the values, the more obvious the differences. MDD values for different locations and fuel types are compared in Table 4. As seen in Table 4, the MDD values between intersection influence areas and road segments were higher than the values between bus stop influence areas and intersection influence areas, and bus stop influence areas and road sections. The finding indicated that there were more obvious differences in emissions between intersections and road segments. As for different fuel types, however, the performance of MDD values varied. For CO and CO 2 , the MDD values between GEHE/CNG buses and EURO-IV/EURO-V buses were the highest; and for HC and NO X , the differences between EURO-IV buses and EURO-V buses were the most obvious. In order to better understand changes in bus emissions at the bus stop influence areas, intersection influence areas, and road sections, heat maps were introduced in this study using the measured data of total emissions including CO, CO 2 , HC, and NO X , as shown in Figure 4 (EURO-IV buses), Figure 5 (EURO-V buses), Figure 6 (CNG buses), and Figure 7 (GEHE buses). To generate a continuous surface from these collected discrete samples, Kriging interpolation was introduced to meet the rigorous demands of practical applications [1,51]. Choosing the appropriate variogram is the most important step in implementing Kriging interpolation. The spherical variation was applied here because it is the most widely used function of Kriging interpolation and it also basically meets the diffusion characteristics of gas. As shown in  Figure 7a, the results of the heat maps for bus stop influence areas showed that for different fuel types, the maximum emissions occurred at the bus stops and in the vicinity of them. This is because buses spend a long time conducting a complete stop and serving the alighting and boarding passengers. In addition, the downstream areas generated more emissions than the upstream areas for bus stop influences areas. This may be due to the fact that vehicle emission rates during the acceleration mode are comparatively higher than the deceleration mode. In the acceleration process, the engine needs more fuel to generate enough power to accelerate. The higher the acceleration rates, the more emissions are generated. In the upstream area, using regenerative breaking for the GEHE bus, the energy consumption and pollutant emission can be reduced, but cannot reach zero emissions. For intersection influence areas, as shown in Figure 4b, Figure 5b, Figure 6b, Figure 7b, the upstream areas generated more emissions than the downstream areas. This occurs because buses joining a queue must come to a stop and undergo several stop-and-go cycles until passing the intersection. As a result, emissions increase because of excessive delays and speed change cycles for approaching traffic. In contrast with bus stop and intersection influence areas, there was no obvious difference in emissions at different locations for road sections, as shown in Figure 4c, Figure 5c, Figure 6c, Figure 7c. Figures 8-10 compared the average values and standard deviations of emissions for different fuel type buses in each location. It can be seen from the figure that the emission results at different locations were similar. EURO4 buses exhausted more than other types of buses, which indicated that the emissions of EURO5 buses were greatly improved compared to EURO4 buses in all driving conditions. CNG buses emitted slightly more than EURO4 vehicles. The emissions of GEHE vehicles were significantly smaller than other models, only about half of those of EURO4 and CNG buses, and its emissions fluctuated less than other models.

Conclusions
This study proposed a mean distribution deviation (MDD) method to evaluate bus emission characteristics at bus stop influence areas, intersection influence areas, and road sections for different fuel types. Three issues were addressed: (1) collection of instantaneous data of speed, acceleration, and emissions under real conditions; (2) identification and comparison of the pollutant emissions at different locations; and (3) utilization of heat maps to explore the changes in emissions for bus stop influence areas, intersection influence areas, and road sections, respectively.
For different locations, no significant difference existed between stop influence areas and intersection influence areas, with p-values varying between 0.087 and 0.357. By contrast, the differences between bus stop influence areas and road segments, and between intersection influence areas and road segments were statistically significant (p-values < 0.050). In addition, results of t-tests (p-values < 0.050) displayed that the differences between EURO-IV, EURO-V, CNG, and GEHE buses were all statistically significant. The findings showed significant effects of fuel types on bus emissions. A t-test was utilized to judge if there was a significant difference between different fuel types and locations (qualitative analysis). In order to further explore the degree of difference (quantitative analysis), MDD was used in this study-the higher the values, the more obvious the differences. MDD values for different locations indicated that there were more obvious differences in emissions between intersections and road segments. For fuel type, the performance of MDD values varied. For CO and CO 2 , the MDD values between GEHE/CNG buses and EURO-IV/EURO-V buses were the highest; and for HC and NO X , the differences between EURO-IV buses and EURO-V buses were the most obvious.
Furthermore, using the measured data of total emissions, heat maps were introduced to explore changes in emissions for different locations in bus stop influence areas, intersection influence areas, and road sections. To generate a continuous surface from these collected discrete samples, Kriging interpolation was utilized to meet the rigorous demands of practical applications. For bus stop influence areas, from the heat maps, the maximum emissions occurred at the bus stops and in the vicinity of them. Meanwhile, the downstream areas generated more emissions than the upstream areas. For intersection influence areas, however, this trend did not occur. The upstream areas generated more emissions than the downstream areas, because buses joining a queue must come to a stop and undergo several stop-and-go cycles until passing the intersection. In contrast with bus stop and intersection influence areas, for road sections, there was no obvious difference in emissions at different locations. We also compared the average values and standard deviations of bus emissions for different fuel types in each location. It can be seen from the figure that the emission results at different locations are similar. EURO4 buses exhaust more than other types of buses, which indicate that the emissions of EURO5 buses are greatly improved compared to EURO4 buses in all driving conditions. CNG buses emit slightly more than EURO4 vehicles. The emissions of GEHE vehicles are significantly smaller than other models, only about half of those of EURO4 and CNG buses, and its emissions fluctuate less than other models.
From the results of this study, the emissions at bus stop and intersection influence areas led to a high exposure level of pollution in a short time. Therefore, transportation agencies could consider the implementation of countermeasures to control the delay time and update the bus facilities to improve the traffic operations at bus stops and intersections, such as exclusive bus lanes, off-line (bay-style) stops, and GEHE/CNG buses for key areas. Using exclusive bus lanes as an example, relevant research studies showed that delay time and bus speed have obvious impacts on bus emissions. For heavy pollutant emission areas and more than two lanes one way, utilizing exclusive bus lanes could reduce bus delay times by 20%-25%, and further control bus emissions.
With the proposed method, it is feasible to conduct the bus emission comparison for diverse locations and fuel types in China. As for different countries, especially for other developed countries, the bus-operation characteristics may be different. Thus, the proposed method can be applied in other countries; to adapt the method, transportation scholars or agencies will need to collect the traffic data in those countries to obtain the corresponding characteristics, such as the bus travel time for different locations.
In addition, the primary objective in this study was to conduct the bus emission comparison for different fuel types. Thus, the selected routes were similar in terms of traffic volume, number of lanes, degree of roads, length of lines, number of stops, and line features. In order to exclude the impacts of other confounding factors on bus emissions and better explore the emission differences in fuel types, the negative externalities of other modes of transport on public transport emissions have been treated as a control variable. It should be noted that the negative externality of other modes on public transport may also increase the emissions (such as at intersections before traffic signal or during peak hour). This important influencing factor will be considered when we explore the influence of contributing factors and impact degree on bus emissions in future research.