Traffic and emissions impact of congestion charging in the central Beijing urban area: A simulation analysis

Traffic and emissions impact of congestion charging in the central Beijing urban area: A simulation analysis

Transportation Research Part D 51 (2017) 203–215 Contents lists available at ScienceDirect Transportation Research Part D journal homepage: www.else...

3MB Sizes 0 Downloads 61 Views

Transportation Research Part D 51 (2017) 203–215

Contents lists available at ScienceDirect

Transportation Research Part D journal homepage: www.elsevier.com/locate/trd

Traffic and emissions impact of congestion charging in the central Beijing urban area: A simulation analysis Kehan Wu a, Yanyan Chen a,⇑, Jianming Ma b, Song Bai c, Xiru Tang d a

Beijing Key Laboratory of Traffic Engineering, Beijing University of Technology, Beijing 100124, China Texas Department of Transportation, 9500 N. Lake Creek Pkwy, Austin, TX 78701-2483, USA c Sonoma Technology, Inc., 1455 N. McDowell Blvd, Suite D, Petaluma, CA 94954, USA d Beijing Research Center of Urban Systems Engineering, Beijing 100035, China b

a r t i c l e

i n f o

Article history:

Keywords: Congestion charging Emissions impact Traffic impact Public transit service Mode choice Macroscopic simulation

a b s t r a c t Congestion charging is being considered as a potential measure to address the issue of substantially increased traffic congestion and vehicle emissions in Beijing. This study assessed the impact of congestion charging on traffic and emissions in Beijing using macroscopic traffic simulation and vehicle emissions calculation. Multiple testing scenarios were developed with assumptions in different charging zone sizes, public transit service levels and charging methods. Our analysis results showed that congestion charging in Beijing may increase public transit use by approximately 13%, potentially reduce CO and HC emissions by 60–70%, and reduce NOx emissions by 35–45% within the charging zone. However, congestion charging may also result in increased travel activities and emissions outside of the charging zone and a slight increase in emissions for the entire urban area. The size of charging zone, charging method, and charging rate are key factors that directly influence the impact of congestion charging; improved public transit service needs to be considered as a complementary approach with congestion charging. This study is used by Beijing Transportation Environment and Energy Center (BTEC) as reference to support the development of Beijing’s congestion charging policy and regulation. Ó 2016 Published by Elsevier Ltd.

1. Introduction Congestion charging is a typical travel demand management approach used to reduce personal vehicle travel in congested areas. Many large cities, such as Singapore, London, and Stockholm (Beevers and Carslaw, 2005; Eliasson et al., 2009; Tuan Seik, 2000) are using this approach to address urban traffic congestion issues. In recent years, Beijing has developed serious traffic congestion. For example, in 2014 the average speeds on Beijing’s urban roadway network during the morning and evening rush hours were approximately 28 km/h and 25 km/h, according to the Beijing Transportation Information Center (BTIC); these speeds are much lower than the roadway free-flow speeds. Beijing has adopted intelligent traffic signal control, traffic restrictions, a license-plate lottery system and many other tools in order to mitigate traffic congestion. However, the effects of these tools are still limited. Recently, the Beijing Municipal Commission of Transport proposed development of a congestion charging policy in the central urban area to further address traffic congestion issues.

⇑ Corresponding author. E-mail address: [email protected] (Y. Chen). http://dx.doi.org/10.1016/j.trd.2016.06.005 1361-9209/Ó 2016 Published by Elsevier Ltd.

204

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

In the meantime, urban air pollution is another challenging issue in Beijing. According to China’s Ministry of Environmental Protection, there were as many as 110 days with moderate or severe air pollution in Beijing during 2014. Motor vehicles are currently one of the largest air pollution sources in Beijing, accounting for more than 30% of the total PM2.5 emissions (Beijing Municipal Environmental Protection Bureau, 2014); traffic congestion and vehicle emissions are closely related. Therefore, to evaluate the impact of a congestion charging policy, both traffic and vehicle emissions changes need to be considered. Congestion charging policies have been studied in depth since the concept was proposed by Pigou (1932). Some studies focused on the economic perspective by employing margin cost analysis (Vickrey, 1969; Walters, 1961) and cost-benefit analysis (Eliasson, 2009; Raux et al., 2012). The traffic impact of congestion charging is another typical research focus. The effects on demand management and relief of congestion in the charging zone have been demonstrated convincingly in real-world examples such as Singapore (Tuan Seik, 2000), London (Sabounchi et al., 2014) and Stockholm (Eliasson et al., 2009). Some of these studies have suggested the importance of public transit service in congestion charging. Jansson (2008) indicated that improved public transport, which reduces private car use, is a necessary condition for successful implementation of congestion charging. Kottenhoff and Freij (2009) indicated that public transport might increase the acceptability and feasibility of a congestion charging policy in Stockholm. With an increased emphasis on air pollution and vehicle emission and their proven association with traffic congestion (Barth and Boriboonsomsin, 2008; De Vlieger et al., 2000; Johansson-Stenman, 2006; Zhang et al., 2011), some studies have focused on assessing the emissions impact of a congestion charging policy. Johansson (1997) indicated that a road-user should pay a toll not only according to one’s own emissions, but also corresponding to the increased emission and fuel consumption of other road-users. Eliasson (2008) evaluated environmental effects of the pilot test of congestion charging in Stockholm, using vehicle emissions modeling and origin-destination (OD) matrix estimation, indicating that the emissions reduction were greatest in the inner city (between 10% and 15%), and that carbon dioxide emissions were reduced 2–3% across Stockholm County. Beevers and Carslaw (2005) indicated that emissions reduction was significantly associated with an increase in vehicle speed and a decrease in vehicle kilometers traveled (VKT), resulting from congestion charging in London, using a traffic emissions model; the study noted that total nitrogen oxide emissions have also increased by about 1.5% on the inner ring road (a ring road forms the boundary of the charging zone). Percoco (2013) evaluated the emissions impact of road charging policies in Milan with a regression discontinuity design, showing that the charges significantly decreased concentrations of carbon monoxide and particulate matters, but only in the short term, due to motorbikes not being charged. Daniel and Bekka (2000) simulated the effects of congestion charging on emissions for a metropolitan highway network in Delaware with traffic simulation and an emissions model, predicting that vehicle emissions would decrease as much as 10% on a citywide scale and 30% in highly congested areas. Mitchell et al. (2005) indicated that depending upon the charging rate, the charges would reduce traffic emissions up to about 70%, and illustrated a diminishing marginal return in emissions reduction as the charge increased. Most research studies reviewed above indicated that congestion charging has a significant effect on the emissions reduction in the charging zone, and that traffic simulation and vehicle emissions models were typical methods used to assess the impact of congestion charging. Although some studies indicated that emissions outside the charging zone increased slightly during congestion charging, the studies on this phenomenon were relatively insufficient based on their areas of focus or available data. As such, the impact on emissions outside the charging zone, and even within the entire city, deserves more attention. Besides, few studies have looked at how the impact on traffic and emissions varies with different charging strategies, even though this is very useful in policy formulation. The purpose of this study is to (1) assess the impact of congestion charging on traffic and emissions in the charging zone, outside of the charging zone, and the Beijing urban area as a whole and (2) compare potential impacts of various charging strategies to support development of a reasonable congestion charging policy. 2. Methodology We analyzed the traffic and emissions impact of a congestion charging policy in the central urban area of Beijing using a combination of macroscopic traffic simulation and a vehicle emissions model, considering the impact of the size of the charging zone, level of public transit service (PT LOS) and charging methods. As shown in Fig. 1, we first defined two modes of travel, private cars and public transit, in a macroscopic traffic simulation platform, with feedback between traffic assignment and mode choice. Then we converted charging fees to time impedance using a monetary value of time concept, and established the congestion charging model for Beijing’s central urban area. Next, we analyzed multiple scenarios with different charging zones, PT LOS, charging methods and charging rates, obtained the characteristic parameters of charging and traffic under different scenarios, and completed an emissions calculation of carbon monoxide (CO), hydrocarbon (HC) and nitrogen oxide (NOX). Finally, we analyzed the emissions impact of the congestion charging in Beijing by adopting multi-factor analysis of variance (MANOVA), and discussed the development of the congestion charging policy with the consideration of emissions reduction. 2.1. Study area and charging scenarios Beijing’s roadways can be classified into four categories: five ring roads (the 2nd to 6th ring road); nine radial toll expressways that connect the central Beijing urban area and suburban areas; eleven China National Highways that depart from

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

205

Fig. 1. Overview of methodology for assessing congestion charing impacts.

Beijing in all directions; and thousands of local roads and streets; the 5th ring road divides the urban and suburban areas. In order to analyze the impact across the city, we set the urban area surrounded by the 5th ring road as the study area (Fig. 2). The total area within the 5th ring road is 665 km2 and the total permanent population was about 10.5 million in 2014. Traffic congestion within the area during rush hours is currently fairly common. According to BTIC, the average speed during morning and evening peak hours in 2014 were only about 28 km per hour and 25 km per hour, respectively. We first analyzed the impact factors of a congestion charging policy and then designed the scenarios. Charging zone, charging method and charging rate are important factors in developing a congestion charging policy. In addition, since public transit is the major alternative mode of private cars, PT LOS is also an important factor affecting the implementation results of the policy. If a specified target of congestion relief is established, then a charging rate developed from a combination of charging zone, charging method and PT LOS is fairly specific. Therefore, we selected charging zone, charging method and PT LOS as the basic factors for designing scenarios, determining the charging rate according to the expected target of congestion relief.

Fig. 2. Study area of the main urban area in Beijing.

206

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

We evaluated two potential charging methods: a cordon-based method, which considers whether a vehicle enters the charging zone, and a distance-based method, which considers vehicle travel distance within the charging zone. Cordonbased charging is currently widely adopted by cities such as Singapore, Stockholm and Milan. It is easier to implement than distance-based charging, because it is not necessary to track how long a vehicle travels. However, distance-based charging is more precise and reasonable, and could be widely realized with the development of vehicle tracking or monitoring technology. We selected the closed central urban areas based on the Traffic Performance Index (TPI, Beijing Transportation Research Center, 2011) to define the charging zone. TPI is widely used in Beijing for describing degree of traffic congestion; its calculation method is shown in Tables 1 and 2 as well as Eqs. (1) and (2).

Pnetwork ¼

n X pi  VKT i

,

i¼1

n X VKT i

ð1Þ

i¼1

pi ¼ Di;sev erely congested =Di

ð2Þ

where Pnetwork is the network congestion proportion (%), pi is the congestion proportion of road type i (%), VKTi is the VKT the road type i (km), Di, severely congested is the length of severely congested segments of road type i (km), and Di is the total length of road type i (km). In 2014, the annual average TPI during rush hours in the area bounded by the 2nd ring road and the annular area bounded by the 2nd and 3rd ring road reached 8.1 and 7.0, respectively, according to BTIC. Traffic congestion within the 2nd and 3rd ring roads is the most serious in Beijing; therefore, we selected these areas as the charging zones, the area inside the 2nd and 3rd ring road is about 63 km2 and 159 km2. In terms of PT LOS setting, Beijing’s public transit system is composed of ground public transit and rail transit, and the average speed of public transit during rush hours was 21.7 km/h, according to BTIC in 2014. Since Beijing is increasing input in both policy and funding to improve public transit service (such as exclusive bus lanes in the urban expressways), the average speed of 21.7 km/h was assumed as the base operating speed of public transit, and a 3-level PT LOS was set with an incremental interval of 5 km/h. Finally, a total of 13 modeling scenarios were established, including a base scenario (no congestion charging) and 12 charging scenarios that cover 2 types of charging zones, 3 levels of PT LOS, and 2 types of charging methods (Table 3). 2.2. Simulation of traffic and emissions The macroscopic traffic simulation was developed using a four-step travel demand modeling approach, focusing on private vehicle and public transit modes. In this study, trip generation and trip distribution are assumed to be uncorrelated with congestion charging. With the congestion charging policy, the charging fee is considered as increased impedance for private vehicle use in the charging zone and has direct impact on mode choice and trip assignment. In terms of mode and route choices under congestion charging, private vehicle use in the charging zone are assumed to have the following potential changes: (1) retain existing route with payment of charging fees; (2) switch mode to public transit to avoid paying the charging fee; (3) switch to a new route outside of the charging zone; and (4) adjust the existing route to reduce travel distance and charging fee payment. These options are reflected in the macroscopic traffic simulation framework with a Logit model to estimate how travel activities may change within and outside the charging zone. We established the traffic simulation network in PTV VISUM (PTV, 2014), which was also used for providing travel activity inputs for emissions estimation. We simulated the main Beijing urban area bounded by the 5th ring road and took transit traffic into account. The OD data used in traffic simulation, obtained from BTIC, was collected during the evening rush hours (5:00 PM–7:00 PM) in 2014. The simulation network included 32,912 roadway links (including major and minor arterials, radial expressways, ring roads and national highways) and 1911 traffic analysis zones (TAZ) corresponding to the OD data. We developed skim matrixes of travel time for three levels of public transit service. The skim matrix of PT LOS 3, which contains the travel time between each TAZ pair (directed), was calibrated using the Beijing Transit IC Card data obtained from BTIC. The IC Card data contain travel information for ground public transit and rail transit, with approximately 5 million records on an average weekday. Travel time was adjusted to reflect improvement of the public transit speed for simulating an increase of PT LOS with an incremental interval of 5 km/h. A Logit model was used for the mode choice modeling, which was calibrated using the stated preference (SP) survey result of the 4th Beijing Resident Travel Survey (Beijing Transportation Research Center, 2012a,b), the total sample size of the survey is about 6500 people. Five traffic modes were included: self-driving car, ground bus transit, rail transit, taxi and non-motorized travel. We combined self-driving car and taxi

Table 1 Conversion relations between roadway segment TPI and average speed (km/h). Road type

Free flow (TPI 0–2)

Reasonably free flow (TPI 2–4)

Slightly congested (TPI 4–6)

Moderately congested (TPI 6–8)

Severely congested (TPI 8–10)

Urban expressway Arterial road Collector and branch road

v > 65 v > 40 v > 35

50 < v 6 65 30 < v 6 40 25 < v 6 35

35 < v 6 50 20 < v 6 30 15 < v 6 25

20 < v 6 35 15 < v 6 20 10 < v 6 15

v 6 20 v 6 15 v 6 10

207

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215 Table 2 Conversion relations between network congestion proportion and network TPI. Parameters

Value ranges

Network congestion proportion (Pnetwork) TPInetwork

[0, 5%] [0, 2]

(5%, 8%] (2, 4]

(8%, 11] (4, 6]

(11%, 14%] (6, 8]

(14%, 24%] (8, 10]

(24%, 100%] 10

Table 3 Description of scenarios. Scenario

Charging zone

Level of public transit service

Charging method

1 2 3 4 5 6 7 8 9 10 11 12 13

No congestion charging Area bounded by the 2nd ring road Area bounded by the 2nd ring road Area bounded by the 2nd ring road Area bounded by the 2nd ring road Area bounded by the 2nd ring road Area bounded by the 2nd ring road Area bounded by the 3rd ring road Area bounded by the 3rd ring road Area bounded by the 3rd ring road Area bounded by the 3rd ring road Area bounded by the 3rd ring road Area bounded by the 3rd ring road

Level Level Level Level Level Level Level Level Level Level Level Level Level

N/A Distance-based Distance-based Distance-based Cordon-based Cordon-based Cordon-based Distance-based Distance-based Distance-based Cordon-based Cordon-based Cordon-based

3 3 2 1 3 2 1 3 2 1 3 2 1

(avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg. (avg.

speed speed speed speed speed speed speed speed speed speed speed speed speed

21.7 km/h) 21.7 km/h) 26.7 km/h) 31.7 km/h) 21.7 km/h) 26.7 km/h) 31.7 km/h) 21.7 km/h) 26.7 km/h) 31.7 km/h) 21.7 km/h) 26.7 km/h) 31.7 km/h)

modes as a single private cars mode, and combined ground bus transit and rail transit as the overall public transit mode. The main influence factors of the Logit model include travel time, cost, traveler age, education and income, the comfortability and reliability were not directly modeled in the model. The input factors of the Logit model were calibrated as follows: the travel time was calculated by the simulation, the basic cost (not including the congestion charging cost) was calculated by a cost matrix which considered travel distance and traffic modes, the age, education and income were calibrated by different average values distinguished by 78 regions (the data obtained from BTIC in 2014). Due to the large scale of the roadway network, we selected an incremental assignment to simulate private car mode. We established the feedback mechanism between traffic assignment and mode choice. The feedback mechanism was used for correcting the results of mode choice and traffic assignment through an iteration of 25 times or less, and traffic volume was used as the convergence criterion (with a difference of less than 10%). We checked the results of traffic simulation under the base scenario with 17 screen lines on four urban expressways and Chang’an Avenue (Fig. 3). Using the paired t-test to compare the observed volumes (Beijing Transportation Research Center, 2014) and simulation results, we found they were well matched (correlation coefficient 0.879, p-value 0.014). We adopted the Beijing Vehicle Emissions Model (BVEM), developed by Beijing Municipal Environmental Protection Bureau (BMEPB) to calculate vehicle emissions for the simulation scenarios. The model was, originally developed in the 1990s to calculate vehicle emissions in Beijing. BVEM classifies vehicles into five categories corresponding to five tiers of emissions standards (China national standards I–V). BVEM includes vehicle dynamometer testing data, verified against on-board monitor data, and was demonstrated as an appropriate model reflecting emissions from Beijing’s local vehicle fleet and travel activities. Since the model was designed for estimating the vehicle emissions in Beijing locally, so the driving conditions used in the platform test were all optimized and adjusted to be more consistent with the practical conditions in Beijing. In our study, we calculated the vehicle emissions using BVEM at a macro level, linking vehicle emissions with average travel speed for various roadway links (see Eqs. (3) and (4)). Additional input information including vehicle fleet composition and meteorological conditions was also used in emissions modeling. We obtained the vehicle composition data (e.g. vehicle

Fig. 3. Volume check of macroscopic traffic simulation by screen lines.

208

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

Fig. 4. Realization of congestion charging model in macroscopic traffic simulation.

type, vehicle age and fuel type) from Beijing Municipal Transportation Administration Bureau (2013), and the meteorological information was obtained from BMEPB (2013).



a X b X c X VKT i;j;k  F j;k

ð3Þ

i¼1 j¼1 k¼1

VKT i;j;k ¼ Li  Voli;j;k

ð4Þ

where E is the total emission of a pollutant in the area (g), Fj,k is the emission factor (g/km) of speed (j) and vehicle type (k), VKTi,j,k is the VKT (km) of Fj,k emission factor in the i segment, Li is the length (km) of i segment and Voli,j,k is the traffic volume corresponding to the Fj,k emission factor in the i segment. 2.3. Congestion charging simulation We converted the congestion charge into the travel time impedance through the concept of monetary value of time (VOT). To obtain a reasonable VOT, we conducted a stated preference (SP) survey in 12 regions in Beijing’s urban area. The survey results suggested an average VOT of 50 RMB/h in Beijing urban area during peak hours; this VOT is similar to the estimate provided by Beijing Transportation Research Center (average VOT of 54 RMB/h). The survey also showed that VOT of the people lived in the suburban area was about 20% lower than that of the people lived in the central urban area; the VOT in peak hours was about 55% higher than that in off-peak hours. In this study, we used the average VOT in each region (ranging from 43 to 56 RMB/h) to convert the congestion charge into the time impedance. We established a charging model including cordon-based and distance-based charging methods in macroscopic traffic simulation, with charging zones within the 2nd and 3rd ring road. The charging model was achieved by setting additional penalty time on the roadway segments. Additional penalty time was applied in traffic assignment to simulate the impact of congestion charging on path choice and mode choice; in concept, with increased penalty time on one route, vehicle would more likely change their routes to reduce the total time impedance, possibly shifting to using public transit. As shown in Fig. 4, for the cordon-based charging method, we added the penalty time to the entrance segment, which achieved the effect of charging the same rate only to vehicles entering or passing through the charging zone (Eqs. (5) and (6)), and we set seven different charging rates of 1.8–18 RMB (equivalent to 2–20 min). For the distance-based charging method, we added a penalty time on the segments within the charging zone, which was proportional to the length of the segment, and also set seven different charging rates of 0.18–1.8 RMB/km (equivalent to 0.2–2 min/km) (Eqs. (5) and (7)). We then conducted simulations of the 12 charging scenarios 84 times in total.

T assignment ¼ T trav el þ T penalty

ð5Þ

T penalty;area ¼ Parea

ð6Þ

T penalty;distance ¼ Pdistance  Ldistance

ð7Þ

where Tassignment is the travel time through the segment used for traffic assignment, Ttravel is the actual travel time (s), Tpenalty is the additional penalty time (s), Parea is the constant additional penalty time (s) in adopting the cordon-based charging method, regardless of how far the vehicle traveled in the charging zone, Pdistance is the time penalty coefficient (s/km) used for adopting the distance-based charging method, and Ldistance is the length of segment (km). 3. Results In this study, congestion charging is assumed to ease traffic congestion to an expected state (target state), which was used to determine the charging rate (target charging rate). We determined the target state using the roadway network TPI levels (Table 1). Given that the roadway network TPI inside the 3rd ring road during off-peak hours can usually remain at the reasonably free flow level, the reasonably free flow state was set as a target state under the congestion charging policy.

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

209

Since TPI is highly correlated to average network speed, we used average network speed (target speed) as a threshold to determine target charging rates. We set the target speeds as 34 km/h and 37 km/h for areas within the 2nd ring road and between the 2nd and 3rd ring roads, respectively, according to the observed off-peak speed data provided by BTIC. These speeds are approximately 10 km/h higher than the average network speeds in peak hours, and about 15 km/h lower than the average network free-flow speed. For each scenario, we conducted modeling simulations with 10 different charging rates to determine the charging rate associated with the target state, and then extracted characteristic parameters of traffic, charging and emissions within the charging zone, external area (the area outside the charging zone), and main urban area under target states (Fig. 5). We analyzed these relationships and the impact of the congestion charging on traffic, charging and emissions in details in the following analysis. 3.1. Analysis of traffic impact As shown in Fig. 6, compared to the base scenario, the changes in the five traffic parameters were different after implementing congestion charging. The public transit shift ratio (4–13%) is strongly associated with the PT LOS, and the shift ratios of adopting the larger charging zone bounded by the 3rd ring road (Sc 8–13) were about 1.3 times as much as that of adopting the smaller charging zone bounded by the 2nd ring road (Sc 2–7). The impact on average speed within the external area (4 to 2 km/h) was greater than that within the main urban area (1.0 to 0.5 km/h), and the latter exceeded the base scenario only with PT LOS 1 (Sc 7, 10, and 13). The target state was set to represent a substantial improvement of traffic congestion, the congestion charging therefore resulted in a significant reduction in VKT in the charging zone (approximately 35–50%). Moreover, regardless of which PT LOS or charging method was adopted within the same charging zone (Sc 2–7 or Sc 8–13), the total VKT within the charging zone were very similar. Even though the PT LOS was very important to attract the private car travelers to change their traveling mode, but restricted by the improvement limit of the PT LOS, according to the mode-choice model, there were always some private car travelers insist on traveling by cars. The reduction of VKT in the large charging zone was about 16% more than that in the small charging zone. However, the impact on VKT in the external area (up to a 20% increase) and the main urban area (up to a 12% reduction) differed greatly; a larger charging zone tends to reduce VKT in the external area and the main urban area more than in a smaller charge zone (Sc 2–7 and Sc 8–13). In addition, trip generation is another aspect of traffic impact. For example, although Sc 3 and 6 (differing in charging method) suggested only about a 4% difference of VKT in the charging zone, the impact on private car trip generation (origin) in TAZs showed a greater difference (up to about 10%, Fig. 7). This indicates that cordon-based charging reduced private car trips generated outside the charging zone to a greater degree, the distance-based charging was more effective at reducing trip generation within the charging zone. 3.2. Analysis of charging impact We analyzed the charging impact from two perspectives: (a) the impact of charging rate on traffic performance, emissions and total charge revenue and (b) the impact of charging zones, transit LOS and charging methods on charging rate and total charge revenue under the target state.

Fig. 5. Major simulation results (e.g. Sc 2): changes of CO, HC and NOX, average speeds within charging zone and main urban area and the target charging rate.

210

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

Fig. 6. Public transit shift ratio, VKT in the charging zone, external area and main urban area; average speed in external area and main urban area. Note: Sc 1 is used as the base scenario for two types of charging zones, so it has different VKT and average speed. VKT in the main urban area was the total of the VKT in the charging zone and the external area. Average speed in a charging zone was set as the same target speed.

Fig. 7. Difference of private car trip generation (origin) in TAZs between the distance-based and cordon-based charging methods (Sc 3 and 6 for example).

(1) The impact of charging rate on traffic, emissions and total charge As shown in Fig. 8, in Sc 2, with the increase of charging rate, traffic conditions, emissions and total charge revenue change in different patterns. The impact of charging rate can be assessed in five sections from Fig. XX. In Section I, with the increase of the charging rate (up to approximately 0.4 RMB/km), the average network speed in the charging zone increased rapidly to the target speed (about 15 km/h lower than the free flow speed in Section III), and the average network speed of the main urban area declined slightly because some private car travelers shift their routes out of the charging zone. Accordingly, the emissions in the charging zone decreased rapidly, while there was a slight growth in emissions for the entire urban area. The total charge revenue also increased when charging rate was higher. In Section II (charging rates between 0.4 and 0.5 RMB/km), all characteristic parameters continued the trends in Section I, while the total charge revenue reached the maximum. Section III (with charging rates between 0.5 and 0.7 RMB/km) involved increased shift from private car use to public transit for travelers crossing the charging zone. The roadway network in the charging zone approached to a reasonably free flow speed (about 50 km/h) with declined VKT in the main urban area. In the realistic practice, the maximum of the charging rates in this section began to be unpractical because the road space may be underutilized, and the capacity of the public transit may face a severe challenge. Section IV (charging rates between 0.7 and 1.3 RMB/km) showed substantially reduction of private vehicle use in the charging zone, with stable free flow speed, due to high charging rates. If the charging

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

211

Fig. 8. Changes of traffic performance, emissions and total charge associated with charging rate in Sc 2.

rate is even higher, as shown in Section V (charging rates greater than 1.3 RMB/km), the roadway network VKT, speed, and emissions became much less sensitive to respond to congestion charging. (2) The impact of charging factors on charging rate and total charge Different charging zones, PT LOS and charging methods could affect the charging rates and total charge revenues when achieving the target states; we extracted charging rates and total charge revenues from the simulation results (Fig. 9). As shown in Fig. 9, the total charge revenues and charging rates both decreased with the improvement of PT LOS. Total charge revenues of distance-based and cordon-based methods within the charging zone bounded by the 3rd ring road (Sc 8– 10 and 11–13) were about 2.9 and 2.5 times as much as that bounded by the 2nd ring road (Sc 2–4 and 5–7); selecting a larger charging zone may increase the total charge revenue and the charging rate. Within the 2nd ring road, total charge revenues with cordon-based charging (Sc 2–4) were about 15% less than that charged by distance (Sc 5–7), but the revenues were similar within the charging zone bounded by the 3rd ring road (Sc 8–10 and 11–13). 3.3. Analysis of emissions impact Fig. 10 shows the results of changes in emissions under ideal target states; the results are as percentages in order to show the degree of emissions reduction due to congestion charging, and the following analysis are all based on the relative changes of emissions. Changes of CO and HC emissions were generally similar and were greater than changes in NOX emissions. Under the current PT LOS (Sc 2, 5, 8, 11), within the charging zone in Beijing, congestion charging would reduce CO and HC emissions by about 60–70%, as well as NOX emissions by about 35–45%. However, it might also have a negative impact; the congestion charging resulted in increased emissions in the external area and the entire main urban area by as much as nearly 30% and 5%, respectively. Emissions in different areas were simultaneously affected by charging zone, PT LOS and charging method. In order to further understand the impact of charging factors on emissions, we analyzed the impact by adopting MANOVA. Emissions in different areas were set as dependent variables, and charging factors were set as impact factors. Among the three impact factors, both charging method and charging zone have one degree of freedom, but PT LOS has two degrees of freedom;

Fig. 9. Total charge revenues and charging rates by adopting different charging methods.

212

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

Fig. 10. CO, HC and NOx changes within the charging zone, external area and main urban area. Note: Sc 1 is used as the base scenario for two types of charging zones simultaneously; emissions in the 2nd and 3rd ring road of Sc1 were used to calculate the emissions impact of charging scenarios with different boundries.

for the purpose of analyzing the impact of PT LOS, we use the mean square (MS) of each impact factor to capture their combined impacts. The MANOVA results were shown in Figs. 11 and 12. (1) Charging Zone As shown in Fig. 11, emissions in the charging zone were significantly affected by the size of charging zone (at 0.05 level, the same below). When adopting the larger charging zone bounded by the 3rd ring road, CO and HC emissions were reduced by about 8% (mean difference in MANOVA, the same below), and NOx emissions were reduced by about 5% (Fig. 12). The impact of PT LOS and charging method on emissions within the charging zone were not significant. The impact analysis result in the charging zone was consistent with the impact on VKT in the traffic impact analysis above. A reduction in emissions within the charging zone is usually an expected aim of congestion charging. Based on the analysis above, adjusting the charging zone is most effective in order to potentially reduce emissions within the charging zone, setting a larger charging zone may reduce emissions in the charging zone more substantially in Beijing.

Fig. 11. Impact of charging factors on emissions in different areas; each coordinate axis in the radar chart represents one pollutant in one zone.

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

213

Fig. 12. Multiple comparisons of MANOVA. The bar chart represents mean difference due to the charging factors; error bars are standard errors.

(2) External area According to the MANOVA results (Fig. 11), the size of charging zone potentially has the greatest impact on emissions in the external area. The impact of PT LOS was weaker, and the impact of the charging zone was gradually weakened by the improvement in PT LOS. The impact of charging method might be the weakest; when the area bounded by the 3rd ring road was taken as the charging zone, cordon-based charging method would relief the increase in emissions slightly relative to the distance-based method (Sc 8–10 and 11–13, Fig. 10). While congestion charging may significantly reduce the emissions within the charging zone, it may also increase the emissions in the external area, illustrating the phenomenon of ‘‘reduce and relocate.” Assessing the emissions increase in the external area is necessary in formulating a successful charging policy. According to the analysis above, a smaller charging zone associated with a higher PT LOS will potentially reduce the negative impact on emissions in the external area. Moreover, when the area bounded by the 3rd ring road is taken as the charging zone, a cordon-based charging method may further ease the negative effect. (3) Main urban area As shown in Fig. 11, the potential impact of PT LOS on emissions in the entire main urban area might be greatest; the emissions gradually decreased with the improvement of PT LOS, and emissions were even lower than the base scenario when the PT LOS reached Level 1 (Sc 4, 7, 10, 12, Fig. 10). Besides, as the PT LOS increased, the effect of emissions reduction improved. For example, when PT LOS was improved from Level 3 to Level 2 and from Level 2 to Level 1, CO emissions decreased an average of approximately 3% and 7%, respectively (Fig. 12). The impact of charging zone size ranked the second, and the impact under different PT LOS varied: the larger charging zone had a negative effect under current PT LOS (PT LOS 3), but it turned positive under a higher PT LOS of Level 3 (Fig. 10). Taking the analysis in the charging zone and external area into account, a larger charging zone seems to ‘‘enlarge” the impact on emissions as compared with a smaller zone. The impact of charging method might be the weakest; cordon-based charging could ease the increase in emissions to a certain extent (about 2–5%) only when the area bounded by the 2nd ring road was taken as the charging zone and the PT LOS was relatively high (Fig. 10, Sc 3 and 6, 4 and 7). Considering the net impact on emissions in the entire urban area is important in developing the congestion charging policy. Based on the analysis above, when implementing the charging policy in Beijing, improving the PT LOS may be the most effective way to ease the main urban area’s increasing emissions levels. Adjusting the charging zone is another method; adopting a smaller or larger charging zone, respectively, when PT LOS is lower or higher. In addition, cordon-based charging might ease the negative impact slightly when the 2nd ring road is taken as the boundary of the charging zone. We summarized the analysis above in Table 4. It should be noted that a larger charging zone may have an opposite effect in the external area and the main urban area in some scenarios (Sc 3 and 9 for example). In addition, although a smaller charging zone might have a positive impact in some cases, the charging revenue impact should be taken into account; a smaller charging zone could always reduce the total charge revenue.

4.1. Comparison with other cities To assess how representative and practical for a potential congestion charging policy in Beijing, we compared the Beijing’s case in this study with London and Stockholm, two cities with congestion charging already implemented. The charging zone areas of London and Stockholm are approximately 22 km2 and 30 km2, covering 7.3% and 16.1% of the main urban area. In our study, the charging zone area in Beijing was assumed as 63 km2 inside the 2nd ring road and 159 km2 inside the 3rd ring road, which covered 9.5% and 23% of the main urban area. Although the charging area we assumed in Beijing were much larger than that in London and Stockholm, the proportion of the charging area covered the main urban area is similar.

214

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

Table 4 Summary of emissions impact of charging factors in all analysis areas. Areas

Charging zone

Level of public transit service

Charging method

Impact

Extent

Impact

Extent

Impact

Extent

Charging zone External area

Larger charging zone may reduce emissions Larger charging zone may increase emissions

Very effective Very effective

No significant impact

Effectiveless

No significant impact

Effectiveless

Higher PT LOS may ease emissions increase

Effective

Little effective

Main urban area

Larger charging zone may reduce or increase emissions under current or higher PT LOS

Effective

Higher PT LOS may ease emissions increase, even make emissions lower than base scenario

Very effective

Cordon-based charging may reduce emissions slightly when charging in the 3rd ring road Cordon-based charging may reduce emissions slightly when charging in the 2nd ring road

Little effective

Atkinson et al. (2009) and Eliasson et al. (2009) analyzed the traffic and emissions impact of the congestion charging policies implemented in London and Stockholm respectively. The two cities substantially improved public transit services when implementing the congestion charging strategy, such as using designated bus lanes, using articulated buses and increasing bus frequency. The measures on the public transit service also demonstrated that improving the level of public transit service is very necessary when implementing a congestion charging policy, which was also shown in our study. After the congestion charging policy was adopted for a year, the VKT inside the charging zone reduced approximately 30% in London and 15% in Stockholm (the average daily charge in Stockholm was approximately half of that in London). Our scenario analysis showed that the VKT in Beijing’s charging zone could reduce by approximately 40%, given the existing high VKT levels and a more aggressive target roadway network state. The emissions impact of the congestion charging in London was evaluated with the observed air pollution concentration, instead of vehicle emissions, and the congestion charging policy was not the only implemented measure to reduce traffic congestion. Inside the congestion charging zone of London, NO and NO2 concentrations decreased by up to 30%; however, monitoring data collected in the boundary zone suggested an increase in concentrations up to 8% (Atkinson et al., 2009). While the average change of the pollutions concentration of the whole main urban area was 0.2–1.1% according to the study, the researchers indicated that there might be some unexpected influence after the congestion charging policy was implemented. Our study showed similar changes in terms of emissions impact: after implementing the congestion charging, there is reduction of emissions within the charging zone, but emissions may increase outside of the charging zone. Two models were used to access the impact of congestion charging policy on emissions in Stockholm. Since the congestion charging rate in Stockholm was about only a half as that in London, the effects on emission reduction was weaker too. The estimated reduction of the pollutants inside the charging zone varied from 10% to 14% (Eliasson et al., 2009). The results revealed that air quality was improved in many streets in the inner city too, while there was no result indicated about the average change of the pollutants in the whole urban area.

5. Conclusion In this study, we analyzed the impact of a congestion charging policy on traffic and vehicle emissions in Beijing using a combination of macroscopic traffic simulation and vehicle emissions calculation. The congestion charging scenarios we simulated in this study were composed by three charging factors, including charging zone, charging method and associated public transit service level. Our simulation results showed that congestion charging in the central Beijing urban area will possibly increase the use of public transit by up to about 13% and will potentially reduce CO and HC emissions by 60– 70%, as well as reduce NOX emissions by 35–45% within the charging zone. However, it also has a potential negative impact; it might increase the emissions in the external area and the entire main urban area by up to 30% and 5%, respectively. The impact of charging fees was simulated through the concept of monetary value of time in this study. The target charging rates were determined by the ideal effects of congestion relief in this study. We analyzed the impact of charging zone, charging method and public transit service level in a potential congestion charging policy for Beijing. Results show that a larger charging zone can reduce the emissions within the charging area, but it may increase the emissions outside the charging zone; it may also have different impacts in the entire urban area, depending on the public transit service level. Improving public transit service can effectively reduce urban area emissions and is important when developing the congestion charging policy in Beijing. Selecting a larger charging zone with sufficient public transit service may enhance the citywide emissions benefit. In some cases, cordon-based charging may optimize emissions reduction effects slightly: cordon-based charging may reduce emissions in the main urban area slightly when the 2nd ring road is taken as the boundary of the charging zone. Due to data availability, our study was limited to analyzing the impact of congestion charging during the afternoon rush hours; the impact during daytime hours should be further evaluated. Further research is also needed to determine charging rates and to statistically assess the relationships among charging, traffic, and emissions in a practical congestion charging policy.

K. Wu et al. / Transportation Research Part D 51 (2017) 203–215

215

Acknowledgments The authors are sincerely grateful to the Beijing Transportation Information Center and the Beijing Transport Energy and Environment Center for providing reliable basic data. This study is supported by Beijing Municipal Commission of Transport under the project of Development Planning of Motor Vehicles in Beijing. Financial support for this research was obtained through the National Natural Science Foundation of China (National Natural Science Foundation of China) project: The research of connectivity and accessibility matrix optimization model for the urban transport network (No. 51208014). The views expressed in this paper are those of the authors and do not necessarily reflect the official policy or any opinions from government agencies. References Atkinson, R.W., Barratt, B., Armstrong, B., Anderson, H.R., Beevers, S.D., Mudway, I.S., Green, D., Derwent, R.G., Wilkinson, P., Tonne, C., 2009. The impact of the congestion charging scheme on ambient air pollution concentrations in London. Atmos. Environ. 43 (34), 5493–5500. Barth, M., Boriboonsomsin, K., 2008. Real-world carbon dioxide impacts of traffic congestion. Transp. Res. Rec.: J. Transp. Res. Board (2058), 163–171. Beevers, S.D., Carslaw, D.C., 2005. The impact of congestion charging on vehicle emissions in London. Atmos. Environ. 39 (1), 1–5. Beijing Municipal Environmental Protection Bureau, 2014. Beijing Environmental Statement 2014: 5. (in Chinese). Beijing Transportation Research Center, 2011. Urban Road Traffic Performance Index (DB11/T 785–2011). (in Chinese). Beijing Transportation Research Center, 2014. Beijing Transport Annual Report 2014: 60–61. (in Chinese). Beijing Transportation Research Center, 2012. The 4th Beijing Resident Travel Survey Part A: General Report: 222–223. (in Chinese). Beijing Transportation Research Center, 2012. The 4th Beijing Resident Travel Survey Part I: Report of Stated Preference (SP) Trip Survey Analysis: 43–50. (in Chinese). Daniel, J.I., Bekka, K., 2000. The environmental impact of highway congestion pricing. J. Urban Econ. 47 (2), 180–215. De Vlieger, I., De Keukeleere, D., Kretzschmar, J.G., 2000. Environmental effects of driving behaviour and congestion related to passenger cars. Atmos. Environ. 34 (27), 4649–4655. Eliasson, J., 2008. Lessons from the Stockholm congestion charging trial. Transp. Policy 15 (6), 395–404. Eliasson, J., 2009. A cost–benefit analysis of the Stockholm congestion charging system. Transp. Res. Part A: Policy Pract. 43 (4), 468–480. Eliasson, J., Hultkrantz, L., Nerhagen, L., Rosqvist, L.S., 2009. The Stockholm congestion – charging trial 2006: overview of effects. Transp. Res. Part A: Policy Pract. 43 (3), 240–250. Jansson, J.O., 2008. Public transport policy for central-city travel in the light of recent experiences of congestion charging. Res. Transp. Econ. 22 (1), 179–187. Johansson-Stenman, O., 2006. Optimal environmental road pricing. Econ. Lett. 90 (2), 225–229. Johansson, O., 1997. Optimal road-pricing: simultaneous treatment of time losses, increased fuel consumption, and emissions. Transp. Res. Part D: Transp. Environ. 2 (2), 77–87. Kottenhoff, K., Freij, K.B., 2009. The role of public transport for feasibility and acceptability of congestion charging–the case of Stockholm. Transp. Res. Part A: Policy Pract. 43 (3), 297–305. Mitchell, G., Namdeo, A., Milne, D., 2005. The air quality impact of cordon and distance based road user charging: an empirical study of Leeds, UK. Atmos. Environ. 39 (33), 6231–6242. Percoco, M., 2013. Is road pricing effective in abating pollution? Evidence from Milan. Transp. Res. Part D: Transp. Environ. 25, 112–118. Pigou, A.C., 1932. The Economics of Welfare, 1920. McMillan&Co, London. Raux, C., Souche, S., Pons, D., 2012. The efficiency of congestion charging: some lessons from cost–benefit analyses. Res. Transp. Econ. 36 (1), 85–92. Sabounchi, N.S., Triantis, K.P., Sarangi, S., Liu, S., 2014. Dynamic simulation modeling and policy analysis of an area-based congestion pricing scheme for a transportation socioeconomic system. Transp. Res. Part A: Policy Pract. 59, 357–383. Tuan Seik, F., 2000. An advanced demand management instrument in urban transport. Cities 17 (1), 33–45. Vickrey, W.S., 1969. Congestion theory and transport investment. Am. Econ. Rev., 251–260 Walters, A.A., 1961. The theory and measurement of private and social cost of highway congestion. Econ.: J. Econom. Soc., 676–699 Zhang, K., Batterman, S., Dion, F., 2011. Vehicle emissions in congestion: comparison of work zone, rush hour and free-flow conditions. Atmos. Environ. 45 (11), 1929–1939.