Application of system dynamics with GIS for assessing traffic emission management policy

Transcription

1 Application of system dynamics with GIS for assessing traffic emission management policy L. Chen and W.-L. Yang Department of Water Resources & Environmental Engineering, Tamkang University, Tam-Shui, Taipei, Taiwan Abstract A system used for assessing impact from traffic management policy on urban air quality was developed and demonstrated. The system consists of two models, the trip analysis model and traffic emission assessment model. The trip analysis model is used to screen probable trip routes. Three submodels are included in the traffic emission assessment model. There is first, the Logit submodel for preference analysis to derive users' mode selection, second the traffic flow submodel for traffic speed analysis; and third, the traffic emission submodel for calculation of NO, and HC emissions of all modes under various traffic conditions. A Taipei case study was performed by the application of the system. Three traffic management policies, the carpool, rescheduling bus frequency and bus fare compensation, were assessed. The assessment of the case study concluded that carpool policy may result in reducing 15.4% NO, and 14.7% HC of daily emission. The rescheduling bus frequency policy results in reducing 3.4% HC while increasing 5.4% NO, emission. Both bus fare compensation and carpool policies lead to the reduction of both NO, and HC emissions while the carpool policy results in more a significant reduction scale than bus fare compensation does. The simulation model was combined with a GIS system to reveal the simulated spatial distribution of both traffic condition and air pollutant emission change for policy evaluation. The case study demonstrated that the system could be used for assessing a traffic management policy from the perspective of urban air quality management. 1 Introduction How to improve urban air quality has become an important issue for most mega cities. In Taipei, there is a 2.6 million population located in a 372-km2 area. Besides the dense population, there are 1.59 million vehicles with 12 million

2 daily trips traveling on the existing 18 million m* road area. Obviously, the mobile source is the major air pollutant emission in Taipei. A local investigation performed in Taipei found that the air pollutants profile in Taipei has 90.9% HC, 94.5 NO, and 99.5% CO emitted from mobile sources [l]. The strategic plans for Taipei to improve the air quality include cleaning vehicle usage and transportation management. To promote cleaning vehicle usage, the introduction of low pollutant vehicles, such as LGP bus and taxi, and the promotion of electrical motorcycle, has been considered. For transportation management means, plans have been proposed to improve the convenience of and to lower the cost of public transportation use while suppressing the use of private vehicles. To improve the public transportation usage, strategies such as exclusive bus lanes, mass rapid transit (MRT) or bus fare compensation, MRT frequency management may be considered. To suppress private vehicle usage, strategies such as mandated carpooling, increased parking fees or limited parking space may be considered [2] [3]. Success of the transportation management approach requires information as to how a user responds to the proposed transportation management strategy, so that the strategy's result may achieve the aims that the strategy proposes. Moreover, the transportation management strategy will affect many users' habits of choosing transportation means, and sometimes requires an associated infrastructure to implement. For this reason, the citizens' response to the strategy is usually a concern of the strategy planners or the policy decision-makers. The Bureau of Transportation of Taipei City applied transportation planning models TRTSIII and the simulation model TRAF-NETSIM with MOBILE-Taiwan to evaluate the air quality improvement for bus exclusive lanes and holiday transportation management strategies [4] [5]. Since intensive data are required to use MOBILES when it was combined with a traffic simulation model such as TRAF-NETSIM, it can only be performed within a very limited simulation time span, therefore the simulation with the above combination is not suitable for macro-scale application. Current models need to overcome two complications when they were applied in the impact assessment of transportation management strategy. First, current emission factor models such as MOBILE5 require heavy and costly efforts to collect many detailed traffic data for prediction application. Second, the traditional four-stage transportation planning sequence, that is trip generation, trip distribution, modal choice and assignment, does not account for the factor that individuals usually make decisions for modes and routes according to their perception of previous traffic conditions. Utilizing available data and including feedback mechanisms are two important factors in successfully applying a model to predict the impacts of transportation management strategy. Euritt et al. [6] proposed a framework for evaluating the impact of transportation control measures (TCMs). The framework integrates a mode choice model, a traffic simulation model, an emissions model, a fuel consumption model and a dispersion model to evaluate the impact of the TCMs such as bus lane, high-occupancy-vehicle and bus pricing. The system proposed in this paper used a framework similar to the one Euritt proposed with the addition of several factors to assess the impact of transportation management strategies on a macro-scale city area. In the modal choice sub-model, the

3 comfortable level that includes space, weather and driving factors is considered along with travel time and cost. In the traffic flow sub-model, this study considered modes that include car, bus, motorcycle, taxi and MRT, as opposed to the car and bus only of TRAF-NETSIM applied in Euritt's study. In this study, a spatial analysis of travel routes was also included. ASystem Dynamics approach then integrated the three sub-models, namely the Logit sub-model, the traffic flow sub-model and the traffic emission sub-model to construct the framework of this system. Detail of the framework is explicated in the methodology section. A Taipei case study was demonstrated for the use of the system to assess the impact of the MRT and taxi management strategies on the city traffic condition and the emission change of air pollutants. To present the assessment results, a GIS was combined with the system to reveal the simulated spatial distribution of both traffic patterns and air pollutant emission change for the assessed strategies. 2 Model methodology The system framework is constituted with the three sub-models as illustrated in Fig. 1. The Logit sub-model is for estimating fleet composition from user's preference before and after implementation of the proposed transportation management strategy. The traffic flow sub-model is for estimating the vehicular speed of each of the en-route modes in the above fleet composition. The traffic emissions sub-model is for estimating individual vehicular emissions from the above estimated modal speed and composition. In this study five modes (motorcycles, cars, buses, taxis and MRT) and two emissions (NO, and HC) were considered. l I System Dynamics d I Preference 1.Logit submodel 2.Traffic flow submodel 3.Traffic emission submodel Figure 1: System Dynamics based system sub-models and interface In this study, Taipei City was divided into 13 traffic zones based on current administration zones and the average trips between those zones in time frame were obtained from past investigation [7]. Based on the total road length 1, km and total road area 17,778,136 m*, those routes with its width over 6 m in the city's traffic network were modeled by 130 nodes and 216 links. The network includes 32 major arteries, 7 highways and 3 MRT routes. The trips J

4 between zones were converted into trips between nodes of the network. Path analysis is used to determine possible routes between the origin node and the destination node. Two criteria are considered for the path analysis to screen possible routes. One is that the routes shall pass through the least number of intersections and the other is the routes shall have the least total length. Based on both the least number of intersections and the least total length criteria, the top five of the most probable routes were found for further analysis. Possible modes for each link of the routes were checked and available modal combinations planning were then analyzed. Since there is transfer service for the MRT system, there are 17 combinations available for each route. The possible modal combinations are listed in Table 1. The utility of available modes on each route then determines the en-route fleet composition, speed and mode amount, therefore the accumulation of NO, and HC emission change were estimated. Table 1. Possible modal combination of trip routes Single mode Motorcycle I Two modes combination Motorcycle + MRT Three modes combination - -- Motorcycle + MRT + Bus Car Car + MRT Car + MRT + Bus Bus Bus + MRT Motorcycle + MRT + Tax1 Tax1 Tax1 + MRT Car + MRT + Tax1 MRT Bus + MRT + Motorcycle Bus + MRT + Car Tax1 + MRT + hlotorcycle Tax] + MRT + Car In the Logit sub-model, the user's preference is parameterized as a utility function of influential factors. The considered factors are cost, travel time and comfortable level. The index for the cost factor is the cost/income ratio. The travel time indexes considered include both the travel time-in-vehicle and travel time-out-vehicleldistance ratio. Space, weather and driving effort are three considered factors to reflect comfortable level. A linear equation is used to combine the considered factors of each mode. A user's choice then was quantified by the chosen probability or the utility value obtained from the equation. With the output of the Logit sub-model, the fleet composition, then the individual en-route vehicle speed can be calculated from the traffic flow sub-model. With the three sub-models, a System Dynamics approach was used to integrate the three sub-models. The causal feedback loop diagrams (CFLD) as shown in Fig. 2 are used to describe the procedure to estimate the NO,/HC emission and the relevant data flow. The factors for cost and travel time-in-vehicle of a car are similar to those of the motorcycle. However, a car user usually spends more travel time-out-vehicle than a motorcycle user because a car user usually requires more parking and walking time. In the comfortable level, a car user has the advantage

5 in both weather and space factors. The only disadvantage for a car user is the driving factor. The bus ticket price is the only cost of a bus user. For bus, the parameter for travel time-in-vehicle is dependent on the bus speed while the travel time-out-vehicle is dependent on the frequency of buses. Figure 2: Modules of the System Dynamics based strategy assessment system. The frequency of buses determines the number of en-route buses, however the nuniber of buses has no direct relation with number of bus users. Considering the comfortable level, a bus user expends no driving effort and has weather protection; however, there is the possibility of a crowded bus, which degrades its space factor. The comfortable level of a taxi is high since a taxi user can enjoy private space, weather protection and expends no driving effort. A taxi user can also enjoy no parking time; therefore it requires less travel time-out-vehicle. To enjoy the advantages, a taxi user must pay more cost than other modal users. Contrasted to the bus, the number of en-route taxis is not determined by frequency as bus, but determined by both the choice of a taxi driver's en-route incentive and taxi user's choice. A taxi driver's en-route incentive mechanism is considered in the system. The factors that affect a taxi driver's en-route incentive include the number of taxi users, traffic condition and profit obtained on the previous day. If there are too many vacant en-route taxis, it will reduce a taxi driver's income expectation. The experience of crowded traffic conditions will also reduce the number of en-route taxis for the next day. The availability of the MRT is determined by its frequency only. Since the service area for the MRT is less than the bus, MRT users usually need to transfer by connecting modes. Therefore the cost for the MRT user includes the cost of both the MRT and connecting modes. The cost, travel time and comfortable level of the

6 combination mode is determined by the summation of the individual mode within the trip routes of the combination mode, hence the utility of the combination modes can be determined. The travel time combined with the cost and comfortable factors then feedback to the modal utility calculation sub-model to update the modal utility for the modal choice of the next day. In the traffic emission sub-model, the emission of each link was obtained from the emission due to each modal amount and speed on the link. By accumulating the emission of each link in the route network, the total emission of the city then was obtained. Base on the above framework of the methodology, the system was coded with Visual Basic computer code and the GIS related database was constructed by a MapInfo GIS package. Simulation of the impacts of the transportation management strategies on air pollutant emission was demonstrated. The simulation results were used to assess the consequences of air quality and traffic condition when the assessed strategies were implemented. Three cases were simulated and assessed in this study. They are: (1) MRT Fare Compensation, (2) Rescheduling of MRT Frequency, (3) Modification of Taxi Rate. Results of the simulation are discussed below. 3 Results and discussion 3.1 Model validation To validate the assessment system, the simulation results of daily modal users distribution were compared with on-site measured data [2]. In Fig. 3, the simulated and the measured traffic conditions were compared where the difference between the simulated fraction of each chosen modal users and the measured fraction is within S%, an acceptable level. From the simulated time distribution of NO, and HC emissions in Fig. 4, two peak emission periods were found to be coincident with two daily peak periods. From the above comparisons I Motorcycle a Car 1 BUS and MRT ' Tan 1 Figure 3: Comparison of measured and simulated daily modal distribution

7 it is confirmed that the assessment system can reveal both the assessed city's daily traffic conditions and the emission scenario; hence the system is validated. 3.2 Background emission To establish the referenced emission data, current Taipei road and traffic data were input to obtain the background NO, and HC emissions. The equilibrium state is shown in Fig. 4. The estimated total daily NO, emission of Taipei is 7,622 kg/day and total HC emission is 39,520 kdday. From the emission distribution pattern, one may find that more pollutants were emitted along the main routes than along the local routes in downtown because of the difference in the carry capacity of the routes. Figure 4:Daily background NO, and HC emission with time Once a transportation management strategy is imposed on the system, a new equilibrium state will occur. By comparing the difference between the emission data of the background state with the new equilibrium state, we then can assess the impact of the imposed strategy on the system in terms of NO, and HC emission change. In the following, the results of the three strategies, namely MRT fare compensation, rescheduling MRT frequency and modified taxi rate are discussed Strategy assessment MRT fare compensation To promote the use of the MRT, compensation of the MRT fare may be considered. One may anticipate that such a strategy would attract more users to choose the MRT for transportation; this then results in less en-route modal loading and indirectly reduces the pollutant emission for the MRT emits no pollutants. For the comparison of the cost between different vehicles, the distance rate of the MRT fare was normalized into the unit of NTh. The performed simulation scenario is to reduce the MRT fare from NT2Jkm to NTllkm. The simulation results show that the strategy does attract many users to shift from other modes to the MRT especially in off-peak periods. The motorcycle and bus users increase during peak periods. This is because the limited capacity of the MRT to carry all shifted users. Those users who are not able to catch the MRT then shifted to choose the motorcycle and bus instead.

8 From the amount of change in car distribution, we found that the amount of cars increases in the areas nearby the MRT routes. This indicates that the strategy may promote long trip users to use the MRT while it increases short trip usage of other modes because of MRT transfer needs. Summarized all emission change distribution in Fig. 5, we concludes that the total daily NO, emission of Taipei was reduced 4.9 kglday because of the reduced amount of cars. The total daily HC emission of Taipei was reduced 29.6 kg/day, which is the result of the balance between the decreased car emission and the partially increase of motorcycle emission. A NOx Em~ss~on Change D~sirlbutlon (kgiday) l 73-0 hplday) -( 1 B HC Em~ss~on Change Distribuhon (kgiday) Wdv) Figure 5: The distribution of NO, and HC emission changes for MRT fare compensation Rescheduling MRT frequency With rescheduling to increase the MRT frequency, one may anticipate that with less travel time-out-vehicle the MRT will be taken and a more comfortable space can be enjoyed. Hence to increase MRT frequency will promote more users to choose the MRT. In this study, the peak periods were defined for 7-9 am. and p.m. while the others were the off-peak periods. The strategy assessment tried to simulate the impact when the MRT patrol interval in off-peak periods was reduced from 10 minutes to 5 minutes. The simulated results indicate that motorcycle users are reduced and shifted to MRT, as shown in Fig. 6. Contrasted to the MRT fare compensation strategy, increasing the off-peak MRT frequency directly increases the MRT carry capacity. The simulation results indicate that both car and motorcycle amounts increased along the MRT route connected areas. This implies that promoting MRT usage also induces an increase of MRT transfer vehicular amount. The pattern of total trip change distribution also confirms that total trip mostly concentrates on the areas along the MRT routes as shown. The simulated emissions change for rescheduling the MRT results in a NO, reduction of 8.4 ky'day and for HC a reduction of kg/day Modified taxi rate Since the taxi rate includes the beginning, distance and time rates, the cost of using taxi is the summation of the three rates. In this strategy assessment. the

9 A. NOx Erniss~on Change D~str~but~on (kqiday) (kqiday) B HC Erniss~on Change Dlstr~butlon - +0-l (kg/dsy) (kgday) 3gure 6: Distribution of NO, and HC emission changes for rescheduling MRT taxi rate was modified only for the distance part, that a taxi rate increase from NT15/km to NT 20h. The simulated results indicate that more change in trips, especially for car and motorcycle, occurred in peak periods than those in off-peak per~ods. When the taxi rate is raised, short trip taxi users tend to be promoted; on the other hand long trip users are suppressed. Notice that many taxis gather in the downtown area after raising the taxi rate. This implies that there is an increase in short trip taxi needs. Many long trip users shifted to the MRT and car from bus and motorcycle were found from the vehicular amount distribution change. The distribution pattern that the total trip is more concentrated in the downtown areas than in suburban areas also indicates that more trips occurred in short trips than in long trips. Summarized all emission change distribution in Fig. 7, the total HC emission reduction is estimated to be kg/day, which is the result of the balance between the increased car emission and the decreased motorcycle and taxi emission. For NO, emission change, there is an increase of 84.4 kg/day. The increase is the result of the balance between the increased NO, emission from cars and the reduced NO, emission from taxis. Figure 7: The NO, and HC emission change after raising taxi rate

10 Table 2 concludes the total daily NO, /HC emission of Taipei change due to the three assessed strategies. Both MRT fare compensation and MRT frequency rescheduling lead to NO, and HC emissions reduction at the same time. However, raising the taxi fare results in a HC emission reduction while resulting in an increase in the NO, emission. The contradiction between HC reduction and NO, increasing shows the need for an assessment system to predict potential impact from proposed management strategy. Table 2. Daily NO, and HC emission change for the three strategies l l l l I MRT Fare Reschedule MRT Modified Taxi Rate ANOx((kdday) Conclusion A System Dynamics based decision support system for assessing transportation management strategy on urban air quality was developed. In the system, three sub-models are included, namely the Logit sub-model for user preference, the traffic flow sub-model for vehicle speed and flow simulation, and the traffic emission sub-model for individual vehicle emission. The feature of the system is its capability to simulate daily traffic conditions and emission change dynamically with a feedback mechanism. A GIS combined with the decision support system was used to reveal the simulated spatial distribution of both traffic conditions and air pollutant emissions change for strategy evaluation. The system was demonstrated for a Taipei case study in assessing the impact of three transportation management strategies. Except for the taxi rate modification strategy, both MRT fare compensation and frequency rescheduling result in reducing NO, /HC emission. The simulation results for taxi rate modification strategy indicate an increase for NO, emission but a reduction of HC emission. The contradictory emission reduction and increasing implies that the need for strategy assessment before its implementation. The demonstrations show how the proposed system can be used for the assessment. The cases studies presented here are the selected examples to demonstrate the application of the system. More cases studies such as bus management, parking fee setting; parking space supply and combined transportation management strategies can also be performed. The application of the system can provide useful information for policy decision-makers regarding how the transportation management strategies affect urban mobile emission. References 1. Tzeng, G H., and Teng, J. Y., "Evaluation of TSM strategies for improving air quality in urban area: application of ELECTRE I1 MODEL and group decision-making technique," Transpormrion, 4,9-30(1989).

11 .41r Pollzlriol~ 1.Y Tzeng, G H., Chen, J. J., Yen, Y. K., "The Strategic Multicriteria Decision-Making Model for Managing the Quality for the Environment in Metropolitan Taipei," Asian Journal of Erlvil-onrnental Management, 4(1), 41-54(1996). 3. "Proposal for Taipei City Green Transportation," Bureau of Transportation, Taipei City Government, (1999). 4. Lai, J. L., "Effects of Transportation System Management on Energy Consumption and Environmental Pollution. Institute of Transportation," Ministry of Transportation and Communications, Taiwan, (1997). 5. "Evaluation Report of the Air Quality Improvement by Traffic Source Reduction in Taipei City," Bureau of Transportation, Taipei City Government, (1999). 6. Euritt, M. A., Qin, J., Meesomboon, J., Walton, C. M,, "Framework for Evaluating Transportation Control Measures: Mobility, Air Quality, and Energy Consumption Trade-offs," Transportation Research Record, (1444), (1994). 7. Lee, C. M., "Taipei Metropolitan Household Interview Survey," Institute of Transportation Ministry of Transportation and Communication, (1992).