Energy Saving and CO 2 Reduction Potential from Partial Bus Routes Reduction Model in Bangkok Urban Fringe

: Bus networks in many capital cities are long distances and partially overlapping with each other. As a result, waiting time is high and energy consumption e ﬃ ciency is poor. Bus operators in many countries tried to reform their bus routes to reduce waiting time and fuel costs by reducing overlapping bus routes. However, most of the reformed bus routes were complicated, which caused discomfort to passengers to use the bus service. To overcome this problem, this study proposed a new bus reformed model called the Partial Bus Routes Reduction in Urban Fringe Model (PBRU) which was a simple and passenger-friendly route operation. It resulted in 14% of total inbound and 16% of total outbound passengers receiving the beneﬁt of waiting time reduction. Most passengers wait twice at the resident bus-stop and transfer point. As a result, the overall waiting time increased by 0.72–3.75 min. The reduction of fuel consumption was consistent with increasing load factors and dependent on the time period. The bus reform operations during the o ﬀ -peak hours had more beneﬁts in terms of waiting time reduction, energy-saving potential, and CO 2 reduction than during the rush hours.


Introduction
Nowadays, the quantity of energy consumption is at a high level despite a global concern on adverse environmental issues. The transportation sector is one of the major contributors to a high proportion of energy consumption that also produces a significant amount of particulate matter (P.M.10 and P.M. 2.5) and greenhouse gas (GHG) [1,2]. Government agencies around the world attempt to reduce GHG emissions by promoting public transportation that can decrease the number of private vehicles on the road and the air pollution problem. A popular type of public transportation that many people choose to travel in urban cities is bus transit. In general, a structure of bus routes in many capital cities is a long-distance route and passengers can travel to their destinations without any transfer [3,4]. Some bus routes are circuitous and overlapping with other bus lines as they compete for passengers [5,6]. Bus operators have tried to improve the bus network to overcome the overlapping bus routes problem. Normally, the bus route reform can be divided into two types. The first reform type is the bus route reorganization with a dedicated lane for buses which can be found in Brazil, China, Turkey, and Columbia [7][8][9][10]. This measure can reduce the total travel time of passengers, Rama 2 Road is located in western Bangkok. It serves as a major route in the urban fringe area in the Chomthong and Bangkhunthian districts. In 2016, the population density of Chomthong and Bangkhunthian districts was 1490 and 5851 people/km 2 , respectively [17]. Typically, the number of passengers/km for general bus routes in the central business district (CBD) is higher than the urban fringe and the earning income in the urban fringe is lower than the CBD zone [18].
Most of the bus routes in Bangkok are long-distance routes that allow passengers to travel to their destinations without any transfer. Currently, the bus network in the urban fringe area consists of many overlapping bus lines on each major road. The overlapping bus lines lead to high waiting times and high competition for passengers that cause low revenue for each bus line [19,20]. The study research on bus reform indicated that the result benefited the urban fringe more than CBD because the transport disadvantages were higher there [21]. Motivated by these problems, this study proposed the bus network improvements by reducing the overlapping bus lines on Rama 2 Road. The study boundary covered 12.9 km from Samaedam terminal to Rama 2 T-junction as shown in Figure 1. Public transportation in this study consisted of 72% bus transit services provided by Bangkok Mass Transit Authority (BMTA) and 28% from a private bus company (from survey data from 14 August 2018 to 11 September 2018). Rama 2 Road has 14 overlapping bus lines including buses No. 17,68,76,85,101,105,140,141,142,147,172,529, 558, and 720 [22]. At present, all inbound buses start from Samaedam terminal to Rama 2 T-junction and continue to different destinations. In the opposite direction, all outbound buses conjoin at Rama 2 T-junction before heading to Samaedam terminal as shown in Figure 1.

Study Methodology
This study illustrated the Partial Bus Routes Reform in Urban Fringe Model (PBRU) by introducing a single bus line along the overlapping corridor (Rama 2 Road). Bus No. A was assigned to run on Rama 2 Road from Samaedam terminal to Rama 2 T-junction. At Rama 2 T-junction, all passengers transferred back to the regular bus lines to continue their trips. The bus route structure after reform is shown in Figure 2. The analysis methodology and calculation framework for this study are shown in Sections 3.1-3.4 and Figure 3, respectively.   [22]. At present, all inbound buses start from Samaedam terminal to Rama 2 T-junction and continue to different destinations. In the opposite direction, all outbound buses conjoin at Rama 2 T-junction before heading to Samaedam terminal as shown in Figure 1.

Study Methodology
This study illustrated the Partial Bus Routes Reform in Urban Fringe Model (PBRU) by introducing a single bus line along the overlapping corridor (Rama 2 Road). Bus No. A was assigned to run on Rama 2 Road from Samaedam terminal to Rama 2 T-junction. At Rama 2 T-junction, all passengers transferred back to the regular bus lines to continue their trips. The bus route structure after reform is shown in Figure 2. The analysis methodology and calculation framework for this study are shown in

Analysis of the Passenger Load Profile for Inbound and Outbound Directions
In this study, average passenger numbers of 27 bus samples (sample size determined by Yamane's formula with the confidential level as 80%) that operated on the overlapping section on Rama 2 Road were collected from 23 January 2018 to 13 February 2018. The number of passengers getting on and getting off all buses along Rama 2 Road was recorded from random individual buses to analyze the passenger load profile for both inbound (Samaedam terminal to Rama 2 T-junction) and outbound directions (Rama 2 T-junction to Samaedam terminal). The relationship between the total passengers on the bus, the number of passengers getting on and getting off is illustrated in Equation (1).
T 0 = The number of passengers at the bus-terminal (Inbound) or transfer point at Rama 2 T-junction (outbound) T passenger = The total passengers on the bus. The number of passengers/bus on Rama 2 Road and the number of buses from the field survey was collected from 24 August 2018 to 27 September 2018 to analyze the number of passenger throughput on Rama 2 Road. The number of passengers/bus in this study during the a.m. peak, off-peak, and p.m. peak was 62, 69, and 62 samples for inbound buses and 63, 70, 61 samples for outbound buses, respectively. The number of existing passengers from the field survey (passengers/hour) can be determined by multiplying the load factor (passengers/bus) with the number of buses (buses/hour).

Bus Frequency after Increasing the Load Factor
If the bus operators increased the load factor of the bus from the existing value to 0.41, 0.61, and 0.82, bus frequency on Rama 2 Road can be calculated by the number of passengers (from Section 3.2) and load factor characteristics (full/not full vehicle capacity). The relationship between each variable is illustrated in Equation (2) [23]. f = m/pb max (2) where f is the bus frequency (buses/period), m is the number of passengers (passengers/period), b max is the ratio between the number of passengers and the vehicle capacity (load factor), and p is the vehicle capacity (passengers/bus). In general, the vehicle capacity is 98 passengers/bus. In this research, the number of passengers did not exceed 80 passengers/bus to support the bus delay from traffic fluctuation that causes the overcrowding of passengers at the bus-stop. The relationship between the number of passengers on the bus and the load factor is illustrated in Table 1. The life cycle analysis (LCA) and life cycle cost (LCC) of diesel and NGV engines found that they were slightly different [25]. The quantity of fuel consumption in transportation can be calculated from a multiplication of the number of buses, travel distance, and fuel consumption rates as shown in Equation (3).
A = quantity of fuel consumption (liter or kg/period time) N = the number of buses on Rama 2 Road (buses/period time) D = travel distance (km/bus). In this study, the distance from Samaedam terminal to Rama 2 T-junction is 12.9 km or 25.8 km for a round trip.

Total CO 2 Emission from Fuel Consumption
The total CO 2 emission depends on the quantity of fuel consumption and emission factors. The relationship between each factor is shown in Equation (4) [26].
E = total CO 2 emission (kg CO 2 /period time) A = quantity of fuel consumption (liter or kg per period time) EF = emission factor (kg CO 2 /weight or volume) that can be calculated using emission factor per heating value of fuel proportion (kg CO 2 /TJ) and energy content value (MJ/liter or kg). The emission factor of diesel and NGV engines are 2.70 kg CO 2 /liter of diesel and 2.528 kg CO 2 /kgNGV, respectively [27][28][29]. The calculation details of emission factors of diesel and NGV are illustrated in Appendix A.

Kilogram of Oil Equivalent (kgoe) Conversion
A kilogram of oil equivalent (kgoe) can be calculated using a multiplication of the quantity of fuel consumption and energy content of fuel (kgoe = A × EC). Where A is the quantity of fuel consumption (kgNGV or liter/time period) and EC is energy content of fuel. Energy content of diesel is 36.42 MJ/liter or 0.86 kgoe/liter (1 toe = 42.244 GJ) [27]. Energy content of NGV is 42,710 Btu/kgNGV [28] or 1.07 kgoe/kgNGV (1 toe = 40.047 × 10 6 Btu) [27].

Waiting Time
The waiting time has a relationship to the number of overlapping routes and bus frequency. The waiting time can be estimated by the equation t wait = 60R/f. From this equation, t wait is the waiting time of passengers (minutes), R is the number of overlapping routes. In the case of Rama 2 Road, R = 14 and f = the bus frequency (buses/hour).

Passenger Load Profile for Inbound and Outbound Directions
The profile of total passengers on the bus each day was collected from 27 bus samples in the urban fringe area (Rama 2 Road) between 23 January 2018 to 13 February 2018. The passenger load profiles for the inbound and the outbound directions were different. The inbound buses gained more passengers as they traveled closer to Rama 2 T-junction, which caused increasing in-vehicle passengers in this direction. Therefore, Rama 2 T-junction had the highest number of passengers on Rama 2 Road as shown in Figure 4. The average number of passengers getting off and getting on the bus on Rama 2 Road was 6 and 35 passengers, respectively, which is also considered as 14% and 86% of total passengers, respectively. On the contrary, the buses traveling in the outbound direction to the urban fringe area (from Rama 2 T-junction to Samaedam terminal) tended to have passengers get off rather than get on the bus, as shown in Figure 5. The average number of passengers getting on and getting off the bus was 6 and 32 passengers, respectively, which is also considered as 16 and 84% of total passengers, respectively.
Energies 2020, 13, x FOR PEER REVIEW 7 of 18 and getting off the bus was 6 and 32 passengers, respectively, which is also considered as 16 and 84% of total passengers, respectively.

Travel Demand Analysis
From this study, the number of passengers on the bus samples along Rama 2 road (Rama 2 T-junction) was collected from 24 August 2018 to 27 September 2018. In this analysis, 62, 69, and 62 samples of the inbound buses and 63, 70, and 61 samples of the outbound buses were collected during the a.m. peak, off-peak, and p.m. peak times, respectively. The field survey results are illustrated in Table 2. It is found that the maximum travel demands of all time periods do not exceed the vehicle capacity (98 passengers). The long-range of the standard deviation of the load factor (the number of passengers on the bus) might result from uncontrollable reasons, such as (1) the fluctuated traffic on Rama 2 Road, (2) the competition of each bus driver to find more passengers by and getting off the bus was 6 and 32 passengers, respectively, which is also considered as 16 and 84% of total passengers, respectively.

Travel Demand Analysis
From this study, the number of passengers on the bus samples along Rama 2 road (Rama 2 T-junction) was collected from 24 August 2018 to 27 September 2018. In this analysis, 62, 69, and 62 samples of the inbound buses and 63, 70, and 61 samples of the outbound buses were collected during the a.m. peak, off-peak, and p.m. peak times, respectively. The field survey results are illustrated in Table 2. It is found that the maximum travel demands of all time periods do not exceed the vehicle capacity (98 passengers). The long-range of the standard deviation of the load factor (the number of passengers on the bus) might result from uncontrollable reasons, such as (1) the fluctuated traffic on Rama 2 Road, (2) the competition of each bus driver to find more passengers by

Travel Demand Analysis
From this study, the number of passengers on the bus samples along Rama 2 road (Rama 2 T-junction) was collected from 24 August 2018 to 27 September 2018. In this analysis, 62, 69, and 62 samples of the inbound buses and 63, 70, and 61 samples of the outbound buses were collected during the a.m. peak, off-peak, and p.m. peak times, respectively. The field survey results are illustrated in Table 2. It is found that the maximum travel demands of all time periods do not exceed the vehicle capacity (98 passengers). The long-range of the standard deviation of the load factor (the number of passengers on the bus) might result from uncontrollable reasons, such as (1) the fluctuated traffic on Rama 2 Road, (2) the competition of each bus driver to find more passengers by trying to reach the bus stop before others, and (3) the difference of driving speeds and driving behaviors. The average passenger demand per hour is a product between the average number of passengers on the bus (passengers/bus) and the average number of buses on Rama 2 Road (buses/hour). The inbound passenger demand during the a.m. peak, off-peak, and p.m. peak times were 3652, 1088, and 1653 passengers/hour, respectively. The outbound passenger demand during the a.m. peak, off-peak, and p.m. peak times were 1728, 1280, and 3124 passengers/hour, respectively. The inbound passenger demand during the a.m. peak time was higher than the outbound one, therefore, the inbound direction was selected for the bus frequency analysis during this period. On the other hand, the outbound direction was selected for the bus frequency analysis during off-peak and p.m. peak times due to a higher demand than the inbound one.

The Relationship between the Average Bus Frequency and Load Factor with Different Time Periods
The bus frequency varied by different load factors as shown in Figure 6. During the a.m. peak time, the average bus frequency was 83 buses/hour with an average load factor of 0.45. If the bus operator increased the load factor from 0.45 to 0.61 and 0.82 by using the PBRU on Rama 2 Road, the bus frequency would reduce to 61 and 46 buses/hour, respectively. During the off-peak time, the average number of buses at normal operation was 64 buses/hour with an average load factor of 0.20. If the bus operator increased the load factor from 0.20 to 0.41, 0.61, and 0.82 by using the PBRU on Rama 2 Road, the average number of buses on Rama 2 Road would reduce to 32, 21, and 16 buses/hour, respectively. During the p.m peak time, the average number of buses at normal operation was 71 buses/hour with an average load factor of 0.45. If the bus operator increased the load factor from 0.45 to 0.61 and 0.82 by using the PBRU on Rama 2 Road, the average number of buses on Rama 2 Road would reduce to 52 and 39 buses/hour, respectively. .00 kgoe/km-hour, respectively. The PBRU on Rama 2 Road during the off-peak period had more energy-saving potential than the others due to its low load factor. The load factor during the off-peak period could be increased from 0.20 to 0.82 to save energy. On the other hand, the range of increasing load factor at the a.m. peak and p.m. peak times was limited from 0.45 to 0.82. .00 kgoe/km-hour, respectively. The PBRU on Rama 2 Road during the off-peak period had more energy-saving potential than the others due to its low load factor. The load factor during the off-peak period could be increased from 0.20 to 0.82 to save energy. On the other hand, the range of increasing load factor at the a.m. peak and p.m. peak times was limited from 0.45 to 0.82.  Table 3 illustrates the total emissions before and after the reform of the bus network by using the PBRU on Rama 2 Road. It was found that the bus network after reform has a lower total emission than the bus network before reform. The analysis results of the air pollution reduction for each time period are shown in Figure 8. During the a.m. peak time, the total emission reduction from the existing load factor (LF = 0.45) to 0.61 and 0.82 would be 27.86 and 46.85 kgCO2/hour-km, respectively. During the off-peak time, the total emission reduction from the existing load factor (LF = 0.20) to 0.41, 0.61, and 0.82 would be 40.52, 54.44, and 60.77 kg CO2/hour-km, respectively. During the p.m. peak time, the total emission reduction from the existing load factor (LF = 0.45) to 0.61 and 0.82 would be 24.06 and 40.52 kg CO2/hour-km, respectively.   Table 3 illustrates the total emissions before and after the reform of the bus network by using the PBRU on Rama 2 Road. It was found that the bus network after reform has a lower total emission than the bus network before reform. The analysis results of the air pollution reduction for each time period are shown in Figure 8. During the a.m. peak time, the total emission reduction from the existing load factor (LF = 0.45) to 0.61 and 0.82 would be 27.86 and 46.85 kgCO 2 /hour-km, respectively. During the off-peak time, the total emission reduction from the existing load factor (LF = 0.20) to 0.41, 0.61, and 0.82 would be 40.52, 54.44, and 60.77 kg CO 2 /hour-km, respectively. During the p.m. peak time, the total emission reduction from the existing load factor (LF = 0.45) to 0.61 and 0.82 would be 24.06 and 40.52 kg CO 2 /hour-km, respectively.   Figure 9 illustrates the percentage of energy or CO2 reduction potential for a different time of a day. The relationship between the percent of daily energy or kgCO2 reduction after the applied PBRU (y-axis) and load factor (x-axis) from the existing load factor to the maximum load factor is shown in the least polynomial degree at SSR/SSE or R 2 > 0.99. The relationship between the percentage of energy or CO2 reduction and load factor during the a.m. peak and the p.m. peak time was YA.M./P.M.= −214.6 X 2 + 394.16 X − 133.92 for the load factor ranging between 0.45 to 0.82. The potential of energy or CO2 reduction when the bus operators increased the load factor from the existing (LF = 0.45) to 0.61 and 0.82 would be 27% and 45%, respectively. The percentage of energy or kg CO2 reduction during the a.m. peak and the p.m peak time was similar due to their identical load factors as shown in Table 2. The relationship between the percentage of energy or CO2 reduction and load factor during the off-peak time was YOff-peak = 464.27 X 3 − 960.8 X 2 + 696.45 X − 106.03 for the load factor ranging between 0.20 to 0.82. The potential of energy or CO2 reduction during the off-peak time when the bus operators increased the load factor from existing (LF = 0.20) to 0.41, 0.61, and 0.82 would be 50%, 67%, and 75%, respectively.  Figure 9 illustrates the percentage of energy or CO 2 reduction potential for a different time of a day. The relationship between the percent of daily energy or kgCO 2 reduction after the applied PBRU (y-axis) and load factor (x-axis) from the existing load factor to the maximum load factor is shown in the least polynomial degree at SSR/SSE or R 2 > 0.99. The relationship between the percentage of energy or CO 2 reduction and load factor during the a.m. peak and the p.m. peak time was Y A.M./P.M. = −214.6 X 2 + 394.16 X − 133.92 for the load factor ranging between 0.45 to 0.82. The potential of energy or CO 2 reduction when the bus operators increased the load factor from the existing (LF = 0.45) to 0.61 and 0.82 would be 27% and 45%, respectively. The percentage of energy or kg CO 2 reduction during the a.m. peak and the p.m peak time was similar due to their identical load factors as shown in Table 2. The relationship between the percentage of energy or CO 2 reduction and load factor during the off-peak time was Y Off-peak = 464.27 X 3 − 960.8 X 2 + 696.45 X − 106.03 for the load factor ranging between 0.20 to 0.82. The potential of energy or CO 2 reduction during the off-peak time when the bus operators increased the load factor from existing (LF = 0.20) to 0.41, 0.61, and 0.82 would be 50%, 67%, and 75%, respectively.  Table 4 illustrates the waiting time before and after combining the bus lines on the overlapping section of Rama 2 Road. The PBRU model could cut down the waiting time by 93% or 14 times if the load factor between before and after reform was identical. During the a.m. peak time, the waiting time of the existing load factor (LF = 0.45) was 10.12 min. If the bus operators increased the passenger load factor to 0.61 and 0.82, the waiting time after reform would be 0.98, and 1.30 min, respectively. The waiting time before and after reform during off-peak and p.m. peak times is illustrated in Table  4.  Table 5. The annual fuel-saving potential when the bus operators increased the load factor from existing (LF = 0.45) to 0.61 and 0.82 would be 136,942 (27%) and 230,312 USD (45%), respectively (based on the market price of diesel and NGV on 5 May 2019, 28.09 THB/liter for diesel and 16.01 THB/kg for NGV) [30]. In general, the cost of the fuel was 30.22% of the total operating cost, which consists of the drivers' and employees' remuneration, fuel cost, maintenance, depreciation, and other expenses   Table 4 illustrates the waiting time before and after combining the bus lines on the overlapping section of Rama 2 Road. The PBRU model could cut down the waiting time by 93% or 14 times if the load factor between before and after reform was identical. During the a.m. peak time, the waiting time of the existing load factor (LF = 0.45) was 10.12 min. If the bus operators increased the passenger load factor to 0.61 and 0.82, the waiting time after reform would be 0.98, and 1.30 min, respectively. The waiting time before and after reform during off-peak and p.m. peak times is illustrated in Table 4.

Economic Benefits of the Bus Reform Using the PBRU in the Bangkok Urban Fringe
A.M. Peak Time The analysis results for the a.m. peak time are illustrated in Table 5. The annual fuel-saving potential when the bus operators increased the load factor from existing (LF = 0.45) to 0.61 and 0.82 would be 136,942 (27%) and 230,312 USD (45%), respectively (based on the market price of diesel and NGV on 5 May 2019, 28.09 THB/liter for diesel and 16.01 THB/kg for NGV) [30]. In general, the cost of the fuel was 30.22% of the total operating cost, which consists of the drivers' and employees' remuneration, fuel cost, maintenance, depreciation, and other expenses [31]. The total annual operating cost-saving potential if the bus operator increased the load factor from existing to 0.61 and 0.82 would be 14,469,104 THB (453,151 USD) and 24,334,402 THB (762,117 USD), respectively. (1 USD = 31.93 baht) [32]. Off-Peak Time From analysis results in Table 6, the annual fuel-saving when the bus operator increased the load factor from existing (LF = 0.20) to 0.41, 0.61, and 0.82 would be 531,170, 713,759, and 796,754 USD, respectively. This leads to the total annual operating cost-saving potential of 1,757,676, 2,361,877 and 2,636,513 USD, respectively.  Table 7. The annual fuel-saving potential when the bus operator increased the load factor from existing (LF = 0.45) to 0.61, 0.82 would be 118,268 (27%) and 199,189 USD (45%), respectively. The annual total operating cost-saving potential if the bus operator increased the load factor to 0.61 and 0.82 would be 391,357 and 659,128 USD, respectively. When comparing the saving benefits between different peak times, it was found that the off-peak time could provide the highest potential for fuel reduction than other peak times. This was because the existing load factor during the off-peak was lower than during rush hours as shown in Table 2. This provided more opportunity for the bus operators to increase the load factor to the desired level. Moreover, the off-peak time accounted for 8 h of operations while each peak time accounted for only 3 h, e.g., 6.00-9.00 a.m. or 5.00-8.00 p.m.

Waiting Time Reduction
Bus routes after the reform in this study reduced the waiting time for passengers who traveled within Rama 2 Road. From Table 4, the range of waiting times for buses on Rama 2 Road after the reform of the routes throughout the day was 0.72-3.75 min, which is similar to the hub and spoke model in Bangalore city [15] and much lower than the bus network after the reform in Barcelona (waiting time = 6.18 min) and Santiago (waiting time = 8.20 min) [12][13][14]. However, only 14% of total inbound passengers and 16% of total outbound passengers received the benefit after the reform of the routes (see Figures 4 and 5). The majority of passengers lost the benefit because they waited at the bus-stop twice (comprised of the bus-stop near the residence and the transfer point). At the peak time, most passengers would slightly lose the benefit from waiting twice. From the results in Table 4, the maximum waiting time increases did not exceed 1.30 min at the a.m. peak and 1.54 min at the p.m. peak times. At the off-peak time, most passengers have a waiting time increase of 1.88-3.75 min. The bus operator can reduce the waiting time by increasing the load factor from the existing to 0.41 which will cause the overall waiting time to increase by only 1.88 min.
However, an advantage of the PBRM, is that passengers do not necessarily walk for the transfer. It is different from the bus reform in Barcelona that reduced overlapping routes to increase bus frequency, but the passengers must walk for the transfer. The general loss time of a bus-to-bus transfer was 5-50 min and the average was 22 min [33].

Energy Saving Potential and CO 2 Emission Reduction
Many literature reviews illustrate the energy-saving potential and CO 2 emission reduction after bus route reform over one day. In actuality, the energy-saving potential and CO 2 emission reduction depend on the time of the day.
From the current bus operations data in Table 2, the a.m. and the p.m. peak time already had a relatively higher load factor than the off-peak hours. Therefore, the benefits of the bus reform on the energy-saving potential and CO 2 emission reduction during peak hours would be little and could expect some limitations. If the bus operators choose to increase the load factor during rush hours, the level of service (LOS) for passengers would be lower. Consequently, the passengers would be likely to be dissatisfied with the bus service.
During the off-peak hours, although the bus operators attempted to reduce the number of fleets due to low travel demand, the average load factor was still low (existing load factor = 0.20). The bus operators had a high potential to increase the load factor from the existing value (LF = 0.20) to the maximum value (LF = 0.82). Therefore, the energy-saving potential and CO 2 emission reduction during the off-peak hours were likely to be higher than the rush hours.
Apart from the direct benefits, the bus route reform by the PBRU also had indirect benefits. The reformed bus network also had a lower risk of service disruption. When the connecting bus lines experienced any delay from traffic or accidents, services on the bus line on Rama 2 Road would not be affected. Therefore, the reformed bus routes structure would be more reliable than the overlapping bus routes structure.

Strategy of the Transfer Point Management
The PBRU can reduce the overlapping bus routes in an urban fringe area. As a trade-off, all passengers have to connect to other bus lines at the transfer point to go to their desired destinations.
In this case, the designated transfer point (Rama 2 T-junction) is expected to be overcrowded with transit passengers. Therefore, the bus operators should re-design the transfer point to accommodate all passengers in the waiting area. If the bus operators are not able to build a large bus-stop at the transfer point, one strategy could be to arrange the bus-stop into multiple sections with a designated group of bus lines to spread out the demand as shown in Figure 10. The other possible problem at the transfer point is the uncertainty in the bus headway from traffic congestion that causes overcrowding of passengers at the transfer point. The suggestion for prevention of this problem is sparing a small area for the bus terminal at the transfer point to serve the passengers when the transfer point has overcrowding passengers.

Application of the PBRU for Other Bus Networks
In general, the pattern of the bus network in the capital city can be divided into three types. The first group is traveling from the urban fringe to another urban fringe area. The second group is traveling from the urban fringe to the central business district area (CBD) and the third group is traveling between the CBD area. The bus routes in Bangkok have 207 routes. Over 75.36% of total bus routes operate from urban fringe to urban fringe and urban fringe to the CBD area. Approximately 50.72% of total bus routes operate in the urban fringe zone and overlap with other bus routes. For example, there are 16 overlapping bus routes in Phetkasem Road, 15 overlapping bus routes in Sukhumvit Road, 10 bus routes in Bangna-Trad, and Phahonyothin Road, etc. [22,34]. Therefore, the bus route reform by the PBRU can be applied in many roads in Bangkok that would cause a high potential for energy-saving in the public transportation system and a significant reduction of air pollution in the city.

Suggestion for Further Research
The PBRM can reduce energy and CO 2 in the bus network. However, the benefit of the waiting time reduction effect is only 14% of the total inbound passengers and 16% of the total outbound passenger, as shown in Figures 4 and 5. The methodology to increase the passenger benefit is expanding the transfer point 1 km by (1) combining 9 bus routes that turn left at Rama 2 T-junction, (2) combining 5 bus routes that turn right at Rama 2 T-junction. The result of increasing the benefits for the passenger is (1) a lot of passengers on Rama 2 Road can exchange with the other bus routes on Suksawat Road, such as buses No. 20, 21, 35, 37, 75, 82, and 195, and (2) expanding the transfer point can increase the possibility of passenger destinations. The waiting time, energy, and CO 2 reductions of further research should investigate the precise results to explore the optimal bus routes structure in the urban fringe area.

Conclusions
This research illustrates the bus reform strategy by the Partial Bus Routes Reduction in Urban Fringe Model (PBRU) along the overlapping section on Rama 2 Road. The advantages of the PBRU bus routes are the decrease in fuel consumption and CO 2 emissions when compared with the regular bus route. However, most passengers' waiting time slightly increased from regular routes (0.72-3.75 min), but the PBRU is better than other bus route reform in that some passengers always get confused with the new network and suffer from transferring from bus-to-bus (5-50 min). The PBRU can be applied for other overlapping bus routes from the urban fringe areas to the central business district area. The results from this study would be useful information for any bus operators (public and private buses) wanting to restructure their bus operations.

Acknowledgments:
The authors acknowledge the Bangkok Mass Transit Authority for sharing information and permitting the authors to collect data on their buses.

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

Emission Factor for Diesel and NGV
The details of the source emission factors of diesel and NGV are shown below.

1.
The source of emission factors for diesel.

2.
The source of emission factors for NGV.