Potential of Ecological Benefits for the Continuous Flow Intersection

Energy conservation and emission reduction from the transportation sector are of great significance in coping with the global energy and environmental crisis. As the bottleneck of urban road traffic, intersection burdens the urban environment greatly. When the volume of left-turn traffic is large, the continuous flow intersection (CFI) can effectively improve intersection operation efficiency. This paper first put forward the definition and application conditions of CFI. Then its mechanism for energy saving and emission reduction was analysed. CFI transformation was designed taking a typical intersection in Xi’an as an example. Operating efficiency, energy consumption and emissions of the intersection before and after CFI transformation were evaluated using the VISSIM model. The results show that energy consumption and emissions in the intersection are greatly reduced after CFI transformation. Queue length is reduced by more than 41%. Energy consumption and pollutant emission are reduced by about 8%. Through the simulation analysis, the emission reduction benefits most when the volume of left-turn traffic is 80%–85% of the design capacity, and the ratio of left-turn traffic over through traffic is maintained between 50% and 100%. This study suggests that CFI is suitable for large-scale promotion with careful examination.


INTRODUCTION
The share of energy consumption from the transportation sector has risen from second place in the 1990s (about 25%) to first place in 2018 (29%), overtaking the industrial sector (see Figure 1). Meanwhile, the transportation sector is now the second-largest emitter of CO 2 and is growing steadily, as shown in Figure 2. With the increasing number of cars per capita in developing countries, energy consumption and emissions from the transportation sector will continue to increase in the coming years. Therefore, a sustainable transportation system can make a great contribution to tackling global climate change and the energy crisis. The potentials for energy and emission reduction from vehicle energy efficiency improvement, green travel mode structure change [1], alternative clean fuels shift [2] and intelligent control [3] have been addressed in recent literature. The potential from the infrastructure form side [4], however, is little mentioned.
Intersection, as the bottleneck of urban road traffic, burdens the urban environment due to traffic delay, energy consumption and emissions caused by vehicle brake-start behaviours. Improving intersection efficiency could contribute to a sustainable transportation system. Continuous flow intersection (CFI) [5][6][7] is a promising intersection design when left-turn traffic is large. CFI works in this way: when left-turn traffic is about to arrive at an intersection, a pre-signal light is set at an appropriate position to make them move to the opposite exit lane in advance. This left-turn traffic waits for the signal light to turn green at the main signal intersection, then is released together with through traffic (as shown in Figure 3). This design makes full use of the space and time resources of the intersection. By reducing the conflict between left-turn vehicles and vehicles in other directions, the primary intersection does not need to set the dedicated signal phase for leftturn traffic, thus shortening the signal cycle. China's first CFI intersection was successfully transformed in 2017 on Shenzhen Caitian-Fuhua Road, which greatly relieved the intersection traffic pressure at that time. Initially, studies on CFI put more emphasis on operating efficiency, focusing on intersection geometric design and signal timing optimisation [8,9]. To improve the capacity of intersections, Coates et al. [10] improved the geometric design of intersections and established an optimisation model. Subsequently, considering pedestrian safety, Coates et al. [11] reassessed the impact of intersection geometric design on CFI operation efficiency. Among geometric design elements, displaced-left lane length design is a research focus area [12]. Moreover, signal timing optimisation of CFI has also attracted the attention of scholars [13][14][15][16][17]. In 2015, Zhao et al. [18] proposed a generalised lane-based comprehensive design optimisation model for CFI types, lane markings, left-turn lane length and signal timing. A large number of numerical analysis results show the effectiveness of this method. In the same year, Sun et al. [19] proposed a simplified continuous flow crossover design (called CFI-Lite), which can let left-turn and through traffic go simultaneously without installing a sub-signal light, laying a foundation for the further promotion and application of CFI.
With the implementation and gradual promotion, safety and mobility assessment are important considerations. In 2020, Qu et al. [20] studied the operation safety of CFI in the United States based on its accident data from 2011 to 2018. The results showed that although some new problems emerged, such as traffic signs and access control management, CFI significantly reduced accidents related to left and right turns and did not increase the overall collision frequency. It is suggested that traffic engineers need to carefully consider different aspects of CFI design, including access management, traffic signal coordination and driver acceptance in the implementation of CFI. In 2021, Ahmed et al. [21] evaluated the mobility of pedestrian and bicycle treatments at complex CFIs.
Although the studies above confirmed the high operating efficiency and safety of CFI, the existing studies did not mention the potential of CFI in energy conservation and emission reduction, which directly determines whether CFI can be vigorously promoted and applied in the context of carbon peak and carbon neutral. It has been widely suggested that CFI can decrease travel time and delay vehicles passing through the intersection. As a result, traffic flow is generally more stable, decreasing fuel consumption and emissions. However, due to the existence of the sub-intersection, left-turn vehicles and some cleared-through vehicles increased the number of stops. Due to constant braking and starting (especially starting) behaviours, fuel consumption and emissions will increase. The final energy-saving and emission-reduction effect of CFI depends on the tradeoff between the pros and cons mentioned above.
The purpose of this paper is to identify the energy-saving and emission-reduction potential of CFI and to put forward standards for CFI transformation from traditional intersections to enhance environmental benefits. A typical intersection in Xi'an city, China, was selected for CFI transformation, including geometric design and signal control design. Finally, a quantitative analysis based on the VISSIM simulation was conducted. Operational efficiency, energy consumption and emission differences were compared before and after the intersection transformation. Quantitative relationships presented can provide transportation professionals and officials with more accurate and precise guidance to facilitate better decisions in considering, evaluating and designing a CFI.

Overview
To ensure the operation efficiency of CFI, transformed intersections should meet certain index requirements, as shown in Table 1. According to the requirements, an intersection in Beidajie, Xi'an, Shaanxi Province, China, was selected as the object of CFI transformation. The current situation of the intersection is shown in the third column of Table 1. The current signal phase design of the intersection is illustrated in Figure 4. Signal period, signal duration and current traffic volume are shown in Tables 2 and 3, respectively. The data was collected by on-site video recording in the evening peak hour (from 17:00 to 18:00).  While there is a bus stop near the intersection, it can be displaced due to enough lane length.

Section length design
According to the requirements shown in Table 1, we only designed the displaced left-turn lanes for the South-North direction, as shown in where l 1 is the section length of displaced left-turn lanes (m); Q el is the traffic flow from the South-North direction (pcu); h s is the headway of left-turn vehicles in the queue (m); k is the non-uniformity coefficient of left-turn vehicles that arrived in one signal period. Commonly, it will take the value from 1.5 to 2. m is the number of left-turn lanes; n is the number of signal cycles within one hour.
Length of crossover lane section l 2 : At the secondary intersection, vehicles need to enter the displaced left-turn lanes through crossover lanes. The length of the crossover lane section is related to lane width, double yellow lines width, and minimum turning radius length of vehicles. l 2 should satisfy the following condition (Jiang et al. 2019): where l 2 is the section length of crossover lanes (m); r is the minimum turning radius of vehicles (m); s is the intermediate variable in the geometric calculation process; W 1 is the width of one lane (m); W c is the width of double yellow lines (m).
Length of storage lane section l 3 : Since displaced left-turn vehicles need to stop once at the secondary intersection, it is necessary to design the section length for storage lanes (l 3 ). l 3 should satisfy the same requirement as shown in Equation 1. At the same time, the sum of three lengths should satisfy the following requirement: where l is the length of the intersection with the adjacent upstream intersection (m).

Traffic signal design
Signal phase: Compared with the traditional intersection, CFI can let left-turn traffic and through traffic from the South-North direction go simultaneously, reducing the number of signal phases. However, a secondary signal is required. The phase design of this two-way control is shown in Figure 6. S i1 is the saturation flow rate of the first traffic direction in phase i (pcu/h). Based on the information in Figure 6 and where g ei is the effective green time for phase i (s). The signal timing of CFI in this case study is illustrated in Figure 7. Moreover, to improve the efficiency of the intersection and reduce the number of vehicle stops, the relationship between the main signal and the pre-signal should meet the requirements shown by where t 1 , t 2 , t 3 , t' 4 are shown in Figure 7; l r is the path length of left-turn vehicles from eastern and western directions at the intersection (m); V el is the average speed of vehicles turning left from eastern and western directions (m/s); v ss is the average speed of through vehicles from southern and northern directions (m/s). Equation 10 is to guarantee that left-turn vehicles from eastern and western directions entering the intersection at phase 2 can completely pass the crossover lane section, avoiding the second stop. Equation 11 is to allow displaced left-turn lanes to be closed later than the opening of phase 1. However, the time gap has an upper bond as shown in Equation 11.
Signal cycle Phase 1

Phase 2
Phase 3 Presignal at the secondary intersection

SIMULATION EXPERIMENT
In this study, the VISSIM simulation model was used to evaluate the energy-saving and emission-reduction potential of CFI transformation. To make simulation results comparable, it is necessary to ensure that the current situation in the simulation world is consistent with what is in the actual situation. Therefore, parameters in the simulation software need to be calibrated. Referring to the method proposed by Sun and Yang [22], the parameter calibration process is shown in Figure 8.
In the calibration process, two indicators were selected to judge whether the simulation world is consistent with what is in the real world. The travel time of vehicles required to pass 200 m through the intersection from the east to the west was selected as one verification indicator. Moreover, the maximum queue length of through traffic at the east approach was selected as another indicator. The two indicators can be easily measured in the actual road network and can be simulated in VISSIM. The core parameters of driving behaviour in VISSIM are those from a car-following model and lane-changing model [22], as shown in Tables 4 and 5, respectively. We only calibrated these parameters. The influence of each parameter, however, is different. Parameters with higher sensitivity should be selected for calibration. We used variance analysis to select parameters with high sensitivity. At last, the orthogonal experiment was used to determine the optimal parameter combination.  Maximum forward-looking distance The maximum distance that a driver can observe when looking ahead. The default value is 250 m.

Mean stop spacing
The average resting distance between the front and rear vehicles. The default value is 2 m.
An addition to the safe distance As a factor participating in the calculation of safe distance. The default value is 2 m.
Multiples of the safe distance As a factor participating in the calculation of safe distance. The default value is 3 m.

Parameter Description
The wait time before eliminating The maximum amount of time a vehicle can wait for a lane change on a road segment before being removed from the network. The default value is the 60 s.

Minimum vehicle spacing
The minimum spacing in a stop required to successfully change lanes for the vehicle behind. The default value is 0.5 m.
Maximum deceleration Maximum deceleration of the vehicle. The default value is -3 m/s 2 .
Acceptable deceleration When below the maximum deceleration, the desired deceleration is taken as the maximum deceleration. The default value is -1 m/s 2 .
We used the ANOVA approach to select calibrated parameters based on the simulation results. The simulation duration time was set to be 4000 s. 400 s were used to warm up the intersection. After that, simulation results were saved once every 600 s, for a total of 6 times. The parameter representing multiples of the safety distance of a vehicle was taken as an example to show the analysis process. Values of the parameter and simulated travel time in different periods are shown in Table 6. Simulated maximum queue lengths in the different periods are shown in Table 7. According to the results of ANOVA for the two indicators shown in Tables 8 and 9, vehicle travel time was significant at the significance level of 10%, and the maximum queue length was significant at the significance level of 5%. Therefore, multiples of the safe distance significantly impacted the simulation results. As a result, it should be calibrated. Similarly, an addition to the safe distance and mean stop spacing need to be calibrated as well.  The orthogonal experimental design was used to find the optimal combination of three parameters calibrated. Five levels for each parameter were designed, as shown in Table 10. Therefore, the L 25 5 3 orthogonal table was produced. Based on experimental results, the combination which produces the closest results to the actual situation was finally selected (coloured in bold in Table 10).

EFFICIENCY AND ENVIRONMENTAL IMPACTS OF CFI TRANSFORMATION
After the CFI transformation, improvements in the intersection are listed in Table 11. Operation efficiency: As can be seen from Table 11, average vehicle travel time can be reduced by 19.5%, and average vehicle delay can be reduced by 11.5%. The queue length of vehicles at the intersection can be reduced by 43.6%, and the maximum queue length can be reduced by 29.4%. The results prove that CFI transformation has a significant effect on improving the operation efficiency of intersections. The efficiency improvement will lead to the reduction of energy consumption and emissions at the intersection.
Parking rate: CFI adds an auxiliary signal control, which results in secondary parking for left-turn traffic. As a consequence, the parking rate will increase, as shown in Table 11. The increase in the parking rate will lead to a rise in energy consumption and emissions. In this case study, CFI transformation will significantly reduce energy consumption and emissions. However, combining the results in operational efficiency and parking rate, it is found that the impact of CFI transformation on intersection emissions and energy consumption is not clear. It depends on the tradeoff between efficiency improvement and parking rate increase. Namely, it is related to left-turn traffic volume and through traffic volume. To enhance the sustainability of CFI transformation (efficiency improvement as well as emissions and energy consumption reduction), the optimal range for traffic flow volume, including the ratio between left-turn traffic and through traffic, needs to be clarified. We adjusted the total traffic volume while keeping ratios of flow from different directions unchanged. The emissions' improvement with the increase of left-turn traffic flow saturation is described in Figure 9a. As can be seen, the reduction of emissions is in a quadratic parabola relationship with traffic volume. If the traffic volume is too large or too small, the emission reduction benefit of CFI transformation cannot be guaranteed. In this case study, when the volume of left-turn traffic is close to 80%-85% of the capacity, the emission reduction benefit is obvious most. Keeping the total volume as the optimal value, we adjusted the ratios of left-turn traffic volume over through traffic volume. The simulation result was described in Figure 9b. With the proportion increasing, the emission reduction firstly witnesses a rising trend, then decreases. When the ratio is maintained between 50% and 100%, the reduction benefit is higher, reaching about 17%.

CONCLUSION
Sustainable development of intersections, as the bottlenecks of road transportation, is critical. Focusing on intersections' reduction potential in terms of energy consumption and emissions, a traditional intersection was transformed into a CFI to fully utilise the time and space resources. After analysing the coordinated relationship between the left-turn signal at the road section and the intersection main signal, the phasing scheme of the improved CFI was designed. The VISSIM model was employed to evaluate the performance of the CFI.
The results confirmed the effectiveness of CFI from both theoretical and empirical perspectives. It can not only significantly improve the operation efficiency of intersections, but also has great potential for energy conservation and emission reduction. We only set displaced lanes at two approaches in the case study. If they are set at four approaches, the improvement in terms of energy consumption and emissions could be higher. By only adjusting the lane function without influencing the land-use area and topological form of the original intersection, CFI transformation can be widely promoted. However, it should be noted that not all intersections are suitable for CFI. To ensure the environmental benefits, factors listed in this study, such as intersection form and traffic volume, should be carefully checked. Moreover, when CFI is implemented empirically, it is suggested to do the simulation first.
The limitations of this study can be summarised as follows. Firstly, using the simulation tool, some measurement errors are inevitable. VISSIM is not specific for evaluating environmental influence. Measurements for different emissions are rough estimates. Environment-specific software (for example EnViVer) needs to be further employed to distinguish the difference in emissions. Secondly, a CFI transformation at four approaches needs to be designed to fully explore the reduction potential for energy consumption and emissions. Finally, more empirical studies are required to justify the methodology developed here and the reasonableness of CFI.