RELATING QUALITY OF SERVICE AND POLLUTANT EMISSIONS AT ROUNDABOUTS Margarida C. COELHO Invited Assistant Lecturer, Department of Mechanical Engineering, University of Aveiro, 3810-193 Aveiro, PORTUGAL, fax: +351-234-370953, [email protected] Nagui M. ROUPHAIL Director, Institute for Transportation Research and Education (ITRE), North Carolina State University, Campus Box 8601, Raleigh, NC 27695-8601, U.S.A., [email protected] Tiago L. FARIAS Assistant Professor, Department of Mechanical Engineering, Instituto Superior Técnico – Technical University of Lisbon, Av. Rovisco Pais, Pav. Mecânica I, 2.º Andar, 1049-001 Lisbon, PORTUGAL, [email protected]

ABSTRACT The main objective of this research is to relate traffic and emission impacts for single lane roundabouts based on experimental measurements of traffic and using the “Vehicle Specific Power” for emission estimation. Experimental data for calibrating the traffic model were gathered from two single lane roundabouts located in Lisbon (Portugal) and Raleigh (North Carolina, USA). The main findings of the research are: (a) the ability to characterize three synthetic vehicle speed profiles that occur at a roundabout approach, (b) the relative frequency of these profiles appears to depend on the prevailing congestion level, and (c) correlations exist between queue length and the number of stop and go cycles experienced by vehicles queued at the roundabout. Two types of stop and go driving cycles were identified: short and long. The duration of each cycle depends on the expected queue length and the frequency of each cycle directly impacts vehicle emissions. Key-words: Emissions, Roundabout, Stop and go 1. INTRODUCTION AND RESEARCH OBJECTIVES Negotiating roundabouts often requires drivers to decelerate from, and accelerate to highway speeds and these may include one or multiple stops. One concern about roundabout operations is that vehicle emissions will increase, because of the occurrence, under certain operational conditions, of excessive delays, queue formation and speed change cycles for approaching traffic. The state of the art review revealed that several widely used models have been developed to describe traffic behavior at roundabouts. One of the most recognized models in this field is aaSIDRA (Akcelik & Associates 2002). SIDRA uses a four-mode elemental model in which synthetic drive cycles based on cruising, acceleration, deceleration and idling are represented. It is not clear whether the model takes into account the effect of multiple stop and go cycles associated with long queues and whether the drive cycles assumed are standard or based on empirical data. The use of standardized driving cycles makes macroscopic models such as COPERT III (Ntziachristos and Samaras 2000), MOBILE6 (US EPA 2002) and EMFAC (CARB 2000) less appropriate, per se, for evaluation of the “micro” scale impact of traffic interruptions (Coelho et al. 2005a, 2005b, 2005c, 2005d) such as roundabouts. The upcoming EPA model MOtor Vehicle Emission Simulator – MOVES (US EPA 2004) is intended to improve the previous modeling tools and can be used to estimate national inventories and projections at the county-level, still not appropriate for emission analysis at the corridor level. The main contributions of this research are (a) the development (based on empirical data) of a set of

limited synthetic speed profiles in roundabouts, (b) an algorithm to predict the relative frequency of their occurrence and (c) conversion between speed profiles and emissions estimation. There are already very good models in the literature that predict performance in roundabouts; this research extends them by calculating emissions without the need for additional/ more detailed inputs. The objectives of this research can be summarized as follows: 1. To quantify traffic and emission impacts of single lane roundabout operations in urban corridors. 2. To develop a methodology that can identify the parameters related to speed change cycles that may have an influence on the added emissions in the system. 3. To explain the interaction between roundabout system operational variables – mainly the approach entry volume (Qin), conflicting or circulating volume (Qconf) and geometry of the roundabout – and the resulting vehicle emissions, in particular, carbon monoxide (CO), nitrogen oxides (NOx), hydrocarbons (HC) and carbon dioxide (CO2) emissions, and queue length. 2. METHODOLOGY An approach based on experimental measurements and on the modeling of traffic and emission performance of roundabouts is presented. The methodology is summarized in Figure 1.

Figure 1: Methodology overview Parameters related to traffic flow in roundabouts were characterized from the analysis of videotapes taken at two roundabouts described in section 2.1. Key variables, such as queue length, elapsed time between two successive acceleration-deceleration cycles and number of stops a vehicle experiences until it reaches the yield line of the roundabout were measured. Synthetic speed profiles and typical stop and go cycles experienced by approaching vehicles at a roundabout were then obtained. Then, any traffic model such as aaSIDRA (Akcelik & Associates 2002) or HCM (Robinson et al. 2000; TRB 2000) could be used to estimate the queue length. Finally, the VSP methodology is applied to calculate pollutant emissions (NCSU 2002; Frey et al. 2003). The integration of the traffic and emission estimation models provides an overall pollution estimate for a roundabout approach. A simple comparison of the quality of service and emissions between roundabouts and comparable two-phase signal control is also performed. The traffic model calibration and emissions calculation are explained in some detail in the next sections.

2.1 Traffic Model Calibration Several traffic variables are needed to calculate the performance variables and to represent the speed profiles (see Figure 1). These include the approach traffic volume (Qin), the conflicting volume (Qconf) and the resulting vehicle dynamics. To calibrate the traffic model, measurements of traffic parameters were taken at two urban single lane roundabouts located in Lisbon (Portugal) and Raleigh (North Carolina, USA). The approach cruise speed is 30 km/h and 56 km/h, for the Lisbon and Pullen-Stinson (Raleigh) roundabouts, respectively. The approach leg for the Lisbon roundabout has an entry width of 6 m and an inscribed circle diameter of 23 m; the corresponding values for the Pullen-Stinson roundabout are 4 m and 27 m, respectively. Field measurements of variables were gathered using video cameras over several days at each site. Queue length, idle time and accelerations at the entry of single lane roundabouts were also taken from the videotapes. In addition, measurements of the dynamic behavior of vehicles were performed at several single lane roundabouts in the region of Lisbon with the purpose of characterizing the typical stop and go situations, using a Microwave Doppler Sensor (DATRON 2000) – DATRON M 2, from DATRON-MESSTECHNIK GmbH – for speed and distance measurement combined with data treatment software (DLS). Two types of stop and go cycles were observed: short (SSG) and long (LSG) (Coelho et al. 2005c, 2005d). The chosen value to define the distance threshold between these two types was 10 m, based on a histogram of the frequency of the stop and go distance. Table 1 depicts the characteristics of typical stop and go cycles. Table 1: Characterization of Stop and Go Cycles for Single Lane Roundabouts Typical parameters LSG SSG Maximum speed (km/h) 6.6 3.8 Distance (m) 15.1 5.2 Time (s) 8.2 4.1 Idle time before each stop and go cycle (s) 5.2 4.5 The acceleration and deceleration profiles were also derived from experimental data on vehicle dynamics. Some measurements of the vehicles’ speed at the entrance of the roundabout were performed, using a speed radar gun. The yield line idle time is estimated from a theoretical function of Qconf and the drivers’ critical gap and is described in Coelho et al. (2005d). Based on the empirical evidence, the research team identified three representative speed profiles for a vehicle approaching a single lane roundabout (see Figure 2). The three speed profiles represent: (I) an unstopped vehicle that approaches the roundabout, decelerates to handle the roundabout geometry and then accelerates as it exits the roundabout; (II) a vehicle experiencing a single stop, in which there is a full deceleration to a stop at the yield line, while waiting for an acceptable gap in the circulating lane, then accelerating to the circulating and cruise speeds; and (III) a vehicle that experiences multiple stops on the approach as it moves up the queue. The relative occurrence of each profile in the approaching traffic stream depends on the level of congestion on the approach. According to the measurements of the dynamic behavior of vehicles performed at several single lane roundabouts, the actual time spent by vehicles traveling at the circulating speed is negligible.

Figure 2: Typical speed profiles through a single lane roundabout A detailed statistical analysis (using F and Student t-Tests and residuals analyses) was performed to develop suitable regression models that best describe the relative occurrence of the 3 profiles. These proportions were found to be dependent on the prevailing levels of congestion, expressed as the sum of the approach and conflicting flow rates over short time intervals. Figure 3 presents the comparison of the experimental data with the regression models that provided the best fit. It should be noted that the percentage of vehicles that stop once was obtained as the difference between 100 and the percentage of vehicles that do not stop and those that stop more than once.

Percentage of vehicles that do not stop

100

% = 100-6.11×10-5(Qconf+Qin)2

(a)

Experimental

R2 = 0.85

90

Model

80 70 60 50 40 30 20 10 0 0

200

400

600 800 Qconf + Qin (vph)

1000

1200

1400

Percentage of vehicles that stop once

50

(b)

Experimental Model

% = 100 - % vehicles that do not stop - % vehicles that stop more than once R2 = 0.42

45 40 35 30 25 20 15 10 5 0 0

200

Percentage of vehicles that stop more than once

100

400

600 800 Qconf + Qin (vph)

1000

1200

1400

(c)

90 80 70 60

Qconf+Qin ≤ 300 vph ⇒ % = 0

50

Qconf+Qin > 300 vph ⇒ % = exp[0.00123(Qconf+Qin-300)1.2]-1 R2 = 0.70

40 30 20

Experimental

10

Model

0 0

200

400

600 800 Qconf + Qin (vph)

1000

1200

1400

Figure 3: Predictive models for the relative occurrence of speed profiles I (a), II (b) and III (c) From the experimental measurements it was also possible to correlate the measured queue length with the number of stop and go situations experienced by vehicles that experience multiple stops, as depicted in Figure 4. It is evident that when the queue length is short, there are very few long stop and go situations. On the other hand, as the queue increases, the number of long stop and go situations becomes significant. The predicted number of short and long cycles from Figure 4 is incorporated into the estimation of speed profile (III) in Figure 2.

18 Total stop&go (TSG) Short stop&go (SSG) Long stop&go (LSG)

Number of stop and go cycles

16 14

TSG = 1.997 exp(0.1124 QL)-1

12 10 8 6 4

SSG = 1.834 exp(0.0759 QL)-1

2

LSG = 1.997 exp(0.1124 QL)-1.834 exp(0.0759 QL)

0 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20

Average queue length (QL) [veh]

Figure 4: Queue length vs. number of stop and go cycles 2.2 Emissions Calculation The emissions estimation method adopted in this study is based on Vehicle Specific Power (VSP), which is a proxy variable for the engine load (NCSU 2002; Frey et al. 2003). Emission rates using this approach are based on on-board measurements of US light duty vehicles. Thus the analysis presented in this paper is limited to light-duty vehicles only. The considered pollutants are CO, NOx, HC and CO2. To ensure uniformity in comparing emissions under various scenarios, the region of influence of the roundabout was defined as the sum of the deceleration distance from cruise speed, an anticipated maximum queue length (of 20 vehicles) and an acceleration distance back to cruise speed (downstream of the roundabout). VSP is estimated from a second-by-second speed profile using the following model (NCSU 2002; Frey et al. 2003): VSP = v ⋅ [1.1 ⋅ a + 9.81 ⋅ sin(arctan( grade )) + 0.132] + 0.000302 ⋅ v 3 (1) Where, VSP is the Vehicle Specific Power [kW/metric ton], v is the instantaneous speed [m/s], a is the instantaneous acceleration or deceleration [m/s2] and grade is the terrain gradient [%]. To execute the VSP method, one must have second by second speed data for each of the 3 representative profiles: no stop, stop once or stop more than once (see Figure 2). For each secondby-second calculation of VSP, discrete VSP bins are defined (see NCSU 2002; Frey et al. 2003). The emission rates for each bin (in g/s) that were used for emission calculations are presented in NCSU (2002). These take into account the variation in odometer reading and engine displacement. For the calculations included in this paper, the emission rates from a vehicle with engine size lower than 3.5 L and with mileage less than 80,500 km were considered. Following the computations of VSP and the assignment of VSP modes, emissions are allocated and summed for each speed profile: Ni

Ei = ∑ EFn

(2)

n =1

in which: EFn is the emission factor (g s-1) assigned to the n’th second of the speed profile based on

the instantaneous VSP; Ni is the number of seconds in profile i, Ei represents the total emissions associated with each of the three profiles in Figure 2 (i.e., i = I, II or III). These values include the cruise emissions related to time spent in cruise mode required to cover the difference between the total influence area of a roundabout and the distance to perform the deceleration from cruise speed, queue length and acceleration, which is about 450 m for the considered cruise and circulating speeds. As stated earlier, for those vehicles that stop more than once, emissions related with the stop and go cycles are multiplied by their number (according to their type – short or long), which is a function of the existing queue length (Figure 4). To assess the overall emissions impact of a change in average vehicle trajectory, the emissions per vehicle due to the three different speed profiles (Figure 2) are aggregated. By estimating the proportion of vehicles that experience each of the three speed profiles (see Figure 3), the approach hourly emissions caused by vehicles that enter the roundabout (Eroundabout) can then be calculated: Eroundabout = E I ⋅ PI ⋅ Qin + E II ⋅ PII ⋅ Qin + E III ⋅ PIII ⋅ Qin (3) in which EI, EII and EIII are the emissions per vehicle due to the three different speed profiles (Figure 2); PI, PII and PIII are the proportions of vehicles that experience each of the three speed profiles (predicted from Figure 3); and Qin is the approach entry flow rate in vph. 3. NUMERICAL APPLICATION In this chapter some results are presented concerning the application of the developed methodology to the Lisbon roundabout. We first show the variation of queue length with the entry flow for different types of intersection. Next a comparison between emissions from an approach controlled by a roundabout or a signalized intersection is shown. Figure 5 shows the variation of queue length with Qin (for a Qconf of 750 vph in the roundabout) for different types of intersection control (roundabout, fixed time traffic signal and flow optimized traffic signal). The HCM-based delay formulation to calculate queue length was used for this comparison. The principal conclusion of Figure 5 is that the queue length appears to be consistently higher for traffic signals under the same flow conditions. 8 Roundabout Fixed time traffic signal Flow optimized traffic signal

7

Queue length (veh)

6 5 4 3 2 1 0 250

300

350 Qin (vph)

400

450

Figure 5: Evolution of queue length with Qin: Roundabout vs. Fixed time traffic signal vs. Flow optimized traffic signal

The presence of a roundabout was compared, in terms of emission levels, to the situation where there is a signalized intersection. In this comparison two situations were considered (Figure 6): fixed time signal and flow-optimized signal, for Qconf = 750 vph. For the fixed time, the signal settings are: Yellow time = 3 s, Red time = 42 s, Green time = 22 s. In the flow optimized traffic signal, the signal settings vary with Qin and Qconf. (a)

CO, CO2

NO, HC

Signal - Roundabout (g/h)

100000

4 3.5

10000

3

CO2

1000

2.5

100

2 10

CO

1.5

1

1

NO

0.1

0.5

HC

0.01 0

50

100

150

200

250

300

350

400

0 450

Qin (vph) (b)

CO, CO2

NO, HC

Signal - Roundabout (g/h)

100000

4 3.5

10000

3

CO2

1000

2.5

100

2 CO

10

1.5

1

1

NO

0.1

0.5

HC

0.01 0

50

100

150

200

250

300

350

400

0 450

Qin (vph)

Figure 6: Comparison of relative CO, NO, HC and CO2 emissions (g/h) between a roundabout and a signalized intersection – (a) Fixed time signal and (b) Flow optimized signal

For fixed traffic signal settings, the emissions increased when the traffic signal replaced the roundabout because the former yields a higher average queue length. From Figure 6(a), the increase of emissions caused by stopping at traffic signals (with high Qin), when compared with the situation where there is a roundabout, is 34 g/h, 4 g/h, 4 g/h and 10 kg/h (for CO, NO, HC and CO2, respectively) when Qconf is 750 vph. However, when a flow-optimized (say actuated or adaptive) traffic signal is considered, the increase of emissions caused by stopping at traffic signals (with high Qin), when compared with the situation where there is a roundabout is less pronounced. The differences are 24 g/h, 3 g/h, 3 g/h and 7 kg/h (for CO, NO, HC and CO2, respectively) when Qconf is 750 vph – see Figure 6(b). Thus, for higher entry flow values, the fixed time traffic signal will yield higher emission differentials than a flow optimized signal, when compared with a roundabout. It must be pointed out that when the traffic signal settings are well optimized, in terms of traffic service rate, the emissions can be controlled and reach values that are comparable to the ones observed in the case of a roundabout. 4. CONCLUSIONS The main motivation of this research was to quantify the impact of (single lane) roundabout operations on pollutant emissions. The parameters related to stop and go behavior (such as length of the queue, elapsed time between two successive acceleration-deceleration cycles and number of stops until a vehicle departs the roundabout) were characterized based on empirical and experimental observations. Synthetic speed profiles and their distribution were defined as a function of congestion levels. The development of this methodology involved videotaping, traffic characterization, development of speed profiles and emissions calculation. The main conclusions of the present research can be summarized as follows: 1. There are three possible and distinct speed profiles for a vehicle approaching, negotiating and exiting a single lane roundabout. These profiles cover all possible combinations of stop / no stop conditions through the roundabout. 2. For the observed single lane roundabouts, vehicles are subjected to two types of stop and go cycles, short and long. Their frequency depends on the queue length. 3. The comparison between emissions at roundabouts and traffic signals depends on the signal settings, namely if the signal is fixed time or is flow-optimized. The main conclusion is that emissions appear to be higher in the case of the traffic signal for the cases studied, although this cannot be generalized without exploring additional signal optimization schemes. The approach presented herein links variables that are routinely available in capacity analysis (entry and conflicting flows along with gap acceptance parameters) with predictions of emissions. As such, it can be considered as an extension of the HCM method to assess environmental impacts. The proposed methodology can also be viewed as a decision support tool for decision makers in the following areas: quantification of the environmental consequences of the installation of a roundabout; comparison of the emissions between several intersection systems (such as roundabouts and signalized intersections); selection of the geometric parameters of roundabouts in order to achieve balance in predefined objectives concerning traffic calming without unduly causing excessive delay for drivers, while at the same time minimize the deleterious effect of the stops on pollutant emissions. Of course, further extensions of the approach to other vehicle types (emission results cannot be generalized without exploring other classes of vehicles – size, age, mileage and type of fuel), and to multilane roundabout conditions need to be completed prior to any large scale implementation.

ACKNOWLEDGMENTS The research work of Margarida C. Coelho was supported by a PhD scholarship (SFRH/BD/4809/2001) of the Portuguese Science and Technology Foundation (FCT) and European Social Fund, within the Third Framework Programme. This research had the support of the Portuguese Science and Technology Foundation – FCT (PhD scholarship SFRH/BD/4809/2001 of FCT and European Social Fund, within the Third Framework Programme; Project POCTI/MGS/37601/01, approved by FCT and POCTI, financially supported by the European Community Fund FEDER), Luso-American Foundation (Project 193/2003) and the U.S. National Science Foundation (Project CMS-0230506). The authors would like also to thank OPEL – General Motors. REFERENCES AKCELIK & ASSOCIATES (2002). aaSIDRA User Guide. Akcelik and Associates Pty Ltd, Melbourne, Australia. CARB. EMFAC – Public Meeting to Consider Approval of Revisions to the State’s On-Road Motor Vehicle Emissions Inventory. California Environmental Protection Agency, California Air Resources Board, 2000. COELHO, M.C., FARIAS, T.L. and ROUPHAIL, N. M. (2005a). A Methodology for Modelling and Measuring Traffic and Emission Performance of Speed Control Traffic Signals. Atmospheric Environment 39(13), pp. 2367-2376. COELHO, M.C., FARIAS, T.L. and ROUPHAIL, N.M. (2005b). Impact of Speed Control Traffic Signals on Pollutant Emissions. Transportation Research Part D 10(4), pp. 323-340. COELHO, M.C., FARIAS, T.L. and ROUPHAIL, N.M. (2005c). Measuring and Modeling Emission Effects of Toll Facilities. Transportation Research Record: Journal of the Transportation Research Board, in press, TRB, National Research Council, Washington, D.C. COELHO, M.C., FARIAS, T.L. and ROUPHAIL, N.M. (2005d). Effect of Roundabout Operations on Pollutant Emissions. Submission to the 85th Transportation Research Board Annual Meeting, to be held in Washington, D.C., January 2006. DATRON. DATRON M 2 / M 3 Operating Manual. DATRON-MESSTECHNIK GmbH, 2000. FREY, H., UNAL, A., CHEN, J. and LI, S. (2003). Modeling Mobile Source Emissions Based Upon In-Use and Second-by-Second Data: Development of Conceptual Approaches for EPA’s New MOVES Model. 96th Annual Conference and Exhibition of the Air and Waste Management Association, San Diego. NCSU. Methodology for Developing Modal Emission Rates for EPA’s Multi-Scale Motor Vehicle and Equipment Emission System. EPA Contract No. PR-CI-02-10493. Office of Transportation and Air Quality, U.S. Environmental Protection Agency, 2002. NTZIACHRISTOS, L. and SAMARAS Z. COPERT III: Computer Programme to Calculate Emissions from Road Transport – Methodology and Emission Factors (Version 2.1). European Environment Agency, 2000. ROBINSON, B.W., RODEGERDTS, L., SCARBOROUGH, W., KITTELSON, W., TROUTBECK, R., BRILON, W., BONDZIO, L., COURAGE, K., KYTE, M., MASON, J., FLANNERY, A., MYERS, E., BUNKER, J. and JACQUEMART, G. Roundabouts : An Informational Guide. Publication No. FHWA-RD-00-067. FHWA, U.S. Department of Transportation, 2000. US EPA. User’s Guide to Mobile6.1 and Mobile6.2: Mobile Source Emission Factor Model. Office of Transportation and Air Quality, U.S. Environmental Protection Agency, 2002. US EPA. MOVES2004 User’s Guide – Draft. Office of Transportation and Air Quality, U.S. Environmental Protection Agency, 2004. TRB (2000). Highway Capacity Manual 2000. Transportation Research Board, Ed. National Academy of Sciences, Washington, D.C., USA.