You are on page 1of 16

Improving left intersections

turn

safety

at

signalized

design guidelines are identified and recommendations are made for further investigation and research. Résumé La sûreté du trafic qui vire à gauche aux intersections signalisées dépend d'une interaction complexe du volumes du trafic, la géométrie de l’intersection, la vitesse de fonctionnement et mise en phase des signaux et des détails de synchronisation. Cet article développe et décrit un modèle quantitatif qui peut être employé pour évaluer l'impact potentiel de ces facteurs sur la sûreté et la capacité de mouvements du trafic qui vire à gauche aux intersections signalisées. En particulier le modèle est employé pour évaluer l'effet des choix de conception d'intersection sur la sûreté et la capacité potentielle d'un mouvement de virage à gauche. La recherche prouve qu'aux intersections conçu en accordance aux normes provinciales de l’Ontario (sans virage à gauche entièrement protégé par une indication), dans certaines conditions, les intersections n'ont pas la distance proportionnée de vue pour un virage à gauche, et que dans ces cas-ci la sûreté et la capacité d'un mouvement de virage à gauche dépend non seulement du volume de virage à gauche (comme assumé par le manuel de capacité de route) mais également du volume d'opposition de virage à gauche et du volume traversant adjacent. Puisqu'un guide de capacité tel que le manuel de capacité de route et le logiciel d'analyse de trafic tel que Synchro® n’expliquent pas explicitement les effets de la géométrie d'intersection et des distances limitées de vue, ils peuvent aussi surestimer la capacité d'un mouvement de virage à gauche. Des mesures d'amélioration de la sûreté de virage à gauche sont identifiées. Ces mesures incluent fournir des phases protégées de virage à gauche, plaçant des boucles de détection plus près de la ligne d'arrêt, diminuant ainsi l'excentrage négatif entre les ruelles opposante de virage à gauche et présenter une ruelle d'ombre entre une ruelle de

Calvin J. Mollett M(Eng) Regional Municipality of York: Ontario

Abstract The safety performance of left turn traffic at signalized intersections depends on a complex interaction of traffic volumes, intersection geometry, operating speeds and signal phasing and timing details. This paper develops and describes a quantitative model that can be used to assess the potential impact of these factors on the safety and capacity of left turn movements at signalized intersections. In particular the model is used to assess the effect of intersection design choices on the potential safety and capacity of a left turn movement. The paper shows that at intersections designed according to provincial design standards in Ontario (without fully protected left turn phasing), under certain conditions, do not have adequate left turn sight distance, and that in such cases the safety and capacity of a left turn movement not only depends on the left turn volume and the opposing through volume (as assumed by the Highway Capacity Manual) but also on the opposing left turn volume and the adjacent through volume. Since a capacity guide such as the Highway Capacity Manual and traffic analysis software such as Synchro® do not explicitly account for the effects of intersection geometry and limited sight distances, and may overestimate the capacity of a left turn movement. Measures to improve left turn safety are identified. These measures include providing protected left turn phases, placing detector loops closer to the stop line, decreasing the negative offset between opposing left turn lanes and introducing a shadow lane between a left turn lane and the adjacent through lane. Opportunities for enhancements to existing intersection and signal

not only for left turn vehicles but for all vehicles using an intersection. Finally enhancements to current geometric design standards for signalized intersections will be recommended. Introduction To efficiently accommodate left turn vehicles at signalized intersections a trade off is required between reducing delay and reducing collisions. Fredericton. Des occasions pour le perfectionnement des directives existantes de conception d'intersection et de signal sont identifiées et des recommandations sont faites pour plus de recherche et de dévelopment. Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. that provide inadequate sight distance. operating speeds. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. on safety and capacity for different signal phasing and timing designs. signal phasing and timing design. and different traffic volume combinations. intersection geometry. This paper will assess how well the proposed designs in these Guidelines perform with respect to providing left turn sight distance to vehicles whose sight lines are restricted by vehicles in the opposing left turn lane. For guidance on how to design intersections to accommodate left turn vehicles traffic engineers rely on intersection design guidelines such as TAC’s “Manual for the Geometric Design Standards for Canadian Roads” or Provincial guidelines such as MTO’s “Geometric Design Standards for Ontario Roads”. 2005. June 5-8. 5-8 juin 2005. A quantitative model will be developed to assess the potential impact of designs. Fredericton. The model will be used to justify a hierarchy of strategies to improve left turn safety. for a range of operating conditions.2 virage à gauche et la ruelle traversante adjacente. The safety and operational performance of left turn traffic at signalized intersections depends on a complex interaction of traffic volumes. NB .

A left turn conflict is assumed to be: • A collision between a permissive left turn vehicle and an opposing through moving vehicle The sudden deceleration of a through moving vehicle to avoid a collision with a permissive left turn vehicle The sudden termination of a left turn maneuver by a permissive left turn vehicle to avoid a collision “Blind” Zone • • SDr SDa “Clear” Zone The following measure of exposure to left turn conflicts will be used to assess the potential effect of traffic volumes. signal timing and phasing design and intersection geometry on left turn related collisions: EL = C= Lu = ru = PSD = 3600 ( Lu ρu PSD PZ ) … [1] C Figure 1a: Illustration of left turn conflict Cycle length (sec) Number of permissive left turns during a cycle Probability that the sight distance is restricted by an opposing left turn vehicle during the permissive phase Probability that the restricted sight distance is less than the required sight distance Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. 2005. Fredericton. AND The available sight distance (SDa) is less than the sight distance required (SDr) for at least the median lane of the opposing approach. AND Any one of the three scenarios in Figure 1b applies to the through lanes of the opposing approach. (a) (b) (c) Figure 1b: Illustration of left turn conflict scenarios It is assumed that a vehicle in the “clear” zone does not present a conflict situation as a rational driver is unlikely to commence a turn while there is a clearly visible vehicle in the “clear” zone. AND The left turn sight distance is restricted by a vehicle in the opposing left turn lane. 5-8 juin 2005. NB . NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. June 5-8.3 Problem Statement Figure 1a and 1b illustrate a typical left turn conflict scenario: OR OR • • • • A driver wants to perform a left turn during a permissive left turn phase. Fredericton.

4 Pz = Probability of either scenario (a) . The Highway Capacity Manual (2000) provides procedures for estimating g and gopp for approaches with leading left turn phases. However. The duration of the permissive left-turn phase for qL is equal to the effective green interval for the opposing through movement (gopp). r’ g’ C Length Opposing Approach q’T q’L T g’q g’u qL qT Study Approach Qa Qg Qu Qr r g C Length gq gu Figure 2: Signal phasing. thereby not allowing for the effect of signal progression and platooning • It does not account for phase end left turn “sneakers” during the intergreen interval • It does not account for drivers selecting inappropriate gaps under perfect sight distance conditions In spite of these shortcomings it is postulated that there will be a strong positive correlation between EL and the actual number of left turn related collisions. It is therefore strongly advised that EL should not be used as the denominator in collision rate calculations. 2005. timing and traffic volume parameters The effective duration of the protected left turn phase for the study and opposing approaches are g and g’ respectively. (b) or (c) in Figure 1b occurring when the restricted sight distance is less that the required sight distance Q’a Q’g Q’u Q’r EL is not a perfect measure of exposure. Both left-turn movements are provided with a protected / permissive phase. Theoretical Framework Equations 2 to 9 are based on procedures detailed in the Highway Capacity Manual (2000). NB . Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. It has the following shortcomings: • PSD and PZ will be determined using only the 85th percentile approach speed and the 85th percentile available sight distance • No allowances are made for trucks • The approach speed is treated as a constant (at the 85th percentile level) and not as a variable • Vehicle arrivals are considered to be completely random. June 5-8. and that any measure that is effective in reducing EL will also be effective in reducing left turn related collisions. Consider a signalized intersection with traffic volumes shown in Figure 2. and that a certain % reduction in EL will translate to the same % reduction in collisions. Fredericton. due to the use of an imperfect measure of exposure it cannot be assumed that the relationship between EL and collisions will be perfectly linear. 5-8 juin 2005. Fredericton. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière.

Qu − g u ( s s − q L ) / 3600 ] … [5] The duration of the effective red interval is: r = C − g − g opp qL = sp = s= g= ss = C= … [6] Left-turn flow (vph) Saturation flow rate for protected phase = 0. Fredericton. NB ... left turns are effectively blocked. 2000): 1 Assuming a platoon ratio and lane utilization factor = 1. the average queue at the end of the red interval is: Qa = q L r / 3600 tc = tf = ss = ( q 'T / 3600 ) exp( −q 'T t c / 3600 ) … [7] 1 − exp( −q 'T t f / 3600 ) Critical gap (4. The value of gu can be determined by: g u = g opp − g q … … [3] [9] The average queue at the end of gq is: Qu = Q g + q L g q / 3600 . [2] The average queue at the end of the protected leftturn phase (g) is: Q g = max[ 0. The portion of permissive green blocked by the clearance of an opposing queue is designated gq. 2000) :1: gq = s= q’T = N’ = tL = (q 'T / N ' )( r + g ) − t L … [8] s − q 'T Saturation flow rate (1900 vph) Opposing through volume (vph) Number of opposing through lanes Lost time(4 sec assumed) . Fredericton. The portion of permissive green during which left turns filter through the opposing flow is designated gu. While the opposing queue clears. gq and gu.. Qa − ( s p − q L ) g / 3600 ] The Highway Capacity Manual (2000) provides additional procedures to estimate gq for approaches with leading left turn phases. Proceedings of the Canadian Multidisciplinary Road Safety Conference XV..5 The period gopp consists of two components. 5-8 juin 2005.2 sec assumed) The value of gq can be determined as follows (HCM. For the study approach. 2005.1 sec assumed) Follow up time (2.95s Saturation flow rate (1900 vph assumed) Effective length of protected left-turn phase (sec) Maximum vehicle departure rate during gu (vph) Cycle length (sec) The value of ss can be determined by (HCM. When the permissive green phase is initiated the opposing queue (q’T) begins to move. June 5-8. [4] The average queue at the end of gu is: Qr = max[ 0. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. Once the opposing queue clears left-turn vehicles filter through an unsaturated opposing flow.

ru1 = 1. NB . and in Case 2 gu < g’u. ru – Probability that the left-turn sight distance is restricted during gu To estimate the probability that during the period gu the left turn sight distance is restricted by opposing left turn vehicles. 5-8 juin 2005. It is therefore of no consequence that the sight distance may be restricted during this period. r’ g’ g’q g’u Figure 5: Case 2 – g’u > gu Case 1 is illustrated in Figure 4 and Case 2 is illustrated in Figure 5. hence the value of gu decreases (Equation 9). qL and g r Q’a g gq gu1 Q’u Q’g gu2 gu gu3 Figure 3 illustrates the effect of introducing protected left turn phases on the value of Lu. Fredericton.6 Lu – Number of permissive left turns The number of permissive left turns during gu is given by: Lu = min[ g u s s . i. In Figure 4 during the period gu1 left-turn vehicles filter through the opposing traffic stream. However in Case 2 no left turns are possible during gu1 as the left turn vehicles are still waiting for the opposing queue to clear. Qu + q L g u ] … [10] 9 g= 0 a d g = 0 n ' 8 q ' L Figure 4: Case 1 – g’u < gu Qa Qg Qu Qr = 2 0v h 5 p 7 6 L u 5 4 g= 1 s c a dg = 1 s c 2 e n ' 2 e 3 2 1 0 5 0 0 7 0 0 9 0 0 1 0 1 0 q ' T 1 0 3 0 1 0 5 0 1 0 7 0 1 0 9 0 2 0 1 0 2 0 3 0 2 0 5 0 2 0 7 0 . however for the whole duration of gu1 the opposing left turn queue is stationary. ru2 = 1 . In Figure 5. by the time that left turn vehicles have an opportunity to start turning. i. 2005.e. during the period gu2 the opposing left turn queue clears as vehicles filter through available gaps in the through traffic (qT). leading to a reduction in Lu (Equation 10). It is assumed that while the opposing left turn queue clears the sight distance will remain restricted for 100 % of the time. In Case 1 gu > g’u. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. i. In Figure 4. As q’T increases the value of gq increases (Equation 8). June 5-8. For the Qa Qg Qu Qr r Q’a g gq gu1 gu gu2 Q’u gu3 Figure 4: Case 1 – g’u < gu r’ g’ Q’g g’q g’u Proceedings of the Canadian Multidisciplinary Road Safety Conference XV.O p sn T r u h V l m ( p ) p o i g ho g o u e vh Figure 3: Relationship between Lu and q’T. Fredericton. two different cases should be considered.e.e. Therefore for Case 1 during gu1 the probability that the sight distance is restricted is equal to 1. ru1 = 0. the opposing left turn queue has already been clearing for a maximum period equal to gu1.

SDr . ru2 = 1. June 5-8. 2 During the period gu3 the probability of one or more opposing left turn vehicles in a queue waiting for suitable gap. q’L = s’s = q' L ] … [13] s' s In reality.Probability that restricted sight distance is less than the required sight distance. Fredericton. SD r = 0. 6 u Case 2: [12] g u 2 = max[ 0. It has been observed that often opposing left turn queues “lock horns” – neither queue is moving because the one queue restricts the left turn sight distance of the other and vice versa. even at low left turn volumes. resulting in ru =1. 2 1 . sight distance along the major road should be provided based on a critical gap approach. even when there are adequate gaps in the opposing traffic streams. and therefore restricting the sight distance can be determined using queuing theory (Taha. i. 0 qL q T g g r C 0 1 0 0 5 0 O p p o s i n g L e f tL TT 0 .Probability of restricted sight distance during g u remaining time to clear (if any) the probability that the sight distance is restricted is also equal to 1.7 r u .278 vG … [15] SDr = v= Required sight distance for left turn from the major road (m) 85th percentile speed on major road (km/h) Opposing left turn flow rate (vph) Maximum vehicle departure rate during g’u (See Equation 9) The probability that the left-turn sight distance will be restricted during gu can be estimated by estimating the weighted average of ru1. 0 .Sight Distance Required Harwood et al. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. ru1 and ru1 over the period gu as follows: ρu = ρu1 g u1 + ρu 2 g u 2 + ρu 3 g u 3 … [14] gu Figure 6 illustrates how the value of ru varies with different values of q’L and qT. due to the unwillingness of drivers to accept available gaps when their left turn sight distances are inadequate. In these cases gu3 would be equal to zero. Case 1: gu2 3600 Q' u = min[ g ' u . PSD . ] … [11] s ' s −q ' L 3600 Q'u − g u1 ] … s ' s −q ' L 1 . 1982) as follows: 100 200 300 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 2900 A p p r o a c h L V o l u m e - q Figure 6: Relationship between ru and q’L and qT ρu 3 = min[ 1. the duration of gu2 will be longer and gu3 will be shorter than those calculated. 2005. Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. 4 0 . It is evident that with very high values of qT opposing left turn vehicles are unable to find suitable gaps and restrict the leftturn sight distance for the whole duration of gu. 8 2 5 2 0 0 0 1 5 = 2 5 0 v p h ' = 8 0 0 v p h = 0 ' = 0 = 5 4 s e c = 1 2 0 s e c r n V o l u m e q ' 0 . 0 .e. Fredericton. 5-8 juin 2005. (1996) suggested that at locations where left turns from the major road are permitted at signalized intersections without a protected turn phase. NB 3000 400 0 .

NB .9 sec … [17] Ya = Vw = Yi = These recommended gap sizes will provide enough time for a left turn driver to decide on a course of action.41 -1.41 -0. SDa .46 0.52 4.17 -2.39 -2. During their study McCoy et al. Analyses were conducted for the 95th. 75th. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière.34 0. 2005.5Lw to determine SDa2 For intersection approaches with a raised median it is desirable that the sum of WLL and -x0 be equal to the width of the median.36 4.21 0.28 -1.97 -1.31 0.53 -0.24 -0.46 -2.Available Sight Distance McCoy et al.74 -0.30 0.75 4.48 -0.45 -0.500 vehicles at 6 intersections.15m assumed) Longitudinal distance from the front of the left-turn vehicle to the driver’s eye.36 -0. 85th.40 0.02 -1.36 0.40 4. Regression analysis of the vehicle positioning data was used to determine the relationship between available sight distance and various intersection design parameters as illustrated in Figure 5. (1999) studied. 5-8 juin 2005.0 sec 6.5 sec 8.66 95 -0. (1. Table 1: Critical Gap values Vehicle Type Passenger Cars Single-unit trucks Combination trucks Number of opposing through lanes 1 Lane 2 Lanes 3 lanes 5.9 sec 7. the position of more than 2.35 -2. Fredericton. using video digital technology.58 -0.07 -0.5m assumed) p= Percentile value WLL = Width of left-turn lane line (shadow lane) WOTLT = Width of opposing left-turn lane WLTL = Width of left-turn lane kip = Constant “i” for p-percentile vehicle position x0 = Negative offset between opposing left turn lanes ΔW = 0. (1996).2 sec 7.8 G= Critical gap size for left-turn from the major road (sec) Ybi = (Ya − Yi )( −V w + k1 pW LL + k 2 p x o + k 3 pWOLTL + ∆W + k k 5 p W LL + k 6 p x o + k 7 p W LtL + k 8 p Table 1 shows the G values recommended by Harwood et al. Fredericton.5 sec 6.5Lw to determine SDa1 and = 1.25 4.5 sec 7.97 Proceedings of the Canadian Multidisciplinary Road Safety Conference XV.54 0. June 5-8.2 sec 8. SD ai = Ya +Ybi … [16] Distance between opposing vehicles Width of design vehicle (2. Table 2: Constant values for sight distance equation Constan t k1p k2p k3p k4p k5p k6p k7p k8p 50 -0.97 -0.39 0. and to perform a left turn movement without impeding opposing through moving vehicles.28 -0.5 sec 6. (1999) devised a procedure to calculate the available sight distance for any intersection based on its geometry.02 -1.05 0. 65th and 50th percentile vehicle positions in order to develop guidelines for a range of sight distance design levels and intersection design parameters.50 -0.62 Percentile Position 65 75 85 -0. They define Available Sight Distance (SDa) as the distance from the left-turn driver’s eye to the point at which his/her line of sight intersects the centreline of the near opposing through lane.20 0.

LW WOLTL 9 Roads” and TAC’s “Manual of Geometric Design Standards for Canadian Roads”. June 5-8. This practice however does not necessarily improve the available sight distance for left turn vehicles. PSD = 1 − Percentile / 100 …. as is the common practice. i.Available Sight Distance (m) WLL 4 0 5 0 6 0 P WLTL 7 e r c 0 e n 8 t i l e 0 ( % ) 9 0 1 0 0 Figure 8: Available sight distance percentiles Figure 7: Intersection design parameters The probability that the required sight distance is less than required (PSD) can be determined from Figure 8 and Equation 18. Figure 8 shows the available sight distance percentiles for a typical signalized intersection of 2lane arterials in York Region.e. PZ . Fredericton.44. through lanes widths = 3. PZ = e λ( SDa 1 SDa 2 + − 2tm ) v v −e − 2λ ( SDr −tm ) … v [19] Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. NB .5 m and shadow lane width = 0. The available sight distance percentile values (SDai) can be estimated by substituting the corresponding percentile constants in Table 2 into Equations 16 and 17. offset = -2 m. 5-8 juin 2005. 2005. From Figure 8 the corresponding percentile value is approximately 56 %.Probability of vehicles in “blind” zones The procedure below estimates the probability of either scenario (a). for a speed of 80 km/h the required sight distance = 133 m (from Equation 15). Therefore PSD = 0. and assumes a displaced exponential distribution in the headways between vehicles during the period gu (as recommended by Troutbeck and Brilon). NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. The benefits of a smaller offset (xo) could be decreased or eliminated by the effect of a smaller distance between opposing vehicles (Ya). Fredericton. 1 1 1 1 1 1 1 9 8 7 6 5 4 3 2 1 0 0 0 0 0 0 0 0 0 0 0 Ya xo SDa . (b) or (c) in Figure 1b occurring. intersection width = 35 m. left turn lane widths = 3 m. [18] Equation 17 and the parameters in Table 2 were derived assuming that left turn vehicles do not enter into the intersection to wait for an available gap. designed according to MTO’s “Geometric Design Standards for Ontario For example.

2005. 1 0 0 PZ . which is used to calculate ss assumes that all available gaps are accepted by turning vehicles. Certain warrants for protected left turn phasing rely on the volume to capacity ratio. For example. RT is related to a driver’s willingness to accept “uncertain” gaps and is likely a function of: • • • A driver’s perception that the next gap in traffic is larger than the critical gap (G) required to perform a safe left turn. Fredericton. Overestimation of the capacity will lead to smaller v/c ratios which could result in protected left turn phases not being implemented where they do have the potential to efficiently improve traffic safety and operations. Fredericton. offset and shadow lane width) and the traffic flows q’L and qT on a left turn movement’s capacity. 0 6 0 .10 λ= tm = SDa1 = SDa2 = SDr = v= N’ = ( q 'T / 3600 N ' ) … [20] 1 − t m ( q 'T / 3600 N ' ) Minimum head way = 3600/1900 = 1. 1 4 0 . 0 2 0 . Figure 9 illustrates the non-linear relationship between PZ and q’T for a number of operating speeds.89 sec. 85th Percentile available sight distance to Lane 1 85th Percentile available sight distance to Lane 2 Sight distance required 85th percentile speed on major road (km/h) Number of opposing through lanes Equation 9. 0 . Equation 22 presents a modified capacity equation that takes into consideration the effect of left turn queues and intersection geometry. 0 4 0 . Equation 32 is therefore likely to overestimate the true capacity of a left turn movement. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. The size of the “blind zone” The time already spend waiting to turn left [21] Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. NB . the maximum number of vehicles that can turn left during a cycle is given by: C L = g (s p ) + g u (ss ) + 2 … CL (mod) = g ( s p ) + PSD ρu g u RT ss + ss ρu g u (1 − PSD ) + ss (1 − … [22] RT = Probability that driver will perform a turn when PSD > 0. 0 1 0 0 03 0 05 0 07 0 09 q 0 T 10 1 ' O 01 0 3 p p o 01 05 s s 01 07 i n 01 0 9 g T 02 0 1 h r o 02 0 3 u g 02 05 h 02 0 7 V o 0 l u 0 m e Figure 9: PZ – Probability of vehicle only in blind zone Left turn capacity The capacity of a movement can be defined as the maximum number of vehicles per hour that can perform that movement given the intersection’s traffic flows and design. 5-8 juin 2005. June 5-8. according to MTO’s “Guidelines for Traffic Control Signal Timing and Capacity Analysis at Signalized Intersections” (1989) a protected left turn phase is warranted if the v/c ratio for a left turn movement is larger than 1. 2000). 1 2 v v = 8 0 k m = 9 / h 0 k m / h 0 . Assuming a maximum of two sneakers per cycle (as recommended by HCM. 0 8 v = 7 0 k m / h 0 . In reality drivers do not accept all available gaps as they are not willing to put themselves in danger when their sight distances are restricted by opposing left-turn vehicles. Capacity estimation procedures in the Highway Capacity Manual (2000). and the Canadian Capacity Guide for Signalized Intersections (1995) do not explicitly consider the potential effect of an intersection’s geometry (primarily intersection width.

NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. reducing approach speeds is difficult to achieve and to sustain. Should this not be feasible measures to reduce Ls and ru should be considered. With signal timing parameters: g = 0. Therefore the greatest improvement in safety will likely be attained by making intersection improvements to reduce PSD and PZ. then from Equation 34 the maximum reduction in capacity is 170 vehicles per hour (assuming RT =0). r = 54 sec and r’= 54 sec and C = 120 sec. EL 8 6 q 'L - O p p o s i n g l e f t .t u r n v o l u m e ( v p h 2 0 5 0 0 7 0 0 9 0 0 1 1 0 01 3 0 01 5 0 01 7 0 01 9 0 0 2 1 0 02 3 0 02 5 0 02 7 0 02 9 0 0 q T - A p p r o a c h V o l u m e ( v p h ) Strategy (a) i. As long as there are left turn vehicles using an intersection Ls and ru can never be zero. unless permissive left turn movements are prohibited completely through the introduction of exclusive left turn phases. Improvement Strategies Left turn safety improvement strategies should aim to reduce the values of Ls. qT and q’L Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. Example Consider a typical signalized intersection: • Designed according to TAC and MTO standards. Fredericton.5 m and shadow lane width = 0. which is about 58 % of the HCM capacity (CL) of 293 vehicles per hour. g’ = 0. 1 2 3 P S D = 0 . 2005. 4 4 qL = 2 5 0 v p h q T '= 8 0 0 v p h 0 5 1 0 0 5 4 0 0 1 2 1 0 2 5 2 0 0 1 a) Reduce the sight distance required (SDr) by reducing the approach speed. providing a protected phase that is “called” by appropriately located detection loops should be considered. intersection width = 35 m. Should this not be feasible. ru. A first consideration should be to provide exclusive protected left turn phases. RT = 0 should be assumed. PSD and PZ. • • For this intersection Figure 10 illustrates that because PSD and PZ > 0 the safety (as measured by EL) of a left turn movement not only depends on the opposing through volume (q’T) but also the adjacent through volume (qT) and the opposing left turn volume (q’L). NB . b) Increase the available sight distance (SDa) by reducing the offset between opposing left ) turn lanes (xo) and increasing the width of the shadow lane (wLL). Figure 10: EL vs. With traffic volumes: qL = 250 vph and q’T = 800 vph.5. June 5-8. 1 6 1 4 P Z = 0 .e. due to increases in delay and other collision types. through lanes widths = 3. Through appropriate intersection design the values of PSD and PZ can be reduced to zero.11 The potential reduction in capacity as a result of restricted sight distances is given by: ∆C = g u s s ρu PSD ( RT −1) … [23] If EL = 13. Improving intersection sight distance A combination of two strategies can be employed to reduce PSD and PZ: To calculate the minimum value of CL(mod) and a maximum value for ΔC. with the following geometry: Offset = -2 m. Fredericton. left turn lane widths = 3 m. 5-8 juin 2005.

60 -1.35 0. Speed (km/h) 40 Proceedings of the Canadian Multidisciplinary Road Safety Conference XV.45 0.Available Sight Distance (m) 6 0 0 5 0 0 W L L 4 0 0 = 0 . n 7 d 5 xm 1 0 0 W L L = 0 m 0 = a n2 d m x 0 4 0 5 0 6 0 P e 7 r c 0 e n 8 t i l e 0 ( % ) 9 0 1 0 0 Figure 11: Available sight distance 0. n 2 d 5 xm SDa .5 m. as this may require extensive intersection reconstruction work. introducing shadow lanes and reducing the negative offset between opposing left turn lanes.70 -1. i. NB . NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. 5 0 0 m= 1a .45 -1.6 0.45 The total amount that an intersection has to be widened by (compared to the standard design) to achieve these sight distance improvements is equal to WLL. a shadow lane 0.5 m Lane 3.55 -1.8 0.55 -1. nor feasible. 8 0 0 50 60 70 80 90 Speed (km/h) 40 50 60 70 80 90 Speed (km/h) 40 50 60 70 80 90 7 0 0 W L L = 0 .70 -1.20 -1.8 0. intersection widths and lane widths that will ensure that the 85th percentile available sight distance is equal or greater than the required sight distance.e. Fredericton.70 0. As speeds increase and intersection widths become narrower the need for wider shadow lanes and smaller negative offsets becomes more urgent.8 0.2 -1. 7 5 0 m = 1a .4 0.55 0. 5-8 juin 2005.55 0.25 -1.12 Figure 11 illustrates the dramatic improvements in sight distance that can be achieved with Strategy (b) i. It is however possible to introduce a shadow lane by reducing the widths of the through lanes.35 0. Should this be reduced 3.30 -1.40 -1.6 0.5 0.15. Figure 12 illustrates the improvement in sight distance that this improvement could achieve. on a 2 lane roadway. 2 5 0 m = 1a .3 0.65 -1.45 -1. Table 3 provides values for WLL and x0 for variety of approach speeds. Table 3: Recommended Intersection Design Parameters Intersection Width = 30 m 3.55 0.45 -1. Fredericton. In York Region the standard lane width on arterials is 3. June 5-8. It is evident that the current design standard of W LL = 0 and x0 = -2 m only provides adequate sight distance (at the 85th percentile level) at low approach speeds and very wide intersections.45 -1.35 m Lane WLL Offset WLL Offset 0 -2 0 -2 0 -2 0 -2 0.35 m Lane WLL Offset WLL Offset 0 -2 0 -2 0 -2 0 -2 0 -2 0 -2 0.55 0.75 0.e.45 0.40 Intersection Width = 40 m 3. to reduce the negative offset.35 m each.55 -1. n 5 d 0 xm 3 0 0 W L L 2 0 0 = 0 .25 -1.45 0.75 0.20 -1. 2005.5 0.30 Intersection Width = 35 m 3. PSD = 0.55 -1.4 -1.65 -1.6 -1.35 m Lane WLL Offset WLL Offset 0 -2 0 -2 At existing intersections it may not always be possible.3 m wide is possible.5 -1.50 -1.5 m Lane 3.5 m Lane 3.

As a result the protected leftturn phases are only “called” when there are 3 or more vehicles in the queue after the red interval. 2005. 5-8 juin 2005. and the position of the detection loops. June 5-8. depending on the probability of each scenario occurring. Table 4: Possible left turn phasing scenarios Scenario g = 0 and g’ = 0 g > 0 and g’ = 0 g > 0 and g’ > 0 g = 0 and g’ = 0 Condition Qa < x AND Q’a < x’ Qa >=x AND Q’a < x’ Qa >=x AND Q’a >= x’ Qa < x AND Q’a >= x’ Figure 13: EL by left turn phasing scenario (q’L = 150 vph) The actual value for EL for the design hour will be a combination of the EL values for each scenario. 3 0 = a n 2d mx = = 3 5 8 0 m k m / h 1 4 SDa .Available Sight Distance (m) 1 6 0 W L L 1 = 0 0 2 a =n d 2 x m 1 0 g = 0 a n d g ' = 1 2 s e c 1 4 0 S rD = 1 3 3 m EL 8 g 6 1 2 0 = 0 a n d g ' = 0 1 0 0 4 g = 1 2 s e c a n d g ' = 1 2 s e c 8 0 2 g = 1 2 s e c a n d g ' = 0 6 0 4 0 5 0 6 0 P e 7 r c 0 e n 8 t i l e 0 ( % ) 9 0 1 0 0 0 5 0 0 7 0 0 9 0 0 1 1 0 0 1 3 0 0 1 5 0 01 7 0 01 9 0 0 2 1 0 0 2 3 0 0 2 5 0 02 7 0 02 9 0 0 q T - A p p r o a c h V o l u m e ( v p h ) Figure 12: Improvement in sight distance It is evident from Figure 12 that the value of PSD will decrease from 0. There are four possible left turn phasing scenarios as shown in Table 4. If j >= x’ then EL and E’L is estimated assuming g’ = 12 sec i= Number of vehicles in left turn queue after red interval Number of vehicles in opposing left turn queue after red interval Figure 13 illustrates the reduction in EL that can be achieved by introducing a protected left turn phase.34. j ) … [24] i =0 j =0 sr sr e ( rqL ) (rq L ) i … [25] pi = i! p' j = e ( rq 'L ) ( rq ' L ) j … [26] j! x= x’ = Loop position in approach left turn lane (veh) Loop position in opposing left turn lane (veh) If i >= x then EL and E’L is estimated assuming g = 12 sec. In York Region it is the practice to place the detection loops at the 3rd vehicle position in the left-turn lane. NB . The probability of a scenario appearing during a cycle depends on the duration of the red interval and the left turn flow rate.13 2 0 0 Y v W L L 1 a 6 1 8 0 = 0 . To assess the impact of detection loop positions the following procedure was adopted to calculate EL: E L = ∑∑ pi p' j E L ( i . j= Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. Fredericton.44 to 0. Protected/Permissive Left Turn Phasing Protected phases are “called” by detection loops in the left turn lane. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. Fredericton.

The introduction of protected left turn phases. However.A p p ro a c h vo lu m e (vp h ) Figure 14: Effect of detection loop positions loops on average EL At qT = 1000 vph. 5-8 juin 2005. The position of the detection loops in the leftturn lane influences the likelihood that protected phases will be “called” and therefore has an effect on the value of EL. even when not warranted according to the current guidelines.j) = Exposure to left turn conflicts on study approach where Qa = i and Q’a = j x= Loop position left turn lane x’ = Loop position in the opposing left turn lane opposing left turn volume (q’L) and approach through volume (qT). Figure 14 illustrates the effect of moving the detection loops closer to the left turn lane stop line. changing from a permissive only signal design (no loops) to a protected/permissive design with detection loops at the 3rd vehicle position reduces EL by an average of 63 %. do not provide adequate left turn sight distance (at the 85th percentile level) for all approach speed conditions when there is a vehicle in the opposing left turn lane and exclusive left turn phases are not provided It is likely that the Highway Capacity Manual and traffic analysis software such as Synchro ® overestimate the capacity of a left turn movement as they do not account for drivers not accepting all available gaps as a result of their sight distances being restricted by opposing left turn vehicles.14 pi = Probability of i vehicles in left turn queue after red interval p’j = Probability of j vehicles in opposing left turn queue after red interval EL(i. can have a significant effect in reducing the exposure to left turn collision conflicts. NB . 1 6 q 1 4 L = = = 2 5 0 8 0 0 1 5 0 vp h vp h vp h q 'T q 'L 1 2 1 0 8 6 • N o x = x = x = L o o p s 3 a n d 2 a n d 1 a n d 2 9 0 E L 4 x ' = x ' = x ' = 0 2 3 2 1 0 5 0 0 7 0 0 9 0 0 1 1 0 0 1 3 0 0 1 5 0 0 1 7 0 0 1 9 0 0 2 1 0 0 2 3 0 0 2 5 0 0 2 7 0 0 q T . Compared to placing the loops at the 3rd vehicle position. for signalized intersections. June 5-8. which in turn could cause other collision types to increase. any need to change signal timing and phasing parameters to improve safety. It appears that there may be some justification to place the detection loops at the 2nd vehicle position rather than the 3rd vehicle position. Significant reductions in EL can be achieved by decreasing the negative offset between opposing left-turn lanes and/or increasing the widths of shadow lanes. • Current TAC and provincial design standards in Ontario. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. 2005. In this case a reasonable compromise between safety and delay may be achieved if the loops are placed at the 2nd vehicle position instead. Fredericton. Fredericton. In most cases signal timing and phasing measures to improve left turn safety will have the negative side effect of reducing delay to through moving vehicles. at the expense of overall delay and congestion. Conclusions • At intersections where PSD > 0 the capacity and safety of left turn movements not only depend on the left turn volume (qL) and the opposing through volume (q’T) but also on the • • • Recommendations • Consider incorporating the information in Table 3 into intersection design guidelines such as TAC’s “Manual of Geometric Design Standards for Canadian Roads” and MTO’s Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. can be avoided by taking measures to reduce PSD and PZ instead. placing the loops at the stop line (1st vehicle position) reduces the EL by 17 %.

2nd edition. Unsignalized Intersection Theory.M. M. 1995. Byrd.W. McCoy. Ministry of Transportation. 2000. Institute of Transportation Engineers. New York. National Research Council. NB . Brilon. and Pesti G.tongji. Washington D. Washington D.cn/~yangdy/TrafficFlow/cha p8. Incorporating procedures to estimate changes in delay to left turn traffic and other traffic as a result of changes in left turn phasing.. P... the potential impact of signal timing and phasing design decisions on safety should be considered and evaluated explicitly. 5-8 juin 2005.15 “Geometric Design Standards for Ontario Roads”. Incorporating procedures to perform benefitcost analyses towards achieving an optimal tradeoff between safety and delay. Traffic Signal Timing and Capacity Analysis at Signalized Intersections. W.. Pietruchia.T.C. Mike Horne.L.pdf • • Future Research Future research could aim to improve the usefulness of the model presented in this paper by: • By establishing. Troutbeck R. P. Fredericton.S. Pavement Markings to Improve Opposing Left-Turn Sight Distance. 1985. Ontario (MTO). timing and intersection geometry. Lincoln. http://www. Fredericton. National Research Council.edu. the relationship between EL and left turn collisions. Highway Capacity Manual. R. Ontario (MTO).T. Inc. and Gittings G. Taha. through regression analysis. • As a general design principle intersections should be designed such that at least the 85th percentile available sight distance (when restricted) exceeds the required sight distance. • • Acknowledgements Proceedings of the Canadian Multidisciplinary Road Safety Conference XV.C. th I wish to acknowledge the support of the Regional Municipality of York and in particular the support of my supervisors Zoran Postic and Brian Harrison as well as my colleagues.A. Brydia. Ministry of Transportation. Transportation Association of Canada.. Mason. TRB. Decisions on whether or not to implement protected left turn phasing should not rely on theoretical capacity and delay estimations made by procedures in the Highway Capacity Manual and traffic analysis software such as Synchro® but rather on actual field observations. 1982. Nebraska. 2005.. At intersections where the 85 percentile available sight distance is less than the required sight distance. Operations Research – An Introduction. Canadian Capacity Guide for Signalized Intersections. Duane Carson and Nelson Costa. D. H. Macmillan Publishing Co. Mid-America Transportation Centre. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière.. June 5-8. J. References Harwood. 1996. Manual of Geometric Design Standards for Canadian Roads. . NCHRP Report 383: Intersection Sight Distance. 1989. Geometric Design Standards for Ontario Highways. 1986 Transportation Research Board..J. 1999.E.

5-8 juin 2005. June 5-8. PZb and PZa the following equation can be derived: PZ = e λ( SDa 1 SDa 2 + − 2tm ) v v −e − 2λ ( SDr −tm ) v Proceedings of the Canadian Multidisciplinary Road Safety Conference XV. Fredericton.16 Appendix A With reference to Figure 1b. NB Le compte rendu de la XVe Conférence canadienne multidisciplinaire sur la sécurité routière. Fredericton. NB . (b) and (c) occurring are: PZa = [e − λ PZb = [e − λ PZc = [e − λ ( t a1 − t m ) ( t a 1 −t m ) − e − λ (tr −tm ) ][e − λ (tr −tm ) ] − e − λ ( tr −tm ) ][e − λ (tr −tm ) ] ( t a 2 − tm ) − e − λ (t r − t m ) ][e − λ (t a 2 − t m ) − e − λ ( t r − t m ) ] If ta2 >= tr then ta2 = tr ta1 = SD a1 / v and t a 2 = SD a 2 / v and tr = SDr / v The probability of Scenario (a) or (b) or (c) occurring is given by: PZ = PZa + PZb + PZc Substituting into this equation the equations for PZa. 2005. the probability of Scenario (a).