Traffic Injury Prevention
ISSN: 1538-9588 (Print) 1538-957X (Online) Journal homepage: http://www.tandfonline.com/loi/gcpi20
An empirical bayes safety evaluation of tram/ streetcar signal and lane priority measures in Melbourne Farhana Naznin, Graham Currie, Majid Sarvi & David Logan To cite this article: Farhana Naznin, Graham Currie, Majid Sarvi & David Logan (2015): An empirical bayes safety evaluation of tram/streetcar signal and lane priority measures in Melbourne, Traffic Injury Prevention, DOI: 10.1080/15389588.2015.1035369 To link to this article: http://dx.doi.org/10.1080/15389588.2015.1035369
Accepted author version posted online: 02 Apr 2015. Published online: 02 Apr 2015. Submit your article to this journal
Article views: 46
View related articles
View Crossmark data
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=gcpi20 Download by: [Texas A & M International University]
Date: 06 November 2015, At: 05:34
TRAFFIC INJURY PREVENTION http://dx.doi.org/./..
An empirical bayes safety evaluation of tram/streetcar signal and lane priority measures in Melbourne Farhana Naznina , Graham Curriea , Majid Sarvia , and David Loganb
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
a Institute of Transport Studies, Department of Civil Engineering, Monash University, Clayton, Victoria, Australia; b Monash University Accident Research Centre, Monash University, Clayton, Victoria, Australia
ABSTRACT
ARTICLE HISTORY
Objective: Streetcars/tram systems are growing worldwide, and many are given priority to increase speed and reliability performance in mixed traffic conditions. Research related to the road safety impact of tram priority is limited. This study explores the road safety impacts of tram priority measures including lane and intersection/signal priority measures. Method: A before–after crash study was conducted using the empirical Bayes (EB) method to provide more accurate crash impact estimates by accounting for wider crash trends and regression to the mean effects. Before–after crash data for 29 intersections with tram signal priority and 23 arterials with tram lane priority in Melbourne, Australia, were analyzed to evaluate the road safety impact of tram priority. Results: The EB before–after analysis results indicated a statistically significant adjusted crash reduction rate of 16.4% after implementation of tram priority measures. Signal priority measures were found to reduce crashes by 13.9% and lane priority by 19.4%. A disaggregate level simple before–after analysis indicated reductions in total and serious crashes as well as vehicle-, pedestrian-, and motorcycle-involved crashes. In addition, reductions in on-path crashes, pedestrian-involved crashes, and collisions among vehicles moving in the same and opposite directions and all other specific crash types were found after tram priority implementation. Conclusions: Results suggest that streetcar/tram priority measures result in safety benefits for all road users, including vehicles, pedestrians, and cyclists. Policy implications and areas for future research are discussed.
Received November Accepted March
Introduction Increasing needs for more efficient use of limited roadway space has led to the promotion of preferential treatments for high-occupancy transit vehicles, which is often termed transit priority. The designation of exclusive transit lanes and the provision of transit signal priority measures are 2 major transit priority treatments that have received increasing attention worldwide (Slinn et al. 2005; Smith et al. 2005; Yarra Trams 2010). Trams/streetcars are light rail vehicles operating on tracks located on roads used by general road traffic, often termed a mixed traffic tram operation environment. Mixed traffic road environments decrease running speeds and reduce reliability of schedule adherence for trams. Hence, tram priority measures are often implemented to improve reliability and speed, and a wide range of impacts has been identified (Currie et al. 2012; Yarra Trams 2010; Zhang and Garoni 2013). Trams are often given priority in terms of space and time. Space priority generally involves giving the right-of-way (ROW) to trams, which is termed tram lane priority. At its highest level, they are given exclusive ROW, called Tramway, which is classified as Category B type ROW (Vuchic 1981). General traffic is not permitted along these tramways, but trams have to cross intersections with regular traffic and pedestrians (VicRoads
KEYWORDS
road safety; tram priority; streetcar priority; empirical Bayes method; before–after study; crashes
2012a). At a lower priority level, tram lanes are designed for exclusive use by trams only for a specific time period of the day (e.g., peak only tram lanes) but share with other traffic for the rest of the day. These are classified as Category C type ROW (Vuchic 1981). Other forms of space priority measures include prohibited parking alongside part-time tram lanes (VicRoads 2012b). Tram priority, in terms of time, facilitates the movement of trams through traffic signal–controlled intersections and is termed tram signal priority. The most common signal priority treatments are tram corridor green extension and early green (i.e., crossroad green truncation), and are often practiced internationally; for example, in Toronto, Melbourne, Zurich, and Germany (Association of German Transport 2000; Currie and Shalaby 2008; Nash 2003; Yarra Trams 2010). These types of signal treatments are provided at intersections along tram routes by detecting trams upstream of an intersection. Green extension or early green calls are decided depending upon the current signal light along tram routes. In Melbourne, in addition to green extension and early green, trams are given priority at intersections by banning right turns, introducing hook turns, and inserting clearance phases to clear traffic from tram routes (Currie and Reynolds 2011; Currie and Shalaby 2008; Yarra Trams 2010). A hook turn is a unique approach in signal setting in Melbourne
CONTACT Farhana Naznin [email protected] Institute of Transport Studies, Department of Civil Engineering, Building , Monash University, Clayton, Victoria , Australia. Associate Editor Clay Gabler oversaw the review of this article. Results of this article were presented at the Transportation Research Board th Annual Meeting, . © Taylor & Francis Group, LLC
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
2
F. NAZNIN ET AL.
that relocates opposing turning traffic at intersections (Currie and Reynolds 2011). Trams possess unique characteristics compared to other transit modes regarding design, mass, and operational aspects and have safety concerns (Cheung et al. 2008; Currie and Shalaby 2007; Grzebieta et al. 1999; Vuchic 1981). In addition, the implementation of tram priority measures considerably adjusts the nature of road spaces, yet little research has considered how tram priority measures have acted to impact overall road safety. The few studies that have been undertaken to investigate the safety impacts of tram priority are largely subjective and the outcomes are mixed. Shahla et al. (2009) found that the tram signal priority acted to increase tram involvement and total collisions. Cheung et al. (2008) developed zonal and arterial-level transit involved collision prediction models in Toronto and attempted to examine the impact of a dedicated tram lane on safety. However, they could not conclude with any statistically significant reliable results due to the lack of sample size. At many intersections in Melbourne, trams are provided priority by a unique approach called the hook turn, and Currie and Reynolds (2011) demonstrated that the intersections with hook turns had lower crash rates compared to intersection without hook turns. A more recent and highly relevant study by Richmond et al. (2014) explored the impacts of dedicated streetcar (tram) ROW priority on pedestrian collisions in Toronto using a quasi-experimental design. Overall results suggested a 48% reduction in pedestrian crashes post tram lane implementation. The Richmond et al. (2014) study is one of the first to suggest that tram priority measures in terms of exclusive ROW lanes act to reduce pedestrian crashes. However, this study only considered the physical changes in road environment to explore crash impacts. The effect of changes in pedestrian exposure and traffic volume or speed due to changes in ROW design was not considered, due to unavailability of such data. They also did not consider the expected number of crashes for similar road sections without streetcar ROW in operation (i.e., no control sample was used). Hence, aspects of the research design could be improved to ensure greater accuracy of crash impact estimates. In addition, the Richmond et al. (2014) study was limited to exploring only pedestrian-involved collisions and only explored lane priority for streetcars. The aim of this study is to explore the impact of all forms of priority on all crash types and the empirical Bayes (EB) before–after analysis method is deployed to provide more accurate crash impact estimates. This research is a part of wider international research program “Optimizing the Design and Implementation of Public Transport Priority Initiatives.”1 ” This article is organized as follows: the next section describes the methodology employed followed by the selected priority sites and the crash data sources for this study. The article closes with a discussion of the research findings and conclusions.
Australian Research Council Industry Linkage Program project LP, “Optimizing the Design and Implementation of Public Transport Priority Initiatives,” Institute of Transport Studies, Monash University in association with the Transport Research Group, University of Southampton, UK. The principal chief investigator is Professor Graham Currie, the chief investigator is Associate Professor Majid Sarvi, and the partner investigator is Dr. Nick Hounsell. Farhana Naznin is one of the APA (I) PhD students on the project. The industry sponsors include VicRoads and Public Transport Victoria.
Methodology To evaluate the safety effectiveness of any traffic measure, the most commonly used study design is a before–after study (Elvik 2002; Hauer 1997). Several methods are commonly used for conducting before–after crash studies (Elvik 2002; Hauer 1997; American Association of State Highway and Transportation Officials 2010). However, preference is given to the EB method because it can control for the most important confounding factors (e.g., traffic volume) and account for the regression to the mean (RTM) phenomenon (Hauer 1997; Persaud et al. 2001). RTM is the natural tendency of observed crashes to regress to the mean in the year following an unusually high or low crash counts. This phenomenon is likely to be present when sites are selected for treatments based on their crash records (Gross et al. 2010). The key element of the EB method is to develop a safety performance function (SPF), which is a statistical crash prediction model used to estimate the mean crash frequency for locations (reference group) that are similar to the treatment locations but without treatment in operation. The SPF is used in addition to the observed crash counts at treatment sites to estimate the expected number of crashes and, by doing so, RTM effects are accounted for. Several contributing factors (e.g., geometric, operational variables) can be considered for developing SPF; however, traffic volume has been identified as one the most important factors for crash prediction models (Hadayeghi et al. 2003; Sawalha and Sayed 2001). Because the reference sites were largely similar to treatment sites in terms of geometry (including speed limits), the geometric variables were not considered for developing the SPFs in this study. The functional forms of SPFs are as follows: • For intersections: E(A) = exp (α + β1 ln Q1 + β2 ln Q2 )
(1)
• For road section: E(A) = exp (α + β1 ln L + β2 ln Q)
(2)
where E(A) is the predicted crash count per year; Q1 is the annual average daily traffic (AADT) from major approach of intersections; Q2 is the AADT from minor approach of intersections; Q is the AADT along road sections; L is the length of road sections; α is the model intercept; and β 1 , β 2 are the model coefficients. Traffic count data are considered to be nonnegative, integer, discrete, infrequent, and likely to be overdispersed. The most commonly used crash data modeling approaches are poisson and negative binomial regression models (Washington et al. 2010). In order to select the most appropriate crash prediction model for this study, 4 statistical models—that is, Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial—were tested using SPSS version 21 (IBM Corp. 2012) and STATA version 13 (STATACorp. 2013) statistical software. However, the negative binomial model was found to be suitable for developing safety performance functions for the selected data set of this study and this finding is consistent with previous
TRAFFIC INJURY PREVENTION
3
Table . Summary of tram priority measures for selected tram routes. Priority type Tram route number
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
and
Number of priority positions
Intersection
Lane
Intersection
Route
— —
— —
— —
— —
research (Chengye and Ranjitkar 2013; Goh et al. 2013). Model parameters and overdispersion parameters were estimated using maximum likelihood techniques. The steps for calculating the safety effectiveness of tram priority measures using the EB approach follow the guidelines described in the Highway Safety Manual (American Association of State Highway and Transportation Officials 2010). Finally, the safety effectiveness (θ ) of the tram priority measures is calculated by the equation θ = 100 ∗ (1 − OR) ,
Implementation date September December April February October August May June (signal priority) and August (lane priority)
Figure 1 shows the locations of tram routes in Melbourne with the selected 8 tram routes highlighted on the map. A group of reference sites for both intersections and arterials was selected to develop the SPF. The intersection reference group included intersections with tram lines but no priority for trams. However, the arterial reference group included road sections with tram lines but with trams running in mixed traffic for the entire day. Considering this, 82 intersections and 65 road sections were selected as reference sites for this study.
(3)
where OR is the odds ratio, which represents the unbiased safety effect of tram priority measures and is calculated as the ratio of observed and expected number of crashes in the after period. Positive safety effectiveness for a treatment is obtained when the observed crash counts are less than the expected crash counts during the after period. In this study, a disaggregate-level simple before–after analysis was conducted in addition to the EB approach to investigate the changes in specific crash types following the implementation of tram priority measures by comparing average annual crash counts for both periods. Austroads’ definition of coding accidents, DCA (AustRoads 2009) was used to categorize crash types.
Priority sites selection Melbourne, Australia, has the largest tram/streetcar network in the world by track length—approximately 250 km (155 miles)— and it also has the largest mixed traffic tram operation environment (167 km [104 miles]) (Currie and Shalaby 2007). “Think Tram,” a joint program among the Victorian Department of Infrastructure, VicRoads (the Victorian state road management and construction authority) and Yarra Trams, was introduced in 2003 and included a series of tram improvement initiatives including tram priority treatments (Currie and Shalaby 2007). Twenty-nine 4-leg intersections with tram signal priority and 23 roadway sections with tram lane priority along 8 tram routes in Melbourne were selected for this study to investigate the overall safety impact of tram priority. Signal priority measures include green extension and early green, right turn bans for general traffic, hook turns, and clearance phases. Lane priority includes tramways and tram lanes (part time and full time). Table 1 presents the selected tram priority positions on the selected tram routes along with implementation dates.
Data sources Crash data analyzed for this study were extracted from the public version of CrashStats (VicRoads 2014b), a crash recording system developed by VicRoads and the Victoria Police. It includes graphical and written descriptions of fatal and serious injuries and other injuries related to all transport modes occurring on public roads and reported by police. At each priority location, crash data for 5 years before and 2 years after (except route 59, which only has crash data for one year of the after period) were extracted for analysis. The starting and ending months for both periods were aligned to remove any seasonality effects. For each priority location, data from a buffer period of 3 months after treatment implementation were disregarded to account for any interruption in traffic during the construction period and any adjustment to tram and traffic operations after priority implementation (American Association of State Highway and Transportation Officials 2010). Traffic volume
Figure . Location of selected tram priority routes in Melbourne.
4
F. NAZNIN ET AL.
Table . Descriptive statistics of crash count and AADT for tram priority sites.
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
AADT Treatment sites
Number of locations
Time period
Total number of crashes per year
Minimum
Maximum
Mean
SD
Signal priority
Lane priority
Before After Before After
, , , ,
, , , ,
, , , ,
, , , ,
data were extracted from the VicRoads arterial road traffic volume database (VicRoads 2014a), in the form of AADT. Table 2 shows the descriptive statistics of crash counts and AADT for treatment sites used in this study. One important thing is to note that the value of AADT decreased slightly in the after period.
Results and discussion Table 3 presents the parameter estimates of crash count models for both intersection and arterial reference groups. P values of model parameters are less than .05, which indicates that the parameter estimates are significant at the 95% level. This also shows that the reference sites are large enough to conduct before–after analysis using the EB approach (Park et al. 2010). Table 4 presents the safety effectiveness for tram signal priority, lane priority, and combinations of both priority measures using the EB method. At sites where only tram signal priority treatments were implemented, the OR was 0.861 with a standard error (SE) of 0.082. An odds ratio of 0.806 with a standard error of 0.091 was obtained when only tram lane priority measures were implemented. These odds ratios suggested that the positive safety effectiveness for tram signal priority and lane priority are 13.9 and 19.4%, respectively. When signal and lane priority were considered together, an OR of 0.836 with an SE of 0.061 was recorded, representing a positive safety effectiveness of 16.4%. The safety effectiveness for combined priority and lane priority were statistically significant at the 95% confidence level when t test was applied (American Association of State Highway and Transportation Officials 2010), whereas signal priority was found to be significant at the 90% confidence level. The results suggest that the introduction of tram lane priority yielded better positive safety effectiveness than signal priority. One of the possible reasons is the reduced road space for other
road users due to the introduction of tram lane priority, leading to less frequent lane changing and reduced vehicle speeds. Both factors help to decrease collision risks as well as injury severity in the event of a collision. A recent study by Richmond et al. (2014) on trams in Toronto found that the presence of a dedicated streetcar ROW can reduce the rate of collisions, an outcome in agreement with the present study. The results indicate that crash risk decreased due to tram signal priority in Melbourne, which is contrary to results found by Shahla et al. (2009), who examined transit safety at intersections in Toronto and found that collisions increased at intersections due to the presence of tram priority. In Toronto, tram priority is given by green extension and early green techniques. The green extension can go up to 43% of cycle length, which reduce green time from the cross street and eventually impose frustration to cross road users. In addition, lengthy “Do Not Walk” phases due to long extended green period may lead to pedestrian jay-walking. Shahla et al. (2009) suggested these as possible reasons for safety deterioration at priority intersections. However, in Melbourne, the green extension is shorter than in Toronto (a maximum of 35% of cycle length) and the cycle length is kept fixed (Currie and Shalaby 2008). The SCATS system automatically provides phase compensation for lost green time on crossroads, which can help to reduce frustration and confusion among cross-street users and may act to reduce crash risks. In addition, in Melbourne, a separate “T” light is provided for tram movement, which is another possible reason for reduced crash occurrence, because it creates awareness among other road users that tram is expected to cross the intersection. In addition, hook turns and turn bans in Melbourne represent potential reasons for reduced crash occurrence, as reported by Currie and Reynolds (2011). Moreover, crash exposures at intersections along tram routes are higher for Toronto compared to Melbourne, because the tram headways are less than 5 min in Toronto, whereas in Melbourne typical tram headways are
Table . Results of crash prediction models. Parameters -leg intersections: E(A) = exp (α + β lnQ + β lnQ ) α β β Dispersion parameter, φ Mean Pearson’s chi-square Road segmentsE(A) = exp (α + β lnL + β lnQ) α β β Dispersion parameter, φ Mean Pearson’s chi-square Note. SE = Standard Error
Estimated values (SE)
P value
% Confidence interval
− . (.) . (.) . (.) . (.) .
. . .
− . to −. .–. .–. .–.
− . (.) . (.) . (.) . (.) .
. . .
− . to −. .–. .–. .–.
TRAFFIC INJURY PREVENTION
5
Table . Results of before–after analysis using the EB method. Parameters Number of locations Total observed crash counts in the after period Total expected crash counts in the after period Adjusted OR Standard error of OR OR (% confidence interval) Safety effectiveness
Signal treatment
Type of treatment Lane treatment
Combined effect
. . .–.
. . .–.
. . .–.
+.%
+.%∗
+.%∗
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
Note. ∗ Significant at the % level.
7.5 min. The above-mentioned reasons could be why the introduction of tram signal priority had not led to an increase in crash occurrence in Melbourne. Because the EB before–after analysis outcome indicates that crashes decreased after implementation of both tram signal and lane priority treatments; the results presented in following will show the changes in specific crash types after implementation of both tram signal and lane priority measures combined by comparing average annual crash counts for both before and after periods. Figure 2 show that the overall crashes decreased from 138 per annum (p.a.) to 113 p.a. in the after period (an 18% reduction), and the number of fatal and serious injury (FSI) crashes dropped from 47 to 34 p.a. (a reduction of 28%), and these reductions were found to be significant at the 95% level from a paired sample t test. Vehicle- and pedestrian-involved crashes decreased by 17% (from 93 to 77) and 19% (from 32 to 26), respectively. Car- and motorcycle-involved collisions decreased by 24% (from 188 to 143) and 12% (from 48 to 42), respectively, whereas tram-involved crashes increased from 4 to 8. Although the absolute numbers are low, a possible reason for increased tram-involved crashes is due to higher safety perception by tram drivers, because they expect that other vehicles will not interrupt their progress in exclusive ROW. The consequence is higher tram speeds and a higher chance of collisions. Figure 3 presents a breakdown of total and FSI crashes by specific crash types in an average one-year period before and after
Figure . Changes in crash counts by severity, crash, and vehicle type. ∗ Others: All crashes including striking animals or objects; ∗∗ M/C: Motorcycles, including mopeds and bicycles; ∗∗∗∗ HGV: Heavy goods vehicles, including semitrailers, trucks, (trams), buses, and coaches.
priority implementation according to the DCA. The key observations and possible reasons are explained below. The reduction is higher for on-path crashes (types 160 to 169). The reductions were 64% (11 to 4) and 50% (4 to 2) for total and FSI crashes, respectively. Prohibited parking along the left lane of tram routes is a possible reason for reduced collision incidences between parked vehicles and traffic. Total crashes among vehicles running in opposing directions (types 120 to 129) were reduced by 23% (from 22 to 17). Fatal and serious crashes decreased by 38% (from 8 to 5). Right-through collisions formed the majority of this category. These types of crashes involve vehicles making a right or opposing turn (in Australia) over the tram tracks in front of oncoming traffic from the opposite direction. Due to the introduction of tramways/lanes, vehicles can only make right turns through designated openings along the tram lanes and can reduce exposure to right-through crashes. All types of pedestrian-involved crashes (types 100 to 109) decreased by 13% (from 32 to 28). The introduction of hook turns and right turn bans can act to reduce pedestrian-involved crashes. In conventional traffic opposing turns, crash exposure for pedestrians is higher as vehicles wait for gaps in opposing traffic to complete turns, conflicting with pedestrians crossing roads. Types 130 to 139, which include collisions among vehicles traveling in the same direction, recorded a 50% drop in fatal and serious collisions during the after period. Rear-end, lane change, and side-swipe collisions are major crash types under this category. Dedicated tram lanes can act to reduce lane changing and side-swipe crashes. In addition, in mixed traffic, vehicles following trams in the tram lanes have to stop when trams stop, potentially leading to rear-end collisions. Thus, lane priority for trams can eliminate rear-end collisions of this type. Total crashes of types 140 to 149, which typically involve U-turn and parking maneuver collisions, dropped from 10 to 8 (−20%). Fatal and serious crashes under this category were reduced by 33%. Permitted U-turn locations in designated safe positions along tram priority lanes and parking restriction alongside part-time tram lanes are possible reasons for reduced collisions of these types. It is clear that tram priority results in positive safety benefits to most road users. Major reductions in total and FSI crashes are associated with on-path incident reduction and also among vehicles moving in the same and opposite directions. Pedestrian-involved collisions did not reduce noticeably but the effect was still positive.
F. NAZNIN ET AL.
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
6
Figure . Frequency of crash occurrence by type: (a) frequency of total crash occurrence by type and (b) frequency of fatal and serious crash occurrence by type.
Conclusion This research article investigates the safety effectiveness of tram priority measures using an empirical study of before/after crash rates using the EB methodology to ensure more accurate estimates of crash impacts. Previous research in this field has been largely subjective with mixed outcomes. None has investigated tram signal priority and lane priority together nor explored the
associated factors related to crash frequency changes due to implementation of tram priority schemes. Based on the EB approach, the safety effectiveness for tram signal and lane priority was estimated as 13.9 and 19.4%, respectively. The overall safety effectiveness was 16.4% when lane and signal priority measures were combined. The simple before– after crash comparison identified a reduction in total and serious crashes as well as vehicle-, pedestrian-, and motorcycle-involved
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
TRAFFIC INJURY PREVENTION
crashes. The results also indicate that reductions in on-path collisions and collisions among vehicles traveling in the same and opposing direction were considerably higher than among other specific crash types. The findings of this study present a new and exciting opportunity to evaluate the value of tram priority measures. Road safety benefits have been established for all road users, suggesting that priority measures benefit a far wider section of the community than tram users alone. In addition, the type of benefits conventionally established from priority scheme investment can be increased, from conventional travel time and service reliability benefits to the consideration of road safety benefits to the community as a whole. The results suggest that both transit and road authorities need to consider road safety impacts when planning and evaluating new transit priority schemes. There are numerous opportunities to carry out further research in this field. The authors are implementing a program to explore causal factors in more depth to better understand factors influencing safety benefits such that they can be emphasized in future transit priority design. There is also a need to apply the methods used in this article in a wider range of priority contexts to better understand the overall effects of priority in different cities and types of design.
Acknowledgment The authors thank Yarra Trams and VicRoads for providing information related to tram priority measures implemented in Melbourne.
Funding The authors thank the Australian Research Council for financial support of this research.
References American Association of State Highway and Transportation Officials. Highway Safety Manual. Washington, DC: Author; 2010. Association of German Transport. Light Rail in Germany. Germany: Alba Fachverlag; 2000. AustRoads. Guide to Road Safety Part 8: Treatment of Crash Locations. Sydney, Australia: Author; 2009. Chengye P, Ranjitkar P. Modelling motorway accidents using negative binomial regression. Paper presented at: Eastern Asia Society for Transportation Studies; 2013. Cheung C, Shalaby AS, Persaud BN, Hadayeghi A. Models for safety analysis of road surface transit. Transp Res Rec. 2008;2063:168– 175. Currie G, Goh K, Sarvi M. Exploring the impacts of transit priority measures using automatic vehicle monitoring (AVM) data. Paper presented at: 35th, Australasian Transport Research Forum (ATRF); 2012; Perth, Western Australia, Australia. Currie G, Reynolds J. Managing trams and traffic at intersections with hook turns. Transp Res Rec. 2011;2219:10–19. Currie G, Shalaby A. Success and challenges in modernizing streetcar systems: experiences in Melbourne, Australia, and Toronto, Canada. Transp Res Rec. 2007;2006:31–39.
7
Currie G, Shalaby A. Active transit signal priority for streetcars: experience in Melbourne, Australia, and Toronto, Canada. Transp Res Rec. 2008;2042:41–49. Elvik R. The importance of confounding in observational before-and-after studies of road safety measures. Accid Anal Prev. 2002;34:631–635. Goh K, Currie G, Sarvi M, Logan D. Investigating the road safety impacts of bus rapid transit priority measures. Paper presented at: Transportation Research Board 92nd Annual Meeting; 2013. Gross F, Persaud B, Lyon C. A Guide to Developing Quality Crash Modification Factors. Federal Highway Administration (FHWA); Report No. FHWA-SA-10-032; 2010. Grzebieta R, Rechnitzer G, Daly D, Little P, Enever D. Crash compatibility of trams. Paper presented at: Road Safety Research, Policing, Education Conference; 1999; Canberra, Australia. Hadayeghi A, Shalaby AS, Persaud BN. Macrolevel accident prediction models for evaluating safety of urban transportation systems. Transp Res Rec. 2003;1840:87–95. Hauer E. Observational Before/After Studies in Road Safety. Estimating the Effect of Highway and Traffic Engineering Measures on Road Safety. Emerald Group Publishing Limited; 1997. IBM Corp. IBM SPSS Statistics for Windows. Ver. 21.0. Armonk, NY: Author; 2012. Nash A. Implementing Zurich’s Transit Priority Program. Transp Res Rec. 2003;1835:59–65. Park ES, Park J, Lomax TJ. A fully Bayesian multivariate approach to before– after safety evaluation. Accid Anal Prev. 2010;42:1118–1127. Persaud BN, Retting RA, Garder PE, Lord D. Safety effect of roundabout conversions in the United States: empirical Bayes observational before– after study. Transp Res Rec. 2001;1751:1–8. Richmond SA, Rothman L, Buliung R, Schwartz N, Larsen K, Howard A. Exploring the impact of a dedicated streetcar right-of-way on pedestrian motor vehicle collisions: a quasi experimental design. Accid Anal Prev. 2014;71:222–227. Sawalha Z, Sayed T. Evaluating safety of urban arterial roadways. J Transp Eng. 2001;127(2):151–158. Shahla F, Shalaby AS, Persaud BN, Hadayeghi A. Analysis of transit safety at signalized intersections in Toronto, Ontario, Canada. Transp Res Rec. 2009;2102:108–114. Slinn M, Matthews P, Guest P. Traffic Engineering Design. Principles and Practice. 2nd ed. Butterworth-Heinemann; 2005. Smith HR, Hemily B, Ivanovic M. Transit Signal Priority (TSP): A Planning and Implementation Handbook. Washington, DC: Intelligent Transportation Society of America; 2005. STATACorp. Stata Statistical Software. Release 13. College Station, TX: StataCorp LP; 2013. VicRoads. Sharing the roads. 2012a. Available at: http://www.vicroads. vic.gov.au/Home/SafetyAndRules/SaferRiders/BikeRiders/SharingThe Road.htm. Accessed February 23, 2013. VicRoads. Tram priority & safety. 2012b. Available at: http://www.vicroads. vic.gov.au/Home/Moreinfoandservices/PublicTransport/TramProjec ts/TramPriorityAndSafety.htm. Accessed January 14, 2014. VicRoads. Arterial road traffic volumes. 2014a. Available at: http://www.vicr oads.vic.gov.au/Home/Moreinfoandservices/RoadManagementAnd Design/RoadUseAndPerformance/ArterialRoadTrafficVolumes.htm. Accessed December 22, 2013. VicRoads. VicRoads CrashStats. 2014b. Available at: http://www.vicroads. vic.gov.au/Home/SafetyAndRules/AboutRoadSafety/StatisticsAndRes earch/CrashStats.htm. Accessed November 3, 2013. Vuchic VR. Urban Public Transportation; Systems and Technology. PrenticeHall; 1981. Washington SP, Karlaftis MG, Mannering FL. Statistical and Econometric Methods for Transportation Data Analysis. Boca Raton, FL: CRC Press; 2010. Yarra Trams. Tram Priority—Research, Modelling & Analysis of Tram Priority on the Melbourne Network. Melbourne, Victoria; 2010. Zhang L, Garoni T. A comparison of tram priority at signalized intersections. arXiv preprint arXiv:1311.3590. 2013.
ISSN: 1538-9588 (Print) 1538-957X (Online) Journal homepage: http://www.tandfonline.com/loi/gcpi20
An empirical bayes safety evaluation of tram/ streetcar signal and lane priority measures in Melbourne Farhana Naznin, Graham Currie, Majid Sarvi & David Logan To cite this article: Farhana Naznin, Graham Currie, Majid Sarvi & David Logan (2015): An empirical bayes safety evaluation of tram/streetcar signal and lane priority measures in Melbourne, Traffic Injury Prevention, DOI: 10.1080/15389588.2015.1035369 To link to this article: http://dx.doi.org/10.1080/15389588.2015.1035369
Accepted author version posted online: 02 Apr 2015. Published online: 02 Apr 2015. Submit your article to this journal
Article views: 46
View related articles
View Crossmark data
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=gcpi20 Download by: [Texas A & M International University]
Date: 06 November 2015, At: 05:34
TRAFFIC INJURY PREVENTION http://dx.doi.org/./..
An empirical bayes safety evaluation of tram/streetcar signal and lane priority measures in Melbourne Farhana Naznina , Graham Curriea , Majid Sarvia , and David Loganb
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
a Institute of Transport Studies, Department of Civil Engineering, Monash University, Clayton, Victoria, Australia; b Monash University Accident Research Centre, Monash University, Clayton, Victoria, Australia
ABSTRACT
ARTICLE HISTORY
Objective: Streetcars/tram systems are growing worldwide, and many are given priority to increase speed and reliability performance in mixed traffic conditions. Research related to the road safety impact of tram priority is limited. This study explores the road safety impacts of tram priority measures including lane and intersection/signal priority measures. Method: A before–after crash study was conducted using the empirical Bayes (EB) method to provide more accurate crash impact estimates by accounting for wider crash trends and regression to the mean effects. Before–after crash data for 29 intersections with tram signal priority and 23 arterials with tram lane priority in Melbourne, Australia, were analyzed to evaluate the road safety impact of tram priority. Results: The EB before–after analysis results indicated a statistically significant adjusted crash reduction rate of 16.4% after implementation of tram priority measures. Signal priority measures were found to reduce crashes by 13.9% and lane priority by 19.4%. A disaggregate level simple before–after analysis indicated reductions in total and serious crashes as well as vehicle-, pedestrian-, and motorcycle-involved crashes. In addition, reductions in on-path crashes, pedestrian-involved crashes, and collisions among vehicles moving in the same and opposite directions and all other specific crash types were found after tram priority implementation. Conclusions: Results suggest that streetcar/tram priority measures result in safety benefits for all road users, including vehicles, pedestrians, and cyclists. Policy implications and areas for future research are discussed.
Received November Accepted March
Introduction Increasing needs for more efficient use of limited roadway space has led to the promotion of preferential treatments for high-occupancy transit vehicles, which is often termed transit priority. The designation of exclusive transit lanes and the provision of transit signal priority measures are 2 major transit priority treatments that have received increasing attention worldwide (Slinn et al. 2005; Smith et al. 2005; Yarra Trams 2010). Trams/streetcars are light rail vehicles operating on tracks located on roads used by general road traffic, often termed a mixed traffic tram operation environment. Mixed traffic road environments decrease running speeds and reduce reliability of schedule adherence for trams. Hence, tram priority measures are often implemented to improve reliability and speed, and a wide range of impacts has been identified (Currie et al. 2012; Yarra Trams 2010; Zhang and Garoni 2013). Trams are often given priority in terms of space and time. Space priority generally involves giving the right-of-way (ROW) to trams, which is termed tram lane priority. At its highest level, they are given exclusive ROW, called Tramway, which is classified as Category B type ROW (Vuchic 1981). General traffic is not permitted along these tramways, but trams have to cross intersections with regular traffic and pedestrians (VicRoads
KEYWORDS
road safety; tram priority; streetcar priority; empirical Bayes method; before–after study; crashes
2012a). At a lower priority level, tram lanes are designed for exclusive use by trams only for a specific time period of the day (e.g., peak only tram lanes) but share with other traffic for the rest of the day. These are classified as Category C type ROW (Vuchic 1981). Other forms of space priority measures include prohibited parking alongside part-time tram lanes (VicRoads 2012b). Tram priority, in terms of time, facilitates the movement of trams through traffic signal–controlled intersections and is termed tram signal priority. The most common signal priority treatments are tram corridor green extension and early green (i.e., crossroad green truncation), and are often practiced internationally; for example, in Toronto, Melbourne, Zurich, and Germany (Association of German Transport 2000; Currie and Shalaby 2008; Nash 2003; Yarra Trams 2010). These types of signal treatments are provided at intersections along tram routes by detecting trams upstream of an intersection. Green extension or early green calls are decided depending upon the current signal light along tram routes. In Melbourne, in addition to green extension and early green, trams are given priority at intersections by banning right turns, introducing hook turns, and inserting clearance phases to clear traffic from tram routes (Currie and Reynolds 2011; Currie and Shalaby 2008; Yarra Trams 2010). A hook turn is a unique approach in signal setting in Melbourne
CONTACT Farhana Naznin [email protected] Institute of Transport Studies, Department of Civil Engineering, Building , Monash University, Clayton, Victoria , Australia. Associate Editor Clay Gabler oversaw the review of this article. Results of this article were presented at the Transportation Research Board th Annual Meeting, . © Taylor & Francis Group, LLC
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
2
F. NAZNIN ET AL.
that relocates opposing turning traffic at intersections (Currie and Reynolds 2011). Trams possess unique characteristics compared to other transit modes regarding design, mass, and operational aspects and have safety concerns (Cheung et al. 2008; Currie and Shalaby 2007; Grzebieta et al. 1999; Vuchic 1981). In addition, the implementation of tram priority measures considerably adjusts the nature of road spaces, yet little research has considered how tram priority measures have acted to impact overall road safety. The few studies that have been undertaken to investigate the safety impacts of tram priority are largely subjective and the outcomes are mixed. Shahla et al. (2009) found that the tram signal priority acted to increase tram involvement and total collisions. Cheung et al. (2008) developed zonal and arterial-level transit involved collision prediction models in Toronto and attempted to examine the impact of a dedicated tram lane on safety. However, they could not conclude with any statistically significant reliable results due to the lack of sample size. At many intersections in Melbourne, trams are provided priority by a unique approach called the hook turn, and Currie and Reynolds (2011) demonstrated that the intersections with hook turns had lower crash rates compared to intersection without hook turns. A more recent and highly relevant study by Richmond et al. (2014) explored the impacts of dedicated streetcar (tram) ROW priority on pedestrian collisions in Toronto using a quasi-experimental design. Overall results suggested a 48% reduction in pedestrian crashes post tram lane implementation. The Richmond et al. (2014) study is one of the first to suggest that tram priority measures in terms of exclusive ROW lanes act to reduce pedestrian crashes. However, this study only considered the physical changes in road environment to explore crash impacts. The effect of changes in pedestrian exposure and traffic volume or speed due to changes in ROW design was not considered, due to unavailability of such data. They also did not consider the expected number of crashes for similar road sections without streetcar ROW in operation (i.e., no control sample was used). Hence, aspects of the research design could be improved to ensure greater accuracy of crash impact estimates. In addition, the Richmond et al. (2014) study was limited to exploring only pedestrian-involved collisions and only explored lane priority for streetcars. The aim of this study is to explore the impact of all forms of priority on all crash types and the empirical Bayes (EB) before–after analysis method is deployed to provide more accurate crash impact estimates. This research is a part of wider international research program “Optimizing the Design and Implementation of Public Transport Priority Initiatives.”1 ” This article is organized as follows: the next section describes the methodology employed followed by the selected priority sites and the crash data sources for this study. The article closes with a discussion of the research findings and conclusions.
Australian Research Council Industry Linkage Program project LP, “Optimizing the Design and Implementation of Public Transport Priority Initiatives,” Institute of Transport Studies, Monash University in association with the Transport Research Group, University of Southampton, UK. The principal chief investigator is Professor Graham Currie, the chief investigator is Associate Professor Majid Sarvi, and the partner investigator is Dr. Nick Hounsell. Farhana Naznin is one of the APA (I) PhD students on the project. The industry sponsors include VicRoads and Public Transport Victoria.
Methodology To evaluate the safety effectiveness of any traffic measure, the most commonly used study design is a before–after study (Elvik 2002; Hauer 1997). Several methods are commonly used for conducting before–after crash studies (Elvik 2002; Hauer 1997; American Association of State Highway and Transportation Officials 2010). However, preference is given to the EB method because it can control for the most important confounding factors (e.g., traffic volume) and account for the regression to the mean (RTM) phenomenon (Hauer 1997; Persaud et al. 2001). RTM is the natural tendency of observed crashes to regress to the mean in the year following an unusually high or low crash counts. This phenomenon is likely to be present when sites are selected for treatments based on their crash records (Gross et al. 2010). The key element of the EB method is to develop a safety performance function (SPF), which is a statistical crash prediction model used to estimate the mean crash frequency for locations (reference group) that are similar to the treatment locations but without treatment in operation. The SPF is used in addition to the observed crash counts at treatment sites to estimate the expected number of crashes and, by doing so, RTM effects are accounted for. Several contributing factors (e.g., geometric, operational variables) can be considered for developing SPF; however, traffic volume has been identified as one the most important factors for crash prediction models (Hadayeghi et al. 2003; Sawalha and Sayed 2001). Because the reference sites were largely similar to treatment sites in terms of geometry (including speed limits), the geometric variables were not considered for developing the SPFs in this study. The functional forms of SPFs are as follows: • For intersections: E(A) = exp (α + β1 ln Q1 + β2 ln Q2 )
(1)
• For road section: E(A) = exp (α + β1 ln L + β2 ln Q)
(2)
where E(A) is the predicted crash count per year; Q1 is the annual average daily traffic (AADT) from major approach of intersections; Q2 is the AADT from minor approach of intersections; Q is the AADT along road sections; L is the length of road sections; α is the model intercept; and β 1 , β 2 are the model coefficients. Traffic count data are considered to be nonnegative, integer, discrete, infrequent, and likely to be overdispersed. The most commonly used crash data modeling approaches are poisson and negative binomial regression models (Washington et al. 2010). In order to select the most appropriate crash prediction model for this study, 4 statistical models—that is, Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial—were tested using SPSS version 21 (IBM Corp. 2012) and STATA version 13 (STATACorp. 2013) statistical software. However, the negative binomial model was found to be suitable for developing safety performance functions for the selected data set of this study and this finding is consistent with previous
TRAFFIC INJURY PREVENTION
3
Table . Summary of tram priority measures for selected tram routes. Priority type Tram route number
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
and
Number of priority positions
Intersection
Lane
Intersection
Route
— —
— —
— —
— —
research (Chengye and Ranjitkar 2013; Goh et al. 2013). Model parameters and overdispersion parameters were estimated using maximum likelihood techniques. The steps for calculating the safety effectiveness of tram priority measures using the EB approach follow the guidelines described in the Highway Safety Manual (American Association of State Highway and Transportation Officials 2010). Finally, the safety effectiveness (θ ) of the tram priority measures is calculated by the equation θ = 100 ∗ (1 − OR) ,
Implementation date September December April February October August May June (signal priority) and August (lane priority)
Figure 1 shows the locations of tram routes in Melbourne with the selected 8 tram routes highlighted on the map. A group of reference sites for both intersections and arterials was selected to develop the SPF. The intersection reference group included intersections with tram lines but no priority for trams. However, the arterial reference group included road sections with tram lines but with trams running in mixed traffic for the entire day. Considering this, 82 intersections and 65 road sections were selected as reference sites for this study.
(3)
where OR is the odds ratio, which represents the unbiased safety effect of tram priority measures and is calculated as the ratio of observed and expected number of crashes in the after period. Positive safety effectiveness for a treatment is obtained when the observed crash counts are less than the expected crash counts during the after period. In this study, a disaggregate-level simple before–after analysis was conducted in addition to the EB approach to investigate the changes in specific crash types following the implementation of tram priority measures by comparing average annual crash counts for both periods. Austroads’ definition of coding accidents, DCA (AustRoads 2009) was used to categorize crash types.
Priority sites selection Melbourne, Australia, has the largest tram/streetcar network in the world by track length—approximately 250 km (155 miles)— and it also has the largest mixed traffic tram operation environment (167 km [104 miles]) (Currie and Shalaby 2007). “Think Tram,” a joint program among the Victorian Department of Infrastructure, VicRoads (the Victorian state road management and construction authority) and Yarra Trams, was introduced in 2003 and included a series of tram improvement initiatives including tram priority treatments (Currie and Shalaby 2007). Twenty-nine 4-leg intersections with tram signal priority and 23 roadway sections with tram lane priority along 8 tram routes in Melbourne were selected for this study to investigate the overall safety impact of tram priority. Signal priority measures include green extension and early green, right turn bans for general traffic, hook turns, and clearance phases. Lane priority includes tramways and tram lanes (part time and full time). Table 1 presents the selected tram priority positions on the selected tram routes along with implementation dates.
Data sources Crash data analyzed for this study were extracted from the public version of CrashStats (VicRoads 2014b), a crash recording system developed by VicRoads and the Victoria Police. It includes graphical and written descriptions of fatal and serious injuries and other injuries related to all transport modes occurring on public roads and reported by police. At each priority location, crash data for 5 years before and 2 years after (except route 59, which only has crash data for one year of the after period) were extracted for analysis. The starting and ending months for both periods were aligned to remove any seasonality effects. For each priority location, data from a buffer period of 3 months after treatment implementation were disregarded to account for any interruption in traffic during the construction period and any adjustment to tram and traffic operations after priority implementation (American Association of State Highway and Transportation Officials 2010). Traffic volume
Figure . Location of selected tram priority routes in Melbourne.
4
F. NAZNIN ET AL.
Table . Descriptive statistics of crash count and AADT for tram priority sites.
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
AADT Treatment sites
Number of locations
Time period
Total number of crashes per year
Minimum
Maximum
Mean
SD
Signal priority
Lane priority
Before After Before After
, , , ,
, , , ,
, , , ,
, , , ,
data were extracted from the VicRoads arterial road traffic volume database (VicRoads 2014a), in the form of AADT. Table 2 shows the descriptive statistics of crash counts and AADT for treatment sites used in this study. One important thing is to note that the value of AADT decreased slightly in the after period.
Results and discussion Table 3 presents the parameter estimates of crash count models for both intersection and arterial reference groups. P values of model parameters are less than .05, which indicates that the parameter estimates are significant at the 95% level. This also shows that the reference sites are large enough to conduct before–after analysis using the EB approach (Park et al. 2010). Table 4 presents the safety effectiveness for tram signal priority, lane priority, and combinations of both priority measures using the EB method. At sites where only tram signal priority treatments were implemented, the OR was 0.861 with a standard error (SE) of 0.082. An odds ratio of 0.806 with a standard error of 0.091 was obtained when only tram lane priority measures were implemented. These odds ratios suggested that the positive safety effectiveness for tram signal priority and lane priority are 13.9 and 19.4%, respectively. When signal and lane priority were considered together, an OR of 0.836 with an SE of 0.061 was recorded, representing a positive safety effectiveness of 16.4%. The safety effectiveness for combined priority and lane priority were statistically significant at the 95% confidence level when t test was applied (American Association of State Highway and Transportation Officials 2010), whereas signal priority was found to be significant at the 90% confidence level. The results suggest that the introduction of tram lane priority yielded better positive safety effectiveness than signal priority. One of the possible reasons is the reduced road space for other
road users due to the introduction of tram lane priority, leading to less frequent lane changing and reduced vehicle speeds. Both factors help to decrease collision risks as well as injury severity in the event of a collision. A recent study by Richmond et al. (2014) on trams in Toronto found that the presence of a dedicated streetcar ROW can reduce the rate of collisions, an outcome in agreement with the present study. The results indicate that crash risk decreased due to tram signal priority in Melbourne, which is contrary to results found by Shahla et al. (2009), who examined transit safety at intersections in Toronto and found that collisions increased at intersections due to the presence of tram priority. In Toronto, tram priority is given by green extension and early green techniques. The green extension can go up to 43% of cycle length, which reduce green time from the cross street and eventually impose frustration to cross road users. In addition, lengthy “Do Not Walk” phases due to long extended green period may lead to pedestrian jay-walking. Shahla et al. (2009) suggested these as possible reasons for safety deterioration at priority intersections. However, in Melbourne, the green extension is shorter than in Toronto (a maximum of 35% of cycle length) and the cycle length is kept fixed (Currie and Shalaby 2008). The SCATS system automatically provides phase compensation for lost green time on crossroads, which can help to reduce frustration and confusion among cross-street users and may act to reduce crash risks. In addition, in Melbourne, a separate “T” light is provided for tram movement, which is another possible reason for reduced crash occurrence, because it creates awareness among other road users that tram is expected to cross the intersection. In addition, hook turns and turn bans in Melbourne represent potential reasons for reduced crash occurrence, as reported by Currie and Reynolds (2011). Moreover, crash exposures at intersections along tram routes are higher for Toronto compared to Melbourne, because the tram headways are less than 5 min in Toronto, whereas in Melbourne typical tram headways are
Table . Results of crash prediction models. Parameters -leg intersections: E(A) = exp (α + β lnQ + β lnQ ) α β β Dispersion parameter, φ Mean Pearson’s chi-square Road segmentsE(A) = exp (α + β lnL + β lnQ) α β β Dispersion parameter, φ Mean Pearson’s chi-square Note. SE = Standard Error
Estimated values (SE)
P value
% Confidence interval
− . (.) . (.) . (.) . (.) .
. . .
− . to −. .–. .–. .–.
− . (.) . (.) . (.) . (.) .
. . .
− . to −. .–. .–. .–.
TRAFFIC INJURY PREVENTION
5
Table . Results of before–after analysis using the EB method. Parameters Number of locations Total observed crash counts in the after period Total expected crash counts in the after period Adjusted OR Standard error of OR OR (% confidence interval) Safety effectiveness
Signal treatment
Type of treatment Lane treatment
Combined effect
. . .–.
. . .–.
. . .–.
+.%
+.%∗
+.%∗
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
Note. ∗ Significant at the % level.
7.5 min. The above-mentioned reasons could be why the introduction of tram signal priority had not led to an increase in crash occurrence in Melbourne. Because the EB before–after analysis outcome indicates that crashes decreased after implementation of both tram signal and lane priority treatments; the results presented in following will show the changes in specific crash types after implementation of both tram signal and lane priority measures combined by comparing average annual crash counts for both before and after periods. Figure 2 show that the overall crashes decreased from 138 per annum (p.a.) to 113 p.a. in the after period (an 18% reduction), and the number of fatal and serious injury (FSI) crashes dropped from 47 to 34 p.a. (a reduction of 28%), and these reductions were found to be significant at the 95% level from a paired sample t test. Vehicle- and pedestrian-involved crashes decreased by 17% (from 93 to 77) and 19% (from 32 to 26), respectively. Car- and motorcycle-involved collisions decreased by 24% (from 188 to 143) and 12% (from 48 to 42), respectively, whereas tram-involved crashes increased from 4 to 8. Although the absolute numbers are low, a possible reason for increased tram-involved crashes is due to higher safety perception by tram drivers, because they expect that other vehicles will not interrupt their progress in exclusive ROW. The consequence is higher tram speeds and a higher chance of collisions. Figure 3 presents a breakdown of total and FSI crashes by specific crash types in an average one-year period before and after
Figure . Changes in crash counts by severity, crash, and vehicle type. ∗ Others: All crashes including striking animals or objects; ∗∗ M/C: Motorcycles, including mopeds and bicycles; ∗∗∗∗ HGV: Heavy goods vehicles, including semitrailers, trucks, (trams), buses, and coaches.
priority implementation according to the DCA. The key observations and possible reasons are explained below. The reduction is higher for on-path crashes (types 160 to 169). The reductions were 64% (11 to 4) and 50% (4 to 2) for total and FSI crashes, respectively. Prohibited parking along the left lane of tram routes is a possible reason for reduced collision incidences between parked vehicles and traffic. Total crashes among vehicles running in opposing directions (types 120 to 129) were reduced by 23% (from 22 to 17). Fatal and serious crashes decreased by 38% (from 8 to 5). Right-through collisions formed the majority of this category. These types of crashes involve vehicles making a right or opposing turn (in Australia) over the tram tracks in front of oncoming traffic from the opposite direction. Due to the introduction of tramways/lanes, vehicles can only make right turns through designated openings along the tram lanes and can reduce exposure to right-through crashes. All types of pedestrian-involved crashes (types 100 to 109) decreased by 13% (from 32 to 28). The introduction of hook turns and right turn bans can act to reduce pedestrian-involved crashes. In conventional traffic opposing turns, crash exposure for pedestrians is higher as vehicles wait for gaps in opposing traffic to complete turns, conflicting with pedestrians crossing roads. Types 130 to 139, which include collisions among vehicles traveling in the same direction, recorded a 50% drop in fatal and serious collisions during the after period. Rear-end, lane change, and side-swipe collisions are major crash types under this category. Dedicated tram lanes can act to reduce lane changing and side-swipe crashes. In addition, in mixed traffic, vehicles following trams in the tram lanes have to stop when trams stop, potentially leading to rear-end collisions. Thus, lane priority for trams can eliminate rear-end collisions of this type. Total crashes of types 140 to 149, which typically involve U-turn and parking maneuver collisions, dropped from 10 to 8 (−20%). Fatal and serious crashes under this category were reduced by 33%. Permitted U-turn locations in designated safe positions along tram priority lanes and parking restriction alongside part-time tram lanes are possible reasons for reduced collisions of these types. It is clear that tram priority results in positive safety benefits to most road users. Major reductions in total and FSI crashes are associated with on-path incident reduction and also among vehicles moving in the same and opposite directions. Pedestrian-involved collisions did not reduce noticeably but the effect was still positive.
F. NAZNIN ET AL.
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
6
Figure . Frequency of crash occurrence by type: (a) frequency of total crash occurrence by type and (b) frequency of fatal and serious crash occurrence by type.
Conclusion This research article investigates the safety effectiveness of tram priority measures using an empirical study of before/after crash rates using the EB methodology to ensure more accurate estimates of crash impacts. Previous research in this field has been largely subjective with mixed outcomes. None has investigated tram signal priority and lane priority together nor explored the
associated factors related to crash frequency changes due to implementation of tram priority schemes. Based on the EB approach, the safety effectiveness for tram signal and lane priority was estimated as 13.9 and 19.4%, respectively. The overall safety effectiveness was 16.4% when lane and signal priority measures were combined. The simple before– after crash comparison identified a reduction in total and serious crashes as well as vehicle-, pedestrian-, and motorcycle-involved
Downloaded by [Texas A & M International University] at 05:34 06 November 2015
TRAFFIC INJURY PREVENTION
crashes. The results also indicate that reductions in on-path collisions and collisions among vehicles traveling in the same and opposing direction were considerably higher than among other specific crash types. The findings of this study present a new and exciting opportunity to evaluate the value of tram priority measures. Road safety benefits have been established for all road users, suggesting that priority measures benefit a far wider section of the community than tram users alone. In addition, the type of benefits conventionally established from priority scheme investment can be increased, from conventional travel time and service reliability benefits to the consideration of road safety benefits to the community as a whole. The results suggest that both transit and road authorities need to consider road safety impacts when planning and evaluating new transit priority schemes. There are numerous opportunities to carry out further research in this field. The authors are implementing a program to explore causal factors in more depth to better understand factors influencing safety benefits such that they can be emphasized in future transit priority design. There is also a need to apply the methods used in this article in a wider range of priority contexts to better understand the overall effects of priority in different cities and types of design.
Acknowledgment The authors thank Yarra Trams and VicRoads for providing information related to tram priority measures implemented in Melbourne.
Funding The authors thank the Australian Research Council for financial support of this research.
References American Association of State Highway and Transportation Officials. Highway Safety Manual. Washington, DC: Author; 2010. Association of German Transport. Light Rail in Germany. Germany: Alba Fachverlag; 2000. AustRoads. Guide to Road Safety Part 8: Treatment of Crash Locations. Sydney, Australia: Author; 2009. Chengye P, Ranjitkar P. Modelling motorway accidents using negative binomial regression. Paper presented at: Eastern Asia Society for Transportation Studies; 2013. Cheung C, Shalaby AS, Persaud BN, Hadayeghi A. Models for safety analysis of road surface transit. Transp Res Rec. 2008;2063:168– 175. Currie G, Goh K, Sarvi M. Exploring the impacts of transit priority measures using automatic vehicle monitoring (AVM) data. Paper presented at: 35th, Australasian Transport Research Forum (ATRF); 2012; Perth, Western Australia, Australia. Currie G, Reynolds J. Managing trams and traffic at intersections with hook turns. Transp Res Rec. 2011;2219:10–19. Currie G, Shalaby A. Success and challenges in modernizing streetcar systems: experiences in Melbourne, Australia, and Toronto, Canada. Transp Res Rec. 2007;2006:31–39.
7
Currie G, Shalaby A. Active transit signal priority for streetcars: experience in Melbourne, Australia, and Toronto, Canada. Transp Res Rec. 2008;2042:41–49. Elvik R. The importance of confounding in observational before-and-after studies of road safety measures. Accid Anal Prev. 2002;34:631–635. Goh K, Currie G, Sarvi M, Logan D. Investigating the road safety impacts of bus rapid transit priority measures. Paper presented at: Transportation Research Board 92nd Annual Meeting; 2013. Gross F, Persaud B, Lyon C. A Guide to Developing Quality Crash Modification Factors. Federal Highway Administration (FHWA); Report No. FHWA-SA-10-032; 2010. Grzebieta R, Rechnitzer G, Daly D, Little P, Enever D. Crash compatibility of trams. Paper presented at: Road Safety Research, Policing, Education Conference; 1999; Canberra, Australia. Hadayeghi A, Shalaby AS, Persaud BN. Macrolevel accident prediction models for evaluating safety of urban transportation systems. Transp Res Rec. 2003;1840:87–95. Hauer E. Observational Before/After Studies in Road Safety. Estimating the Effect of Highway and Traffic Engineering Measures on Road Safety. Emerald Group Publishing Limited; 1997. IBM Corp. IBM SPSS Statistics for Windows. Ver. 21.0. Armonk, NY: Author; 2012. Nash A. Implementing Zurich’s Transit Priority Program. Transp Res Rec. 2003;1835:59–65. Park ES, Park J, Lomax TJ. A fully Bayesian multivariate approach to before– after safety evaluation. Accid Anal Prev. 2010;42:1118–1127. Persaud BN, Retting RA, Garder PE, Lord D. Safety effect of roundabout conversions in the United States: empirical Bayes observational before– after study. Transp Res Rec. 2001;1751:1–8. Richmond SA, Rothman L, Buliung R, Schwartz N, Larsen K, Howard A. Exploring the impact of a dedicated streetcar right-of-way on pedestrian motor vehicle collisions: a quasi experimental design. Accid Anal Prev. 2014;71:222–227. Sawalha Z, Sayed T. Evaluating safety of urban arterial roadways. J Transp Eng. 2001;127(2):151–158. Shahla F, Shalaby AS, Persaud BN, Hadayeghi A. Analysis of transit safety at signalized intersections in Toronto, Ontario, Canada. Transp Res Rec. 2009;2102:108–114. Slinn M, Matthews P, Guest P. Traffic Engineering Design. Principles and Practice. 2nd ed. Butterworth-Heinemann; 2005. Smith HR, Hemily B, Ivanovic M. Transit Signal Priority (TSP): A Planning and Implementation Handbook. Washington, DC: Intelligent Transportation Society of America; 2005. STATACorp. Stata Statistical Software. Release 13. College Station, TX: StataCorp LP; 2013. VicRoads. Sharing the roads. 2012a. Available at: http://www.vicroads. vic.gov.au/Home/SafetyAndRules/SaferRiders/BikeRiders/SharingThe Road.htm. Accessed February 23, 2013. VicRoads. Tram priority & safety. 2012b. Available at: http://www.vicroads. vic.gov.au/Home/Moreinfoandservices/PublicTransport/TramProjec ts/TramPriorityAndSafety.htm. Accessed January 14, 2014. VicRoads. Arterial road traffic volumes. 2014a. Available at: http://www.vicr oads.vic.gov.au/Home/Moreinfoandservices/RoadManagementAnd Design/RoadUseAndPerformance/ArterialRoadTrafficVolumes.htm. Accessed December 22, 2013. VicRoads. VicRoads CrashStats. 2014b. Available at: http://www.vicroads. vic.gov.au/Home/SafetyAndRules/AboutRoadSafety/StatisticsAndRes earch/CrashStats.htm. Accessed November 3, 2013. Vuchic VR. Urban Public Transportation; Systems and Technology. PrenticeHall; 1981. Washington SP, Karlaftis MG, Mannering FL. Statistical and Econometric Methods for Transportation Data Analysis. Boca Raton, FL: CRC Press; 2010. Yarra Trams. Tram Priority—Research, Modelling & Analysis of Tram Priority on the Melbourne Network. Melbourne, Victoria; 2010. Zhang L, Garoni T. A comparison of tram priority at signalized intersections. arXiv preprint arXiv:1311.3590. 2013.
