2015jurewiczetal-roadgeometrystudyforimprovedruralsafetyap-t295-15.pdf

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/295010023

Road Geometry Study for Improved Rural Safety Technical Report · June 2015 DOI: 10.13140/RG.2.1.4144.3605

CITATIONS

READS

0

1,528

5 authors, including: Chris Jurewicz

Peter Aumann

Transport Accident Commission

ARRB (Australian Road Research Board)

52 PUBLICATIONS   58 CITATIONS   

1 PUBLICATION   0 CITATIONS   

SEE PROFILE

C- B-w 3 PUBLICATIONS   6 CITATIONS    SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Evaluation of pedestrian crash treatments View project

Development of innovative low-cost Safe System rural intersections View project

All content following this page was uploaded by Chris Jurewicz on 18 February 2016. The user has requested enhancement of the downloaded file.

SEE PROFILE

Technical Report AP-T295-15

Road Geometry Study for Improved Rural Safety

Road Geometry Study for Improved Rural Safety Prepared by

Publisher

Chris Jurewicz, Peter Aumann, Carolyn Bradshaw, Rachel Beesley and Adrian Lim

Austroads Ltd. Level 9, 287 Elizabeth Street Sydney NSW 2000 Australia Phone: +61 2 8265 3000 [email protected] www.austroads.com.au

Project Manager Noel O'Callaghan Abstract

About Austroads

This report draws on literature and crash data analysis to identify and quantify geometric road design elements which contribute to casualty crash occurrence and severity on rural roads, e.g. lack of sealed shoulders, steep downhill grades combined with curves, roadsides with narrow offset to roadside hazards, and high-flow rural at-grade intersections. These findings were supported by a before and after evaluation of casualty crash reductions expected from shoulder sealing, pavement widening and road realignment.

Austroads’ purpose is to:

Combining this evidence and inputs by the Austroads Road Design Task Force, the report proposes a number of possible changes to Austroads road design guides aimed at reducing the casualty crash risk on rural roads. Most proposed changes involve clarification of guidance, e.g. for selection of design speed in challenging alignments, use of speed limits to control speeds, use of sealed shoulders, selection of barriers and clear zones, and greater guidance for design of low speed roundabouts. A Commentary is provided discussing the usefulness of different types of evidence in revision of road engineering guidance. It is intended to make it easier for policy makers to select and commission the most appropriate inputs for consideration. Keywords road design, road geometry, road safety, rural, design guidelines, sealed shoulders, pavement widening

• promote improved Australian and New Zealand transport outcomes

• provide expert technical input to national policy development on road and road transport issues

• promote improved practice and capability by road agencies • promote consistency in road and road agency operations. Austroads membership comprises the six state and two territory road transport and traffic authorities, the Commonwealth Department of Infrastructure and Regional Development, the Australian Local Government Association, and NZ Transport Agency. Austroads is governed by a Board consisting of the chief executive officer (or an alternative senior executive officer) of each of its eleven member organisations:

•

Roads and Maritime Services New South Wales

•

Roads Corporation Victoria

•

Department of Transport and Main Roads Queensland

•

Main Roads Western Australia

•

Department of Planning, Transport and Infrastructure South Australia

•

Department of State Growth Tasmania

ISBN 978-1-925294-42-2

•

Department of Transport Northern Territory

Austroads Project No. TS1543

•

Territory and Municipal Services Directorate, Australian Capital Territory

•

Commonwealth Department of Infrastructure and Regional Development

•

Australian Local Government Association

•

New Zealand Transport Agency.

Austroads Publication No. AP-T295-15 Publication date June 2015 Pages 89

© Austroads 2015 This work is copyright. Apart from any use as permitted under the Copyright Act 1968, no part may be reproduced by any process without the prior written permission of Austroads.

The success of Austroads is derived from the collaboration of member organisations and others in the road industry. It aims to be the Australasian leader in providing high quality information, advice and fostering research in the road transport sector.

Acknowledgements The project authors would like to thank the Road Design Task Force members for their collaboration and input into this project. In particular, thanks are extended to Mr Bill Bui and Mr Richard Fanning of VicRoads, and Mr Mike Whitehead of Department of Transport and Main Roads Queensland for provision of project data. This report has been prepared for Austroads as part of its work to promote improved Australian and New Zealand transport outcomes by providing expert technical input on road and road transport issues. Individual road agencies will determine their response to this report following consideration of their legislative or administrative arrangements, available funding, as well as local circumstances and priorities. Austroads believes this publication to be correct at the time of printing and does not accept responsibility for any consequences arising from the use of information herein. Readers should rely on their own skill and judgement to apply information to particular issues.

Road Geometry Study for Improved Rural Safety

Summary This report identifies and quantifies geometric road design elements which contribute to casualty crash occurrence and severity on rural roads. These findings have been used to identify potential guidance revisions to reduce potential for rural road design contributing to crashes. The report presents a review of Australian and New Zealand rural and remote road casualty crash data, pointing out design areas requiring focussed research: horizontal and vertical alignment, crosssection, roadsides, and rural intersections. Next the report summarises the recent Austroads and international research in these areas, identifying specific road design elements which multiply the risk of rural road casualty crashes, e.g. lack of sealed shoulders, steep downhill grades combined with curves, roadsides with narrow offset to roadside hazards, and high-flow rural at-grade intersections. Published research findings were supported by a before and after evaluation of crash reductions expected from shoulder sealing, pavement widening and road realignment. Combining this evidence and inputs of the Austroads Road Design Task Force, the report proposes a number of possible changes to Austroads road design guides aimed at reducing casualty crash risk on rural roads. Most proposed changes involve clarification of guidance, e.g. for selection of design speed in challenging alignments, use of speed limits to control speeds, use of sealed shoulders, selection of barriers and clear zones, and greater guidance for design of low speed roundabouts. The report makes a number of suggestions about selection of rural intersection types to future revisions of the Traffic Management guides. A number of issues have been identified as requiring further research and/or evaluation before road design guidance can be improved, e.g. application of turbo roundabouts in Australia and New Zealand, and intersection channelisation for left turns. The report also recognises the need to focus on severe crashes in future research and guidance revisions in order to evolve road design guidance towards the Safe System principles. A Commentary is provided discussing the usefulness of different types of evidence in revision of engineering guidance. It is intended to make it easier for policy makers to select and commission the most appropriate inputs for consideration.

Austroads 2015 | page i

Road Geometry Study for Improved Rural Safety

Contents 1.

Introduction.................................................................................................................................... 1 1.1 Background .............................................................................................................................. 1 1.2 Objectives of this Study ........................................................................................................... 1

2.

Methods .......................................................................................................................................... 2 2.1 Crash Data Analysis ................................................................................................................ 2 2.2 Literature Review ..................................................................................................................... 3 2.3 Before and After Evaluation ..................................................................................................... 4 2.3.1

Data Collection ............................................................................................................ 4

2.3.2

Analysis ....................................................................................................................... 5

2.4 Consultation on Design Guidance Changes............................................................................ 5 2.5 Commentary ............................................................................................................................ 5 3.

Rural Casualty Crashes ................................................................................................................ 6 3.1 Extent of the Rural Casualty Crash Problem ........................................................................... 6 3.2 Characteristics of Rural Casualty Crashes .............................................................................. 7 3.3 Rural Casualty Crashes and Geometric Design ...................................................................... 9

4.

Rural Road Design Elements and Crash Risk .......................................................................... 10 4.1 Horizontal Alignment.............................................................................................................. 10 4.2 Vertical Alignment .................................................................................................................. 11 4.3 Cross-section ......................................................................................................................... 11 4.4 Roadsides .............................................................................................................................. 12 4.5 Intersections........................................................................................................................... 13

5.

Evaluation of Selected Road Design Element Improvements ................................................ 16 5.1 Road Design Elements Data ................................................................................................. 16 5.2 Crash Reduction Factors ....................................................................................................... 17 5.3 Crash Severity Ratios ............................................................................................................ 17 5.4 Summary and Interpretation of Results ................................................................................. 18

6.

Design Elements for Improved Rural Road Safety .................................................................. 19 6.1 Horizontal Alignment.............................................................................................................. 19 6.1.1

Design Speed ........................................................................................................... 19

6.1.2

Curvature and Grade ................................................................................................ 19

6.1.3

Curve Delineation and Warning ................................................................................ 19

6.1.4

Perceptual Solutions ................................................................................................. 20

6.2 Vertical Alignment .................................................................................................................. 21 6.3 Cross-section ......................................................................................................................... 21 6.3.1

Sealing Shoulders ..................................................................................................... 21

6.3.2

Providing Unsealed Shoulders ................................................................................. 21

6.3.3

Sealed Shoulder Widening on the Outer Edge of the Curve .................................... 22

6.3.4

Painted Medians ....................................................................................................... 22

6.3.5

Barriers in Painted Medians ...................................................................................... 23

6.3.6

Gateway Treatments................................................................................................. 24

6.4 Roadsides .............................................................................................................................. 24

Austroads 2015 | page ii

Road Geometry Study for Improved Rural Safety

6.5 Intersections........................................................................................................................... 24

7.

6.5.1

General Crash Risk Reduction Principles................................................................. 24

6.5.2

Roundabouts ............................................................................................................. 25

6.5.3

Turbo Roundabouts .................................................................................................. 26

6.5.4

Priority-controlled Intersections ................................................................................ 27

Discussion ................................................................................................................................... 30 7.1 Rural Crash Data Analysis..................................................................................................... 30 7.2 Literature Reviews ................................................................................................................. 30 7.3 Evaluation of Selected Road Design Elements ..................................................................... 31 7.4 Relevance of Findings to Safe System Implementation ........................................................ 31

8.

Conclusions and Proposed Actions.......................................................................................... 32 8.1 Workshop ............................................................................................................................... 32 8.2 Project Findings and Task Force Directions .......................................................................... 32

References ........................................................................................................................................... 35 Appendix A Geometric Design Elements and Crash Risk ............................................................. 40 Appendix B New Road Design Elements ......................................................................................... 60 ..................................................................................................................................... 80

Austroads 2015 | page iii

Road Geometry Study for Improved Rural Safety

Tables Table 2.1: Table 5.1: Table 5.2: Table 5.3: Table 5.4: Table 6.1: Table 6.2: Table 8.1:

RRMA classification system ............................................................................................... 3 Number of treatment and control sites (total length in brackets) ...................................... 16 Before and after casualty crash data (average crash period in years in brackets) .......... 17 Crash reduction factors due to changes in design elements ........................................... 17 Crash severity ratio changes from before to after ............................................................ 18 Example of crash risk-based delineation scheme for rural curves ................................... 20 Section lengths ................................................................................................................. 28 Key crash likelihood and severity factors, and suggested design guidance changes ..... 33

Figures Figure 3.1: The number of rural casualty crashes across Australia and New Zealand between 2005 and 2009 ..................................................................................................... 6 Figure 3.2: Types of casualty crashes on rural roads in Australia and New Zealand (2005 to 2009) .................................................................................................................... 7 Figure 3.3: Midblock casualty crashes on rural roads in Australia and New Zealand (2005 to 2009) ................................................................................................................... 8 Figure 3.4: Intersection casualty crashes on rural roads in Australia and New Zealand (2005 to 2009) .................................................................................................................... 8 Figure 4.1: Flow product range and severe crash relationship for intersection controls at rural cross-intersections ................................................................................................... 13 Figure 4.2: (a) Left-right stagger, and (b) right-left stagger designs ................................................... 15 Figure 6.1: Altered spacing and height of guide posts to make the curve appear more severe ........ 21 Figure 6.2: Wider shoulders on the outside of curves ........................................................................ 22 Figure 6.3: Narrow painted medians deployed in Queensland........................................................... 23 Figure 6.4: A typical 2 + 1 configuration ............................................................................................. 24 Figure 6.5: Basic form of a turbo roundabout ..................................................................................... 27 Figure 6.6: Centre painted median with rumble strips ........................................................................ 28

Austroads 2015 | page iv

Road Geometry Study for Improved Rural Safety

1. Introduction 1.1 Background Road design geometry is a key aspect in the design of safer roads. Analysis of crash data shows that almost 60% of fatal crashes occur on rural roads in Australia, while in New Zealand, the proportion is markedly higher at about 70%. In-depth crash studies have also shown that the road is a causation factor in about 30% of all crashes, while it is known to be a factor in the severity outcome of 100% of crashes. Examination of crash trends has shown that while substantial reductions in fatal crashes have been achieved in recent years on urban roads, such reductions have not been evident on rural roads. A review of road design practice for rural environments would provide an opportunity to significantly improve the safety performance of the rural road network.

1.2 Objectives of this Study The first objective was to identify and quantify geometric road design elements that contribute to crash occurrence and crash severity on rural roads. The second objective was to use these findings to identify solutions and/or future research needs. This information will lead to future revisions of current road design standards to reduce the potential for road design factors to be attributed as a crash factor. For this reason, Austroads Road Design Task Force provided ongoing direction and input into this project.

Austroads 2015 | page 1

Road Geometry Study for Improved Rural Safety

2. Methods The following methods were used to deliver the project objectives.

2.1 Crash Data Analysis Crash data analysis was intended to provide a cross-section of current priority factors associated with rural casualty crashes. The analysis used the methodology in the Guide to Road Safety – Part 5: Road Safety for Rural and Remote Areas (Austroads 2006), but focussed on data from the years 2005 to 2009. The analysis determined key predominating factors associated with rural crashes, while also identifying any changes in the characteristics of heavy vehicle crashes over time. The crash analyses presented in this report sought consistency with the Austroads definition of ‘rural’ and ‘remote’ (displayed in Table 2.1). The definition classes non-urban undivided roads with a speed limit of 80 km/h and all other roads with speed limits greater than 80 km/h as ‘rural’ and ‘remote’. The definition also incorporates rural and remote cities or centres (categories Rural, Remote and Metropolitan Area (RRMA) 3 to RRMA 7). There may be challenges in presenting data from different Australian jurisdictions and New Zealand in terms of the Austroads definition, as each jurisdiction varies with respect to the type and form of crash data collected. It is noted that roads in the Australian Capital Territory (ACT) do not fit the Austroads definition of ‘rural’ or ‘remote’. As such, the crash data contained below does not exclude cases of rural or remote crashes from the ACT. Rather, there were not any rural or remote roads in the ACT. Any of the following analyses that contain data from New Zealand and Tasmania are based on an amendment to the Austroads definition. This was necessary to include crashes on New Zealand and Tasmanian roads in this report. Neither New Zealand, nor Tasmanian data, provided the road classification information needed to ascertain whether or not a crash occurred on a divided or undivided road. As a result, the New Zealand and Tasmanian analyses shown in this report assume that crashes that occurred on 80 km/h roads were rural or remote crashes. Therefore, some caution should be exercised when interpreting and comparing rural and remote crashes between states, territories and New Zealand.

Austroads 2015 | page 2

Road Geometry Study for Improved Rural Safety

Table 2.1:

RRMA classification system

RRMA category

Description of area

Examples

1 – Metro

Capital cities

Canberra ACT, Melbourne VIC, Sydney NSW, Brisbane QLD, Perth WA, Adelaide SA, Hobart TAS, Darwin NT and Wellington New Zealand

2 – Metro

Metropolitan areas / urban with population > 100 000

Geelong VIC, Townsville and the Gold Coast QLD, Newcastle and Wollongong NSW, Auckland New Zealand, Christchurch New Zealand, Hamilton New Zealand and Dunedin New Zealand

3 – Rural

Large rural centre / urban population between 25 000 and 100 000

Mildura and Shepparton VIC, Lismore NSW, Mackay Qld, Launceston TAS, Whyalla SA, Queanbeyan NSW, Rotorua New Zealand and Nelson New Zealand

4 – Rural

Small rural centre / urban population between 10 000 and 25 000

Armidale NSW, Albany WA, Caloundra QLD, Devonport TAS, Mt Gambier SA, Whakatane New Zealand and Ashburton New Zealand

5 – Rural

Other rural area / urban population < 10 000

Swan Hill VIC, Batemans Bay NSW, Ayr QLD, St Helens TAS, Busselton WA, Port Vincent SA, Opotiki New Zealand and Picton New Zealand

6 – Remote

Remote centre / urban population > 5000

Mt Isa QLD, Alice Springs NT, Kalgoorlie WA

7 – Remote

Other remote centre / urban population < 5000

Kununurra WA, Birdsville QLD, Carrieton SA, Strahan TAS, Katherine NT, Murrayville VIC, Karamea New Zealand and Haast New Zealand

Note: For the purposes of this guide, New Zealand cities have been categorised within the RRMA classification system. Source: Austroads (2006).

Data considered rural was then segregated according to available attributes, e.g. crash type, crash location. This permitted discussion by the Task Force and provided direction for further investigation of targeted design elements.

2.2 Literature Review The approach used in this report involves two methods:

•

published and ‘grey’ literature review via library, database and internet searches

•

consultation with practitioners to derive experience within different jurisdictions.

The literature review sought to identify and quantify, where possible, current road design elements that may be over-represented in crash occurrence or crash severity in rural areas, and to identify remedial measures options. The library, database and internet searches were conducted using a range of specific terms related to rural road geometry elements (e.g. sealed shoulders, paved shoulders, hard shoulders) and their effects on crash frequency and severity. Innovative treatments were also pursued in a similar way, seeking their crash reduction effectiveness or proxy measures (e.g. speed).

Austroads 2015 | page 3

Road Geometry Study for Improved Rural Safety

The literature review was conducted using the resources of ARRB Group’s MG Lay Library, the leading land transport library in Australia. These resources included the library’s own comprehensive collection of technical land transport literature and information retrieval specialists with extensive experience in the transport field, as well as access to the collections and expertise of other transport-related libraries throughout Australia and internationally. Used specifically in this literature search were the Australian Transport Index (ATRI) and Transportation Research Information Documentation (TRID) databases, whose content is coordinated by ARRB Group, and the OECD/U.S. Transportation Research Board respectively. Use of these databases ensured wide coverage for quality research material within the subject area from national and international sources. The literature review was used to identify key crash risk factors associated with specific rural road design elements.

2.3 Before and After Evaluation The aim of this part of the study was to add to the literature review by evaluating crash risk reduction potential of prioritised design elements for which data was available.

2.3.1 Data Collection The objective of this task was to gather available information about upgrades of selected rural road design elements and to evaluate their effect on casualty crashes. Evaluation of up to three geometric elements was agreed by the Task Force during Stage 2 of the project. The first step required identification of upgrade works on rural roads completed circa 2008 or after, where clear geometric improvements have been undertaken. The ideal projects for evaluation were to consist of a single design element improvement from substandard to standard level. Information was then sought from jurisdictions, focussing on:

•

what element was improved and the reason for improvement (e.g. type of a funding program)

•

limits of improvements (e.g. road name, chainage, GPS coordinates)

•

the old standard and as-built plans

•

start and end dates of works

•

before and after casualty crash data.

The next step in the process was to collate the available casualty crash performance data and available road design data. There were relatively few projects returned for evaluation as project data was not generally stored centrally. Also the scope of projects varied widely. All projects included multiple element changes. The available data was sorted into like categories (e.g. mainly shoulder sealing, mainly pavement widening). Sites for evaluation were obtained from Queensland and Victoria. Once the design improvement categories were established, control data was collected from surrounding network of similar rural roads where no design improvements have occurred. Crash data was collected for use in the before and after evaluation.

Austroads 2015 | page 4

Road Geometry Study for Improved Rural Safety

2.3.2 Analysis The study method was based on comparison of casualty crash frequencies and severities before and after the upgrades. The works period was excluded. The method was strengthened by the use of nearby control sites where no changes took place. This captured any ‘global’ changes affecting safety, such as increased police enforcement, traffic flow changes and weather patterns. Crash reduction factors and fatal and serious injury (FSI) crash ratios were then calculated, along with an appropriate measure of robustness (p-value below or close to 0.05). Crash reduction factors were calculated for each of the three treatments, adjusting for the expected change in crashes over time based on the control sites. Fisher’s exact test was used to calculate the p-value for the crash reduction factors. A low p-value (p ≤ 0.05) suggests a high degree of confidence that the observed crash change was due to the change in the design element.

2.4 Consultation on Design Guidance Changes This part of the project used the literature review and data analysis to propose a suite of rural road design guidance changes which would reduce the crash risk due to road design. This would mean practice changes to avoid these design parameters that have been shown to be strongly associated with rural casualty crashes. Road Design Task Force was consulted to discuss which proposed changes are realistic and warranted, which require further research, and which are not likely to proceed. Circulation of project outputs in Section 6, and a Task Force workshop to discuss them, identified the potential changes in Section 8.

2.5 Commentary The Commentary was provided to invite discussion and eventually capture the existing state of practice in Austroads guidance review. It is proposed that it may form future informative, high-level guidance for road experts new to the field of guideline revision. The Commentary discusses the general reasons and processes applicable to Austroads guidance review. Importantly, it then focuses on different types of evidence which are used in reviewing technical guidance. The Commentary was developed from a scan of mainstream research methodologies applied by Austroads and individual road agencies to inform reviews of a wide range of guides, supplements and internal policies. This experience was broadened by Task Force member discussions, presentations and working processes of the US Transportation Research Board gathered during a study trip in 2014. [see Commentary 1]

Austroads 2015 | page 5

Road Geometry Study for Improved Rural Safety

3. Rural Casualty Crashes The first objective was to identify and quantify geometric road design elements which contribute to crash occurrence and crash severity on rural roads. The following sections provide a brief analysis of the rural casualty crash problem in Australia and New Zealand based on available data. The focus of the analysis was to determine broad priorities for further investigation of geometric design elements.

3.1 Extent of the Rural Casualty Crash Problem Crash analysis was undertaken as part of a project investigating the relationship between crashes and geometric design elements on rural roads. The analysis was undertaken on crash data for the 2005–09 period for several Australian jurisdictions and New Zealand. Casualty crashes were sourced from rural roads defined as ‘rural and remote’ in Section 2.1. Figure 3.1 shows that approximately 10 400 casualty crashes occurred each year on Australian rural roads, and approximately 4700 on New Zealand rural roads. Analysis also showed that approximately half of the road toll occurred on these road networks. The trend for fatal and casualty crashes showed only a minor sign of decline across the investigated period in Australia. Figure 3.1:

The number of rural casualty crashes across Australia and New Zealand between 2005 and 2009

Fatal crashes accounted for about 5% of all recorded casualty crashes on Australian and New Zealand rural roads. Serious injury crashes made up 51% of rural casualty crashes in Australia, and 23% in New Zealand. The difference may arise from the substantial differences in accounting of serious injury crashes between jurisdictions.

Austroads 2015 | page 6

Road Geometry Study for Improved Rural Safety

3.2 Characteristics of Rural Casualty Crashes Analysis of rural road casualty crash data attributes showed that such crashes occurred:

•

slightly more often in holiday months, December and January

•

more often on Fridays, Saturdays and Sundays

•

in greatest numbers during an extended PM peak between the hours of 14:00 and 18:00

•

mostly at low alcohol risk times

•

mostly in daylight conditions

•

mostly in dry weather conditions

•

mostly in 100 km/h speed zones, although Australia also had significant crash numbers in 80 and 110 km/h speed zones (rural freeways and outback highways, and the default rural speed limit in Western Australia)

•

involving a typical casualty age profile, i.e. showing over-representation of casualties among the 17–25 year-olds for both males and females

•

with most casualties wearing an appropriate restraint (seatbelt/helmet), but data differed significantly between jurisdictions.

The most frequent casualty crash type on rural roads in Australia and New Zealand was the off-path (run-offroad) crash, as shown on Figure 3.2. Off-path crashes formed a similar proportion of rural casualty crashes in Australia (60%) and in New Zealand (55%). The next most significant crash types were:

•

vehicle-opposite (head-on) crashes

•

vehicle-same (rear-end, side-swipe) crashes

•

vehicle-adjacent (intersections) crashes

•

on-path (not leaving the carriageway, e.g. loss-of-control, hitting an animal) crashes.

Miscellaneous crashes not fitting into any of the main groups made up a notable proportion of crashes. Figure 3.2:

Types of casualty crashes on rural roads in Australia and New Zealand (2005 to 2009)

Austroads 2015 | page 7

Road Geometry Study for Improved Rural Safety

Further analysis of casualty crash location attributes indicated that midblock crashes accounted for 80% of the crashes (typically off-path, vehicle-opposite, vehicle-same, on-path, vehicle-overtaking). As shown in Figure 3.3, about half of these crashes occurred on curves. Figure 3.3:

Midblock casualty crashes on rural roads in Australia and New Zealand (2005 to 2009)

For the proportion of crashes that occurred at intersections (vehicle-same, vehicle-adjacent, vehiclemanoeuvring), a significant majority of these were at T-intersections, which tends to be the predominant intersection arrangement found in rural areas. Figure 3.4:

Intersection casualty crashes on rural roads in Australia and New Zealand (2005 to 2009)

Austroads 2015 | page 8

Road Geometry Study for Improved Rural Safety

A sub-sample of heavy vehicle crashes on rural roads was investigated to assess if there were significantly different patterns for this road user group. The proportion of heavy vehicle crashes remained stable across the five-year crash period (about 10% in Australia and 2% in New Zealand). In both countries, the significant majority of rural heavy vehicle casualty crashes also occurred at midblock locations, with less than half on curves. Intersection crashes were under-represented, with the majority occurring at T-intersections. Overall, the heavy vehicle crash patterns did not appear to be very different from that for all vehicles combined.

3.3 Rural Casualty Crashes and Geometric Design From the geometric design perspective, the analysis findings in Section 3.2 show that midblock crashes dominate the rural casualty crash data. In particular, crashes on curves were over-represented in the midblock crash sample, given that this design element constitutes only a relatively small proportion of the overall rural road network 1. The next most concerning design element involved T-intersections due to the concentration of conflicting flows and their proliferation on the rural road network. In summary, the five most common crash types related to the road geometry were: 1. midblock crashes on curves 2. midblock crashes on straight sections 3. vehicles travelling in opposite directions (head-on) 4. vehicles travelling in the same direction (intersections and elsewhere) 5. vehicles travelling in adjacent directions at T-intersections. These observations point to geometric design elements associated with a high incidence of rural casualty crashes:

•

horizontal alignment (points 1 and 3)

•

vertical alignment (points 1, 2 and 3)

•

cross-section, i.e. mid-block design (points 1, 2, 3 and 4)

•

intersection type and design (points 4 and 5).

1

The proportion varies by jurisdiction and road type. For example a recent investigation of a 3000 km sample of rural C roads in Victoria (lowest in the state road hierarchy) revealed that 8% of length was on curves with radius under 1000 m.

Austroads 2015 | page 9

Road Geometry Study for Improved Rural Safety

4. Rural Road Design Elements and Crash Risk Crash trends explored in Section 3 showed that horizontal, vertical, cross-section, and intersection designs have a strong association with the occurrence and severity of crashes on rural roads in Australia and New Zealand. This part of the project sought to identify in more detail which individual geometric design elements contribute strongly to increased casualty crash occurrence or severity. This enabled the project to propose changes to these design elements, which intends to reduce crash risk (Section 6). The geometric road design components for the purpose of this part were divided into:

•

horizontal alignment – circular curves and transition curves, pavement crossfall and superelevation

•

vertical alignment – grades, crest and sag curves, and sight distances

•

cross-section – the number of lanes, lane widths, shoulder widths, verges, batters or embankments, and medians.

•

roadsides – clear zones, roadside slopes, and roadside hazards

•

intersections-comprising type, and orientation of the approach legs.

Austroads (2014a) developed a framework for the assessment of run-off-road crash risk in the Safe System context, given different road design parameters. The framework may be useful in assessment of a combination of horizontal, vertical and cross-section elements on rural road safety outcomes. The following sections provide a summary of the literature review findings. Appendix A provides the detailed findings.

4.1 Horizontal Alignment Horizontal curves have a large impact on the casualty crash risk. Single curves in the 200 m to 600 m range have been found to have the greatest risk. For instance, a sharp curve with a 100 m radius has a 5.5 times higher crash risk than a relatively straight section. Crash severity was found to be somewhat higher on milder curves, however, mainly due to the effect of higher speeds. Crash risk on curves is strongly associated with a high approach speed combined with a large speed change across the curve. For instance, a curve causing a speed reduction of 30 km/h from an approach speed of 100 km/h elevates the risk of a run-off-road casualty crash by 5.1 times. The same speed change but from an approach speed of 60 km/h will increase the risk by 3.1 times (based on Cardoso 2005). This is one of the reasons why out-of-context curves (e.g. isolated, unusually sharp curves) tend to attract run-off-road and head-on crashes. Research by Jurewicz et al. (2014) found that curves to the right are 1.5 times more likely to attract run-offroad casualty crashes than curves to the left. While this is a feature which cannot be designed out on undivided roads, there are options to reduce crash risk by providing wider shoulders and safety barriers on the outside of the curve. An attempt could be made on divided roads to reduce the sharpness of curves to the right, if permitted by the road reserve width. Risk of run-off-road casualty crashes on curves was strongly increased by the effect of the grade, and in particular, of downhill grade (Austroads research on this was summarised in Jurewicz et al. 2014). This aspect is detailed in the next section.

Austroads 2015 | page 10

Road Geometry Study for Improved Rural Safety

It was also found that higher curve frequency was associated with increased risk of run-off-road casualty crashes. A road with 2.5 curves per km had double the crash rate of a straight road. The superelevation deficiencies on a horizontal curve can impact on the crash risk, with the relative risk increasing from 1.06 for deficiencies of 0.02 m/m, to 1.15 for deficiencies of 0.05 m/m. (Harwood et al. 2000). Motorcyclists can be adversely affected by curves, and Schneider et al. (2010) found that an increasing risk was related to the curve length, with a 1% change in curve length increasing the motorcyclist crash frequency by 0.4%.

4.2 Vertical Alignment On vertical grades, casualty crashes increase as the road grade increases, particularly on downhill grades, with a significant increase in the number of crashes and the severity of the crashes when the grade is greater than 6% (Austroads 2014a). Sight distance along vertical curves affects the crash risk and it has been found that where the curve is deficient by greater than 40% compared to the design value, the risk increases by 40%, i.e. a relative risk of 1.40 (Austroads 2010b). Jurewicz et al. (2014) drew on recent Austroads research to suggest a set of run-off-road casualty crash risk values associated with steep grades on rural undivided roads. An uphill grade of 6% increased the risk by 2.6 times compared to the general risk level on a flat road. A downhill grade of the same steepness increased the risk by 5.6 times. Unpublished research by ARRB, based on Victorian road data, highlighted crash risks associated with vertical crests and sags. The relative risk of casualty crashes occurring due to a very sharp sag curve (e.g. radius less than 500 m) was approximately 1.7 as compared with no sag. A similar result was found for crests, but the highest risk category was for crests with vertical radii between 500 and 1300 m. In the cases of both sags and crests, casualty crash risk was elevated by a factor of 2–3 in the presence of a sharp horizontal curve.

4.3 Cross-section Austroads (2011a) showed that the risk of run-off-road casualty crashes was 2.7 times higher on roads with narrow pavements (< 6 m) when compared to roads with 9–10 m pavements. Austroads (2010a, 2010f) provide a relationship between lane width and crash rates, based on Queensland data. The rural results, particularly for two-lane undivided major roads that have a speed limit of 100 km/h, shows that narrow 2.5 m wide lanes have a crash rate that is approximately 50% higher than the wider 3.5 m lanes. Austroads (2014c) draws on previous research in Queensland to identify that rural undivided roads with little or no sealed shoulder (< 0.5 m category) had 1.7 times higher risk of casualty crashes (any type) than roads with 2.0 m sealed shoulders. Safety benefits of sealed shoulders were also evident on rural roads with lower speed limits, e.g. 80 km/h (Austroads 2014c). In a limited study of lane and shoulder widths, Amjadi (2009), found that the allocation of lane and shoulder width on existing pavements can influence the crash rates. The author showed that providing wider lanes has a small advantage over providing wider shoulders.

Austroads 2015 | page 11

Road Geometry Study for Improved Rural Safety

Provision of unsealed shoulders was also shown to provide notable safety benefits. Austroads (2014a) showed that provision of an unsealed shoulder, defined as a 1–2 m wide gravelled surface, was associated with a 50% reduction in run-off-road casualty crash risk. This effect was consistent regardless of the absence or width of a sealed shoulder. Austroads (2010g) showed that casualty crash rates were 1.6 times higher on undivided rural roads than on divided. There are many other differences between the two road stereotypes besides the presence of a median, however, the finding shows that lack of a separation between opposing traffic is a significant risk contributing factor. The finding also showed that severity of crashes on undivided roads was generally higher. This was most likely due to the occurrence of high-speed head-on crashes on undivided roads (very rare on divided roads).

4.4 Roadsides Run-off-road events contribute to about a third of all fatal and serious injury crashes on rural roads (Victorian data based on Jurewicz et al. 2014). Design of road alignment and its delineation play a crucial role in helping drivers to stay on the road. When drivers make errors, or are impacted by events outside of their control (objects/animals on the road, actions of other drivers), design of the roadsides can have a significant contribution to improving safety on rural roads. Geometry of roadsides should be considered as part of any road design. Austroads (2014a) lists a number of roadside design parameters which contribute to increased risk of run-offroad casualty crashes. High speeds are a clear contributor to increased severity of run-off-road crashes. In 80 km/h speed limits, 1 in 25 recorded run-off-road casualty crashes will be fatal. In 110 km/h speed limits, 1 in 15 will result in a fatality. Even where roadsides are clear of conventional hazards, the combination of uneven roadside surface and crash dynamics provide a low chance of safe recovery or stopping. In fact, Austroads (2014a) shows a relationship between reducing crashes with roadside trees and increasing incidence of rollover crashes in wider clear zones. Austroads (2014a) shows limited benefits of providing very wide clear zones on rural roads. Austroads (2014a) and its preceding interim reports provided a list of run-off-road casualty crash risk factors associated with roadside design parameters, as follows:

•

Roads with little or no clear zones (< 2 m) had a 2.2 times higher risk of run-off-road casualty than roads with clear zones > 8 m. Notably, the risk for roads with clear zones in the 4–8 m range was only 1.12 times higher. There were no appreciable changes in risk beyond 8 m. The greatest safety benefits of clear zones were shown to occur in the first 4 m. There was a weak, statistically non-significant link between wider clear zone and higher severity of run-off-road casualty crashes.

•

The run-off-road crash likelihood in the presence of roadside slopes steeper than 1:3.5 was more than double the rate compared to a flat roadside, i.e. slopes1:6 or flatter (Austroads 2011a). Austroads (2014a) shows that the severity of run-off-road casualty crashes into cutting embankments in a 100 km/h speed limit rural road environment was slightly lower than hitting nothing at all or hitting a guardrail (FSI crash ratios of 0.53, 0.55 and 0.60, respectively).

•

Austroads (2014a) showed that increased hazard density (> 50 per 100 m of roadside, or continuous) increased the risk of casualty run-off-road crashes by 1.6 times compared to a density of < 25 per 100 m. The risk level was stable at densities below 25 per 100. Hazard density did not have an appreciable effect on crash severity.

Austroads 2015 | page 12

Road Geometry Study for Improved Rural Safety

Austroads (2014a) highlighted a number of other crash risk factors which are also considered during road design, such as:

•

type of roadside hazards – trees and poles were the most severe (good evidence)

•

type of roadside protection – none, different types of barriers (good evidence)

•

roadside drainage features (good evidence)

•

type of roadside drainage (poor evidence)

•

roadside surfaces and vegetation types (poor evidence)

•

crash attenuators (poor evidence).

4.5 Intersections The form of rural intersection control and traffic flows are the key crash risk factors associated with designing a rural intersection. Figure 4.1 shows the safety performance functions for priority-controlled, signalised and roundabout-controlled cross-roads in New Zealand (severe crashes only). The graphs indicate that prioritycontrolled intersections have the highest risk for the same traffic flow parameters (e.g. product of 1000). Roundabouts are safer than traffic signals. Given that traffic signals are rarely applicable in a rural context, roundabout intersection control should be the first choice when seeking to minimise severe crash risk. Figure 4.1: Flow product range and severe crash relationship for intersection controls at rural crossintersections

Average Qmajor x Average Qminor x 10-4 Source: NZ Transport Agency (2013).

Austroads (2013b) noted that roundabouts frequently have severe crash risk factors leading to right-angle, loss-of-control and rear-end crashes due to:

•

multiple approach and circulating lanes resulting in inadequate approach speed dissipation

•

cyclist and motorcyclist issues relating to lack of visibility (drivers look, but fail to see).

These risk factors could be addressed by several design changes discussed in Section 6.5.2.

Austroads 2015 | page 13

Road Geometry Study for Improved Rural Safety

When roundabouts are not feasible, a number of risk factors can be considered at priority-controlled intersections. The available sight distance at such intersections can be influenced by the form and layout of the intersection. Limited sight distance can be due to the presence of crests or horizontal curves, or roadside infrastructure such as poles and vegetation. Intersecting angles of approaches is an important geometric consideration. Gattis and Low (1998) indicated that compared to 90° intersections, skew angle intersections may cause problems which include:

•

vehicles requiring a longer time to cross, due to the increase in travel distance across the intersection. This may lead to a greater sight distance being required

•

drivers making left or right turns encroaching into the opposing flow lanes

•

drivers not slowing sufficiently to undertake a turn that only requires a small turning angle

•

drivers making right or left turns having more difficulty aligning their vehicles as they turn onto the crossroad

•

drivers’ sight distance being obstructed by other vehicles.

In addition to the angles of the approaches, the alignment of the lanes, both through and turning, can have an impact on rural intersection safety. Misalignment of an intersection was found to be a significant crash risk factor, for similar reasons as above. A purposeful misalignment of approaches has been employed in rural staggered T- intersection design to address high-speed right-angle crashes. This solution has produced evidence of crash reduction benefits of up to 35% under optimal flow conditions (low minor road flows). The design may be executed as a left-thenright-turn sequence (left-right stagger in Figure 4.2a), or as a right-then-left-turn sequence (right-left stagger in Figure 4.2b). A limited study of staggered T-intersections in Victoria (Chia, Jurewicz & Turner 2013) found that crash risk for the right-left stagger design was significantly higher when compared to the left-right-stagger design. This was previously found in a detailed modelling study of Queensland unsignalised intersections by Arndt (2003). The stagger distance between the minor legs suggested in Austroads (2010h) should be at least 15 m for right-left stagger arrangements. The study by Chia, Jurewicz and Turner (2013) found that sites with a shorter stagger distance than 15 m were markedly more likely to have a high annual casualty crash rate. Significant minor road approach curvature and lack of advance warning signs were also found to add to casualty crash risk at these intersections. Superelevation changes through an intersection to accommodate the intersecting road, can affect the lateral acceleration of a vehicle, particularly a heavy vehicle, and lead to increased crash rates for vehicles travelling along the major road (Leonard, Bilse & Recker 1994).

Austroads 2015 | page 14

Road Geometry Study for Improved Rural Safety

Figure 4.2:

(a) Left-right stagger, and (b) right-left stagger designs

(a)

(b) Source: Austroads (2010h).

Austroads 2015 | page 15

Road Geometry Study for Improved Rural Safety

5. Evaluation of Selected Road Design Element Improvements The following stage of the project sought to add to the crash occurrence and severity information relating to design elements collated in Section 4 obtained from literature. This experimental stage was also intended to inform selection of the potential design element solutions in Section 6. The Task Force considered the findings of the literature reviews and availability of road agency data, which led to further analysis of the following design elements:

•

sealed shoulders, through investigation of the effect of their sealing

•

horizontal and vertical alignment, through investigation of the effect of road alignment improvement

•

pavement width, through investigation of the effect of its widening.

The potential for a given road design element to contribute to crash occurrence is often described by its casualty crash reduction factor (CRF) when upgraded from one design category to another. Crash severity reduction can also be measured by comparing before and after FSI crash ratios. Recent road design element improvement sites were nominated by road agencies. These agencies provided basic data such as as-built design plans, scope of works, start and completion dates, and crash data.

5.1 Road Design Elements Data Design element changes were carried out at sites nominated by road agencies in Queensland and Victoria. The numbers of different treatment and matching control sites are shown in Table 5.1. All of the roads in the study were rural undivided roads with a speed limit of 100 km/h. Table 5.1:

Number of treatment and control sites (total length in brackets)

Design element

No. of treatment sites

No. of control sites

Sealed shoulders

12 (93.2 km)

12 (88.2 km)

Horizontal and vertical alignment (improvement)

2 (2.3 km)

2 (3.5 km)

Pavement width (widening)

3 (15.0 km)

3 (14.7 km)

As can be seen in Table 5.1, site nominations provided a more significant number of sites for evaluation of shoulder sealing. All of these sites were from Victoria.

Austroads 2015 | page 16

Road Geometry Study for Improved Rural Safety

Table 5.2:

Before and after casualty crash data (average crash period in years in brackets)

Design element

No. casualty crashes at treatment sites

No. casualty crashes at control sites

Before

After

Before

After

Sealed shoulders

93 (5.0 years)

51 (4.5 years)

57 (5.0 years)

52 (4.5 years)

Horizontal and vertical alignment (improvement)

1 (5.0 years)

0 (4.5 years)

2 (5.0 years)

1 (4.0 years)

Pavement width (widening)

16 (5.0 years)

5 (4.7 years)

6 (5.0 years)

6 (4.7 years)

5.2 Crash Reduction Factors The data described in Section 5.1 was used to calculate crash reduction factors and their statistical robustness. Crash reduction at treatment sites was adjusted for changes at the control sites. Fisher’s exact test was used to calculate the p-value for the crash reduction factors. Low p-value (p ≤ 0.05) suggests a high degree of confidence that the observed crash change was due to the change in the design element. The results are presented in Table 5.3. Table 5.3:

Crash reduction factors due to changes in design elements

Design element

Casualty crash reduction factor (p-value)

Sealed shoulders

41% (0.05)

Horizontal and vertical alignment (improvement)

100% (1.00)

Pavement width (widening)

69% (0.15)

Only shoulder sealing produced a casualty crash reduction that can be considered statistically significant, with a p-value equal to 0.05. Pavement widening produced a strong crash reduction based on three sites with a p-value of 0.15. While this result was not as robust as desired, it may still be considered as indicative. Due to the small number of sites and crashes, the results for horizontal and vertical alignment improvement cannot be considered valid. A comforting observation is that crashes did not dramatically increase after the treatment.

5.3 Crash Severity Ratios Crash severity analysis was carried out to determine the crash severity effects of the treatments. Crash severity was defined as the ratio of fatal plus serious injury crashes to all casualty crashes. The results are presented in Table 5.4. It shows that for sealed shoulders, the severity at the treatment sites reduced in the after period, in comparison with the before. However, a stronger effect in the same direction was observed at the control sites. This difference was tested for significance using the z-ratio test. The pvalue was 0.38, indicating that the difference was not close to being statistically significant.

Austroads 2015 | page 17

Road Geometry Study for Improved Rural Safety

Table 5.4:

Crash severity ratio changes from before to after

Design element

Crash severity ratio at treatment sites

Crash severity ratio at control sites

Before

After

Before

After

Sealed shoulders

0.54

0.47

0.54

0.38

Horizontal and vertical alignment (improvement)

1.00

Zero crashes

0.50

0.00

Pavement width (widening)

0.38

0.40

0.50

0.50

There was insufficient fatal and serious injury crash data to make a statistically meaningful comparison for the other two design element changes.

5.4 Summary and Interpretation of Results The results in Table 5.3 showed that provision of sealed shoulders produced a statistically robust casualty crash reduction on 41%. This means that casualty crash risk is 1.7 times higher on a rural road without sealed shoulders than on a road with sealed shoulders. This result based on Victorian sites exactly matches the findings from the Austroads research based on Queensland data in Section 4.3. Road widening provided an even stronger, but only indicative, casualty crash reduction factor of 69%. This means that roads with narrower pavements had a casualty crash risk 3.2 higher. This compares favourably with the relative risk of run-off-road casualty crashes of 2.7 for a narrow pavement as quoted in Section 4.3. Improvement of horizontal and vertical alignment did not provide meaningful crash reduction findings due to a data sample available for the study. Also, crash severity effects of these treatments could not be established due to insufficient data. The relevance, applicability and robustness of these findings for revision of road design guidance are discussed in Section 6.

Austroads 2015 | page 18

Road Geometry Study for Improved Rural Safety

6. Design Elements for Improved Rural Road Safety The second objective of this project was to use the findings of research in the previous sections to identify geometric design solutions and/or future research needs. This information will lead to future revisions of the current Austroads road design guide series to reduce the potential for road design factors to be attributed as a crash factor. This section also lists solutions that are fully or partially covered in other Austroads guide series, e.g. Traffic Management or Road Safety, and in Australian Standards. These solutions were included as they support safety effectiveness of geometric road design, e.g. through line marking or choice of intersection control type. Some of the proposed solutions are proven road safety treatments, while others are emerging solutions which had a significant level of investigation and trialling.

6.1 Horizontal Alignment Section 4.1 provided a review of evidence related to horizontal alignment design factors associated with an increased incidence of casualty crashes. The following sections propose several guidance-related solutions.

6.1.1 Design Speed Design speed selection guidance already contains consideration of out-of-context curves and other lowspeed elements on an otherwise high-speed road. A strengthening of elements of this guidance could be considered to avoid surprising horizontal alignment features. Consideration of the expected speed limit should also be made in the Road Design series. The issue is being tackled via Traffic Management and Road Safety guides concerned with speed limit setting. Roads with a significant number of curves will perform much more safely with lower speed limits. Lowering the assumed speed limit for roads with curves will influence the design speed.

6.1.2 Curvature and Grade Revision of Austroads (2010d) could consider stronger discouragement of combining sharp curvature (< 600 m) with grades steeper than 6%.

6.1.3 Curve Delineation and Warning Simplify curve delineation and signing schemes according to run-off-road crash risk, rather than being based an individual proxy measures such as curve radius and length. Jurewicz et al. (2014) provides a review of similar schemes from overseas, and outlines an approach adopted in Victoria as a safety measure. Table 6.1 presents the graduated application of delineation. Casualty crash reduction effectiveness of the full high-risk curve package was estimated at 57% for previously untreated curves.

Austroads 2015 | page 19

Road Geometry Study for Improved Rural Safety

Table 6.1:

Example of crash risk-based delineation scheme for rural curves

Low-risk curves

Medium-risk curves

High-risk curves

• •

• • •

Guideposts Edge line (only if pavement width allows) Centreline (only if pavement width allows

• • •

Guideposts Edge line (only if pavement width allows) Centreline (only if pavement width allows)

•

RRPM (only if line marking exists or possible Audio-tactile (only when there is evidence of fatigue related crashes Curve warning signs for isolated or group curves - One side of the road - Both sides of the road - Larger signs

•

RRPM (only if line marking exist or possible) Audio-tactile (only when there is evidence of fatigue related crashes) Curve warning signs for isolated or group of curves - One side of the road - Both sides of the road - Larger signs

•

Guideposts Edge line (only if pavement width allows) Centreline (only if pavement width allows)

• •

Or

• •

Or

•

CAMs

•

Advisory speed signs

Source: Jurewicz, Lim and Phillips (2013).

6.1.4 Perceptual Solutions There is now adequate evidence to allow consideration of perceptual solutions aimed at the control of operating speed on the approach to and through horizontal curves (Austroads 2014d). The main solutions with known effectiveness are listed here:

•

delineation and signage creating the perception of a tighter curve (–10 km/h), see Figure 6.1

•

vehicle-activated signs warning of excessive speed (–6 km/h)

•

transverse rumble strips (–5 km/h)

•

chevron alignment markers (–3.5 km/h)

•

‘slow’ pavement markings ahead of curves (–5%). Austroads 2015 | page 20

Road Geometry Study for Improved Rural Safety

Figure 6.1:

Altered spacing and height of guide posts to make the curve appear more severe

Source: Austroads (2014d) based on Macaulay et al. (2004).

6.2 Vertical Alignment The Austroads guide to geometric design (Austroads 2010d) describes many points of good practice in terms of vertical alignment and vertical/horizontal combinations. Given the risk factor findings in Section 4.2, it is proposed to strengthen the wording on avoidance of substantial changes in vertical alignment and steep grades in rural road design. In particular, the findings repeatedly show the multiplicative casualty crash risk effect of combining grades, curves and sags with sharp horizontal curves. This should be emphasised more strongly in the guidelines.

6.3 Cross-section 6.3.1 Sealing Shoulders There is strong evidence in Section 4.3 and Section 5.2 showing that provision of sealed shoulders reduces the risk of casualty crashes. Provision of adequate pavement width has been shown to reduce the risk of rural road crashes in Section 4.3. Evidence in Appendix A.3 indicates that there would be limited benefits in providing undivided rural road pavement widths wider than 10 m. Thus, assuming the use of standard rural lane widths, the current design guidance for sealing shoulders to 1.5–2.0 m is adequate in Austroads (2010d). The safety reasons for sealing should be highlighted ahead of other considerations (e.g. stopping, asset management). The current advice in the guide remains open-ended on the decision point of when to provide sealed shoulders. More defined guidance could be provided based on traffic volumes or crash rates. Consideration of compromised geometric alignment should also be a strong factor in the decision (see Section 6.3.3). This detail could be developed from existing data sources used in recent Austroads research.

6.3.2 Providing Unsealed Shoulders There is sufficient evidence in Austroads (2014a) to consider that the provision of unsealed shoulders of 1–2 m strongly reduces the risk of run-off-road casualty crashes. The current guidance could be enhanced by noting the positive role of providing this design feature.

Austroads 2015 | page 21

Road Geometry Study for Improved Rural Safety

It is noted that the provision of 2.5 m to 4.0 m shoulders (sealed plus unsealed) effectively provides a minimum clear zone. This addresses the bulk of roadside crash risk identified in Section 4.4.

6.3.3 Sealed Shoulder Widening on the Outer Edge of the Curve This solution could be included in the Austroads guidance in addition to sealing shoulders, especially on higher-volume roads. Wider shoulders on the outside of curves would increase the percentage of erring drivers who recover safely back onto the road. This approach has been successfully applied in New South Wales. Figure 6.2 illustrates the design modification. Application of this solution could also be considered also along roads where full-length shoulder sealing is not warranted due to low traffic volumes. Shoulder sealing could be carried out on approaches through, and on the departures of curves. Figure 6.2:

Wider shoulders on the outside of curves

Source: Levett (2007).

6.3.4 Painted Medians Austroads design guidance could be updated to provide this solution for implementation during design or upgrade of roads. Painted medians provide separation of opposing traffic. They also provide a recovery space for drivers erring to the right, and to the left in the case of over-correction to the right. The painted median solution would be of particular value where the chance of error is high, i.e. on roads with high traffic volumes and/or with high frequency of curves. Narrow painted medians can be installed by trading off existing sealed shoulder width, where this is adequate, or when shoulders are sealed. Wide painted medians would require road widening. Whittaker (2012) carried out preliminary but robust evaluation of the provision of a narrow painted median in Queensland on rural roads with a high incidence of head-on and run-off-road crashes. The study results suggested a 75% reduction in severe head-on and centreline crossover crashes and a 59% reduction in severe run-off-road to the left and total crashes. Figure 6.3 shows a narrow median.

Austroads 2015 | page 22

Road Geometry Study for Improved Rural Safety

Figure 6.3:

Narrow painted medians deployed in Queensland

Source: Whittaker (2012).

The provision of a wide painted median, installed at 0.5 m, 1.0 and 2.0 m wide with a centre wire rope barrier has achieved significant crash reductions for all of these installations. The 0.5 m wide median reduced the number of crashes, but did not reduce cross-over type crashes. The 1.0 m wide median was found to reduce the number and severity of the crashes; and with the 2.0 m wide median with wire rope barrier, the number of crashes reduced but cross-over crashes remained almost constant (but these crashes were predominantly property damage type crashes, with the vehicle colliding with the barrier) (Levett et al. 2009).

6.3.5 Barriers in Painted Medians A ‘2 + 1’ solution consists of a three-lane road with the provision of alternating passing lanes and a wire rope safety barrier located in a narrow median. The treatment has been successfully implemented along significant lengths of high-volume rural highway network in Sweden over the past 15 years. The passing lanes are 1.0 to 2.5 km long, and alternate between each direction of travel, as shown in Figure 6.4. The width of the painted median varies, typically in the range of 2–3 m, although Swedish practice includes many safely operating sections with medians as narrow as 0.9 m 2. This treatment has been found to provide significant reductions in fatalities and serious injuries reported by the Swedes for all vehicle classes, including motorcyclists (Carlsson 2009). The wire rope barrier in the median prevents the majority of head-on and run-off-road to the right severe crashes. Limited Australian and New Zealand evaluations have confirmed and strengthened the safety evidence of this solution.

2

The rationale behind adopting a very narrow median is the safety benefit from reduced impact energy in case of a head-on or runoff-road crash. The higher probability of median barrier hits is partially offset by observed reductions in operating speeds due to the shy-line effect.

Austroads 2015 | page 23

Road Geometry Study for Improved Rural Safety

Figure 6.4:

A typical 2 + 1 configuration

Source: Austroads (2009b).

An alternative solution is to provide a 1 + 1 solution, with overtaking lanes spaced between 4 and 10 km apart. This solution could be reasonably applied in Australia, in the presence of lower traffic volumes than found in rural Sweden.

6.3.6 Gateway Treatments Gateway treatments are typically employed at the point of a speed zoning change from a higher to a lower limit. The gateway can be a single treatment, or a combination of treatments, such as a traffic island, pavement narrowing, raised or painted pavement markings. These treatments have been found to reduce the operating and mean speeds, but need to be accompanied by further changes in the speed environment along the road in order to retain the reduction (Makwasha & Turner 2013).

6.4 Roadsides Reduction of casualty crash risk due to roadside design relates primarily to reducing the likelihood and severity of run-off-road crashes. A number of proposed solutions relating to horizontal and vertical alignment will help to address the issue of likelihood. Based on the geometric roadside elements noted in Section 4.4, revision of the Guide to Road Design – Part 6 (Austroads 2010g) may be based on reconsideration of the existing focus on providing adequate clear zones. Austroads (2014a) concluded from a wide range of evidence that flexible wire rope barriers offered the lowest risk and severity of run-off-road casualty crashes on high-speed roads. Where this solution could not be achieved for economic reasons, the report offered a process for evaluating the risk of fatal and serious injury based on a mix of alternative roadside design solutions, such as other barrier types, wider clear zones, or hazard substitution.

6.5 Intersections 6.5.1 General Crash Risk Reduction Principles Improvement of the safety performance of rural road priority-controlled intersections should be based on general principles. These are already suggested by the current Austroads guides, and by the risk factors reviewed in Section 4.5. Such principles can assist in the identification of inherently lower-risk designs. The principles include:

•

minimise the number of high-exposure, high-speed conflict points

Austroads 2015 | page 24

Road Geometry Study for Improved Rural Safety

•

establish clear priority for movements through the intersection

•

separate conflict points in space (e.g. auxiliary lanes) and time (traffic signals)

•

control the angle of conflict; crossing streams of traffic should intersect at a right-angle or close to it, while merging streams should intersect at small angles to ensure low relative speed between the vehicles

•

control approach speeds using alignment, lane width, traffic control, speed limits, and ITS (e.g. vehicle activated signs)

•

define and minimise conflict areas

•

provide adequate sight distances

•

minimise roadside hazards

•

provide for all vehicular and non-vehicular traffic likely to use the intersection, including where necessary, special provisions for heavy vehicles, public transport vehicles, pedestrians and other vulnerable road users.

Safe System Principles These proposed Safe System principles for designers arise from the concurrent Austroads project SS1958 Safe System Assessment Framework. The practitioner has to assume that road users will make errors and crash. A Safe System infrastructure solution will not allow these road users to be fatally or seriously injured. A Safe System solution must take into consideration all likely road users and vehicle types, and their movements. A Safe System infrastructure solution will seek to: 1. minimise opportunities for impacting other road users by separating conflicting movements (e.g. exclusion, separation) 2. reduce impact speeds to survivable levels (e.g. < 50 km/h for right-angle impacts) if impacts are inevitable 3. minimise impact force transfer to road users in other ways, e.g. by reducing vehicle angles, changing vehicle impact areas (e.g. from side to rear), extending crash duration, or by redirecting vehicles. Solutions which only partially address these criteria are only partially aligned with Safe System principles. They will deliver valuable, but ultimately limited safety benefits and will need to be replaced by Safe System solutions in the future.

6.5.2 Roundabouts A rural roundabout is inherently safer than a comparable priority-controlled or signalised intersection, as discussed in Section 4.5 (Austroads 2010f, NZ Transport Agency 2013). A key contributor to the safer operation of roundabouts is the relatively low intersection negotiation speed and the inherent expectation to give way. Speed of less than 50 km/h should be achieved prior to the vehicle entering the circulating lanes of the roundabout. Thus, approach speeds on rural roads will typically need to be reduced. Methods to achieve this involve using a series of curves on the approach, as outlined in the Guide to Road Design - Part 3 (Austroads 2010d) and in Part 4B (Austroads 2011b). As noted in Section 4.5, this can be difficult to achieve in practice. One of the solutions to approach speed dissipation problems is to avoid multi-lane roundabouts where possible. For the majority of rural roundabouts, delay and queuing are minor considerations, and approach capacity may not be needed – design provisions can be made for future expansion.

Austroads 2015 | page 25

Road Geometry Study for Improved Rural Safety

Where multi-lane approaches and circulating lanes are needed to handle the flows, roundabout design guidance could include the use of extended raised traffic islands on the approaches. This solution would assist in speed dissipation and identification of the roundabout prior to the start of the approach lanes. Such solution would also separate the opposing traffic flows as they either enter or leave the roundabout. The benefits of including extended approach islands could be enhanced within the current guides covering roundabouts, i.e. the Guide to Traffic Management – Part 6 (Austroads 2013c) and the Guide to Road Design – Part 4B (Austroads 2011b). Stronger guidance should be considered for provision of clear run-off areas on departure sides of roundabouts and within the central islands. This is an important consideration in reducing the risk of loss-offcontrol injuries.

6.5.3 Turbo Roundabouts A further evolution of the roundabout design, which addresses a number of the risk factors listed in Section 4.5, is the addition of circulating and approach lane management using traffic islands. The following suggestions are based on ideas raised during a recent Transport Research Board Webinar: Is North America Ready for the Turbo Roundabout: Development and Advantages With and Without Raised Curbs by Mr LGH Fortuijn of the Delft University of Technology (Netherlands). Other inputs come from Fortuijn (2009). Turbo roundabouts were developed to improve the capacity of concentric two-lane roundabouts. The twolane roundabouts were found to not be adequate when there were unequal flows, i.e. a dominant flow in one direction, and drivers were undertaking weaving manoeuvres within the circulating lanes to change lanes, which was impacting on the capacity of the roundabout. The effect was a radical improvement in driver lane discipline and subsequent reductions in crashes. The following are forms of turbo roundabouts:

•

Basic (shown in Figure 6.5)

•

Egg – single-lane approach from the minor intersecting road

•

Spiral – the major road has three approach lanes and there are three circulating lanes at the major road entry, one of which serves the left-turn vehicles and does not continue past the first exit

•

Knee – the dominant flow turns left or right and the circulating lanes are aimed at facilitating this flow

•

Rotor – similar to the spiral roundabout, but accommodates higher flows by having three approach lanes to cater for the higher flows.

Further details of these designs can be found in Fortuijn (2009). The key safety benefits which have been identified are as follows:

•

cutting-off other vehicles is reduced due to the spiral lane marking

•

a small diameter central island results in low entry speeds

•

lane changes are avoided due to the raised lane dividers

•

conflict points are reduced, e.g. a two-lane concentric roundabout has 12 conflict points at the entry to the roundabout, plus weaving and cut-off movements; whereas a turbo roundabout has only 10 at the entry to the roundabout.

It should be noted that European roundabout designs employ radial entry, i.e. there is no speed dissipation on the approaches. The Austroads approach is considered safer (lower entry speeds, lower impact angles) and could be combined with the turbo concept. Desktop and field trials would be needed before adopting turbo design in Austroads practice.

Austroads 2015 | page 26

Road Geometry Study for Improved Rural Safety

Figure 6.5:

Basic form of a turbo roundabout

Source: From presentation delivered by by L.G.H. Fortuijn at TRB Webinar on 20 August (2014).

6.5.4 Priority-controlled Intersections The following sub-sections relate to suggested design element guidance enhancements applicable mainly to priority-controlled intersections. Their application could be extended.

Number of legs Austroads (2010f) shows that crash rates per 10 million entering vehicles are much lower for T-intersections. Where possible, cross-intersections should be avoided. Staggered T-intersections, described further on, provide an intermediate solution under certain conditions.

Channelisation The provision of median islands on the approach to an intersection can assist drivers to identify the location of the intersection and raise their alertness to select their travel path as they travel through the intersection. Median islands provide some protection for turning vehicles, when a turning lane is provided to take the turning vehicle out of the through lane. Austroads (2012b) reports that this treatment can achieve a reduction in head-on, rear-end and right-turn against type crashes by 20%. If the median island is placed through the intersection, thereby removing the cross-movement, head-on, right-turn against and right-angle type crashes can be eliminated. The provision of indented turn lanes with painted islands can achieve a 20% reduction in opposing turn and rear-end crashes; and with a median island, reductions of 40% in rear-end, 30% opposing turn, and 20% loss-of-control crashes can be achieved (Austroads 2012b). A greater emphasis on the use of channelisation on rural roads could be made in the Austroads guide on intersection design (Austroads 2010h).

Austroads 2015 | page 27

Road Geometry Study for Improved Rural Safety

Intersection Narrowing Similar to channelisation, but at lower cost, is the placement of a painted centre median to narrow the lanes and reduce approach speeds. This is supplemented by rumble strips within this median and along the outside of the edge lines of the pavement. Asokan and Bared (2009) propose to reduce the major road lane width from 3.7 m to 2.75 m with a median typically 1.2 m to 1.8 m wide. A typical layout of this treatment is shown in Figure 6.6. Figure 6.6:

Centre painted median with rumble strips

Source: Asokan and Bared (2009).

The rumble strips are applied on the painted centre median and along the shoulder of the road. The lengths of the treatment vary with the posted speed limit, and are shown in Table 6.2. Table 6.2:

Section lengths

Speed (km/h)

Section A (m)

Section B (m)

Section C (m)

70–85

30

60

45

100

45

60

45

Source: Asokan and Bared (2009).

The rumble strips, installed on the edge line, commence 15 m prior to the end of the median to avoid turning traffic from travelling across the strips. On the edge lines, the rumble strips commence 15 m prior to the start of section B. This treatment has been applied at eight sites in the USA, and preliminary evaluation found that the total number of crashes reduced by 32% and fatal/injury crashes reduced by 34%.

Austroads 2015 | page 28

Road Geometry Study for Improved Rural Safety

Staggered T-intersections Austroads (2010h) provides guidance on the design of staggered T-intersections where conspicuity of a cross-intersection is low from the high-speed side roads. The design (see Figure 4.2) is intended to also slow minor road drivers before entering the intersection. As noted in Section 4.5, the study by Chia, Jurewicz and Turner (2013) provided indicative design risk factors for this intersection type. Based on the findings, the authors provided points of general design guidance, which were that staggered T-intersections should have:

•

low major road traffic volumes (< 2000 vpd)

•

no significant curvature of the minor road approaches

•

left-right stagger type

•

stagger distance ≥ 15 m

•

advance warning signs on the major road.

Chia, Jurewicz and Turner (2013) concluded that this form of rural intersection should not be provided where traffic analysis indicates the likelihood of operation at or near capacity within its design life. If excessive delays are anticipated, then other intersection solutions should be considered. A Safe System assessment of the intersection indicated a low-moderate level of alignment with the Safe System principles.

Austroads 2015 | page 29

Road Geometry Study for Improved Rural Safety

7. Discussion The information on rural road design elements which contribute to casualty crash risk has been drawn from three approaches: crash data analysis, literature reviews, and direct analysis of the effect of improving selected design elements. Each approach presented limitations which require discussion to fully appreciate their validity, robustness and overall contribution to revision of future road design guidance.

7.1 Rural Crash Data Analysis Rural crash data analysis in Section 3 was able to provide a crude breakdown of the key aspects of crash risk through analysis of movement types and conditions that were over-represented in the crash databases. There is an inherent bias in this approach, as less severe crashes are under-reported in rural areas, which can result in focussing on the more severe end of the crash problem (not a bad thing in the Safe System context). Also, the definition of ‘rural’ is often problematic and varied between jurisdictions. This might have skewed some aspects of the analysis. Given the large numbers of crashes available for analysis, the results tend to be statistically robust, at least for all-casualty crashes. This means that the observed trends are not likely to be due to chance. This is less the case for some types of serious injury crashes. Fatal crash data rarely provides more than broad policy directions due to low crash counts. The benefit of rural crash data analysis was setting of the direction for further investigations via literature reviews, and then before and after analysis of design element changes.

7.2 Literature Reviews This part of the project built on the key directions emerging from the crash data analysis. Reviews of research and past analysis identified and quantified how various design elements, or their values, affect rural crash risk. This information could be then considered directly as evidence in revisions of guidance to reduce the potential for them to be attributed as a crash factor. Most of the potential solutions emerged from published and internal road agency publications. Relevance of some international research findings to Austroads needs to be considered. Many overseas examples operate under different traffic and fleet conditions. Relationships such as the one between the speed limit and operating speed, for example, may be very different to that experienced in some Australian states. Many studies, including Austroads ones, provide indicative results only, or results based on methods which did not account for errors. Seeking confirmation of findings between different research sources was one of the methods which sought to limit this problem. The contribution and value of Austroads research to this task were significant. Over a decade of road engineering safety research has been drawn on to provide inputs across all identified issue areas. A number of knowledge gaps have also been identified (e.g. in roadside drainage, vegetation and effects of surface on crash risk).

Austroads 2015 | page 30

Road Geometry Study for Improved Rural Safety

7.3 Evaluation of Selected Road Design Elements The results reported in Section 5 need to be considered in the context of their limitations. There were few sites nominated where road design elements were improved. The project showed that it was difficult to obtain project data in sufficient numbers for a robust before and after evaluation. It needs to be noted that a sufficiently long period was allowed for the nomination of sites, and included extensive follow-up with road agencies. The result of insufficient site numbers was that only one of the three evaluated design elements returned a robust study finding. For the sites that were nominated, it was difficult to piece together necessary project information such as:

•

scope of the works to establish which design element(s) were changed

•

start and end chainage of works (or GPS coordinates)

•

start and end date.

There was no centralised catalogue of completed works (e.g. as-built plans). It is proposed that such systems be established by jurisdictions to enable future evaluations of safety effectiveness of road improvements, and also assist in financial and other performance audits in road construction. Another limitation of the study was that treatments were not limited only to the single road design element changes listed. The sealed shoulder group, for example, comprised of sites where shoulder sealing was a major component; however, additional treatments were included such as delineation improvements (most sites), installation of drivable culvert end-walls, safety barriers at selected locations, and tree removal. Similarly, horizontal and vertical alignment improvements included shoulder sealing and tree removal. Pavement widening included shoulder sealing in every case. These inclusions represent a practical limitation to the study: rarely is a single design variable changed when roadworks are funded. Due to design impacts on surrounding features, crash risk and legal liability management often requires designers and road engineers to modify other design elements (e.g. renew the delineation or seal shoulders). For all the reasons discussed, it may be questionable whether before and after crash studies are the optimal method to quantify the contribution of selected crash road design elements to the occurrence and severity of crashes. The applicability, relevance and robustness of the results based on either a limited or highly biased sample of sites, and a combination of several treatments, deserves further discussion.

7.4 Relevance of Findings to Safe System Implementation Much of the presented evidence in this report refers to casualty crashes. More emphasis is needed on road design risk factors associated with fatal and serious injury (severe) crashes. This research is needed to promote Safe System consideration in any future guidance revisions. The severe crash evidence for road design elements is emerging. All new Austroads road safety research is now focussed on these high-severity crashes. Other research funding bodies in Australia and New Zealand also demand findings relevant to severe crashes, where data permits this (e.g. TAC, NZTA). Given that rural roads are typically high-speed environments, the majority of crashes occur at speeds in the 80–110 km/h range. The average severity of a casualty crash in these speeds is very high. For these reasons, the evidence applicable to casualty crashes presented in this report can be cautiously extended to apply to severe crashes on rural roads, until more appropriate evidence emerges. Sense-checking is needed when considering Safe System design implications using the evidence developed using casualty crash data.

Austroads 2015 | page 31

Road Geometry Study for Improved Rural Safety

8. Conclusions and Proposed Actions A workshop was held on 28 January 2015 to discuss the project findings and enable further input from members of the Task Force. The aim of the workshop was to discuss the presented findings of the project.

8.1 Workshop A preliminary draft report was circulated to all Task Force members for review and to inform the workshop discussion. The workshop was then held via video linkup, where Task Force members attended, discussed the presented findings, and provided further directions for the proposed changes to Austroads Road Design guides. Additional improvements to rural road safety were also identified and discussed. Input from this workshop was sought to provide solutions and form the basis for future research if required. This will allow the project to achieve its goal in identifying and quantifying road design elements that contribute to crash occurrence and severity, and will lead to identifying possible revisions to the current Road Design Guidelines to improve rural road safety. The following members attended: Noel O’Callaghan (Austroads Project Manager), Peter Ellis, Richard Fanning, Mike Whitehead, James Hughes and Michael Tziotis. They were joined by Chris Jurewicz (ARRB Project Technical Leader) and Stephen Tofler (Project Team Member).

8.2 Project Findings and Task Force Directions This project has made findings based on literature reviews, evaluation of design improvements, and previous Task Force inputs in four areas of rural road design:

•

horizontal and vertical alignment

•

cross-section

•

roadsides

•

intersections.

The members were asked to consider each suggested design guidance change and agree whether:

•

it is open for consideration in the next review of the relevant road design guide

•

further evidence is needed before a consideration can be made. Gather available or generate new research on the subject

•

it is not likely to be considered in the road design guidelines.

A summary of the Task Force directions and agreed actions arising from the meeting can be found in Table 8.1. The consensus on suggested changes should form the starting point input to future revisions of the Guide to Road Design series.

Austroads 2015 | page 32

Road Geometry Study for Improved Rural Safety

Table 8.1:

Key crash likelihood and severity factors, and suggested design guidance changes

Key crash likelihood and severity factors

Suggested design guidance changes

Consensus on the suggested changes

Horizontal alignment (see Section 4.1) Out-of-context curve (tight radius, high speed reduction) (see Section 4.1)

1. Strengthen design speed selection guidance to avoid out-of-context curves and other low-speed elements (e.g. to force an alignment change). (See Sections 6.1.1, 6.1.2 and 6.1.3).

Open for consideration

High curve frequency along a route (see Section 4.1)

2. Reduce the expected speed limit in recognition of the topography and road hierarchy which may lead to curved alignment. (See Section 6.1.4).

Open for consideration

Curve combined with grade (see Section 4.1 and 4.2)

3. Stronger discouragement of combining sharp curvature (< 600 m) with grade steeper than 6% in highspeed environments. (See Sections 6.1.2 and 6.2).

Open for consideration

Curves to the right (see Section 4.1)

4. See item 10.

Vertical alignment (see Section 4.2) Downhill grade > 6% (see Sections 4.1 and 4.2)

5. Strengthen guidance to avoid downhill grade > 6% in high-speed environments. Suggest a lower speed limit if conditions continue over a significant distance. (See Sections 6.1.1 and 6.1.2).

Open for consideration

Crests and sags with a sharp vertical curve (< 600 m) (see Section 4.2)

6. Strengthen guidance to avoid sharp vertical curves (crests and sags) in high-speed environments. Suggest a lower speed limit if conditions continue over a significant distance. (See Section 6.2).

Not likely to be considered (already in the guidelines)

Narrow pavement width (< 6 m) (see Section 4.3)

7. Consider minimum 9 m pavement width for high-speed roads (See Section 6.3.1).

Open for consideration

Narrow lane widths (< 3.5 m) (see Section 4.3)

8. Strengthen the importance of using adequate lane widths (≥ 3.5 m). See Point 8. (See Section 6.3.1).

Open for consideration

Little or no sealed shoulder (see Section 4.3)

9. Suggest wider shoulders on the outside of curves (RMS example). (See Sections 6.3.2 and 6.3.3).

Open for consideration

10. Emphasise positive safety benefits of sealing shoulders on high-speed roads. (See Sections 6.3.2 and 6.3.3).

Open for consideration

11. Suggest provision and maintenance of unsealed shoulder wherever practicable. (See Section 6.3.2).

Open for consideration

Cross-section (see Section 4.3)

Austroads 2015 | page 33

Road Geometry Study for Improved Rural Safety

Key crash likelihood and severity factors

Suggested design guidance changes

Consensus on the suggested changes

13. Implement changes to AGRD Part 6 in line with recommendations of Improving Roadside Safety (Austroads 2014a), i.e.:

Open for consideration

Roadsides (see Section 4.4) High speeds (See Section 4.4) Presence of roadside slopes steeper than 1:3.5 (See Section 4.4) Roads with narrow clear zones (See Section 4.4) Severe roadside hazards such as trees, poles, embankments (See Section 4.4)

• removing focus on clear zones • greater use of wire rope safety barrier (WRSB) • use other road design features to reduce FSI risk Safe System + economic optimisation approach for each road location/stereotype. (See Section 6.4).

High density of hazards (> 50 per 100 m) (See Section 4.4) Intersections (see Section 4.5) Priority-controlled rural intersections (see Section 4.5)

Multiple approach and circulating lanes at roundabouts; high speeds in roundabouts (see Section 4.5)

Right-left staggered T-intersection design (see Section 4.5)

14. Emphasise safety benefits of roundabouts within the current guidelines; promote their application ahead of priority intersections. (See Section 6.5.2).

Not likely to be considered (belongs in the Austroads Guide to Traffic Management)

15. Strongly discourage cross-intersections in favour of T-intersections. (See Section 6.5.4).

Not likely to be considered (belongs in the Austroads Guide to Traffic Management)

16. Promote intersection channelisation and side road traffic islands.(See Section 6.5.4).

Further research needed

17. Strengthen/clarify guidance on achieving low approach and circulating speeds at roundabouts (more examples, typical scenarios). (See Section 6.5.2).

Open for consideration

18. Develop guidance on use of extended raised traffic islands on roundabout approaches for improved channelisation. (See Sections 6.5.2 and 6.5.4).

Not likely to be considered (already in the guidelines)

19. Trial turbo roundabouts. (See Section 6.5.3).

Further research needed

20. Discourage right-left staggered T-intersections in favour of left-right staggered T-intersections. (See Section 6.5.4).

Not likely to be considered (belongs in the Austroads Guide to Traffic Management)

Austroads 2015 | page 34

Road Geometry Study for Improved Rural Safety

References Arndt, OK 2003, Relationship between unsignalised intersection geometry and accident rates, Queensland University of Technology, Brisbane, Qld, viewed 10 February 2015, . Arnold, ED & Lantz, KE 2007, Evaluation of best practices in traffic operations and safety: phase I: flashing LED stop sign and optical speed bars, report VTRC 07-R34, Virginia Transportation Research Council, Charlottesville, Virginia, USA. Austroads 2005, Road safety in rural and remote areas of Australia, AP-R273-05, Austroads, Sydney, NSW. Austroads 2006, Guide to road safety: part 5: road safety for rural and remote areas, AGRS05-06, Austroads, Sydney, NSW. Austroads 2009a, Heavy vehicle safety in rural and remote areas, AP-T136-09, Austroads, Sydney, NSW. Austroads, 2009b, Evaluation of the safety impact of centre-of-the-road wire rope barrier (WRB) on undivided rural roads, AP-T135-09, Austroads, Sydney, NSW. Austroads 2010a, Infrastructure/speed limit relationship in relation to road safety outcomes, AP-T141-10, Austroads, Sydney, NSW. Austroads 2010b, Road safety engineering risk assessment: part 1: relationships between crash risk and the standards of geometric design elements, AP-T146-10, Austroads, Sydney, NSW. Austroads 2010c, Safe intersection approach treatments and safer speeds through intersections: final report: phase 1, AP-R363-10, Austroads, Sydney, NSW. Austroads 2010d, Guide to road design: part 3: geometric design, 2nd edn, AGRD03-10, Austroads, Sydney, NSW. Austroads 2010e, Road safety engineering risk assessment: part 8: rural head-on crashes, AP-T153-10, Austroads, Sydney, NSW. Austroads 2010f, Road safety engineering risk assessment: part 7: crash rates database, AP-T152-10, Austroads, Sydney, NSW. Austroads 2010g, Guide to road design: part 6: roadside design, safety and barriers, AGRD06-10 Austroads, Sydney, NSW. Austroads 2010h, Guide to road design: part 4A: unsignalised and signalised intersections, 2nd edn, AGRD04A-10, Austroads, Sydney, NSW. Austroads 2011a, Improving roadside safety: stage 2: interim report, AP-R387-11, Austroads, Sydney, NSW. Austroads, 2011b, Guide to road design: part 4B: roundabouts, AGRD08-11, Austroads, Sydney, NSW. Austroads 2012a, An introductory guide for evaluating effectiveness of road safety treatments, AP-R421-12, Austroads, Sydney, NSW. Austroads 2012b, Effectiveness of road safety engineering treatments, AP-R422-12, Austroads, Sydney, NSW. Austroads 2013a, Expanded operating speed model, AP-T229-13, Austroads, Sydney, NSW.

Austroads 2015 | page 35

Road Geometry Study for Improved Rural Safety

Austroads 2013b, Improving the performance of safe system infrastructure: stage 1: interim report, AP-T25613, Austroads, Sydney, NSW. Austroads 2014a, Improving roadside safety: summary report, AP-R437-14, Austroads, Sydney, NSW. Austroads 2014b, Australian national risk assessment model, AP-R451-14, Austroads, Sydney, NSW. Austroads 2014c, Model national guidelines for setting speed limits at high-risk locations, AP-R455-14, Austroads, Sydney, NSW. Austroads 2014d, Methods for reducing speeds on rural roads: compendium of good practice, AP-R449-14, Austroads, Sydney, NSW. Baldock, MRJ, Kloeden, CN & McLean, AJ 2008, In-depth research into rural road crashes, report CASR057, Centre for Automotive Safety Research, Adelaide, SA. Barua, U, Azad, AK & Tay, R 2010, ‘Fatality risk of intersection crashes on rural undivided highways in Alberta, Canada’, Transportation Research Record, no. 2148, pp 107-15. Candappa, N & Corben, B 2011, Intersection study task 2 report: targetted literature review, report 316b, Monash University Accident Research Centre, Clayton, Vic. Cardoso, J 2005, ‘Safety assessment for design and redesign of horizontal curves’, International symposium on highway geometric design, 3rd, Chicago, Illinois, Transportation Research Board, Washington, DC, USA, 20 pp. Carlsson, A 2009, Evaluation of 2+1 roads with cable barrier: final report, Swedish National Road and Transport Research Institute (VTI), report 636A, VTI, Linköping Sweden. Charlton, SG 2003, ‘Restricting intersection visibility to reduce approach speeds’, Accident Analysis & Prevention, vol. 35, no. 5, pp. 817-23. Charlton, SG & Baas, PH 2006, Speed change management for New Zealand roads, research report 300, Land Transport New Zealand, Wellington, New Zealand. Charman, S, Grayson, G, Helman, S, Kennedy, J, de Smidt, O, Lawton, B, Nossek, G, Wiesauer, L, Fürdös, A, Pelikan, V, Skládaný, P, Pokorný, P, Matĕjka, M & Tučka, P 2010, Self-explaining roads literature review and treatment information, Speed Adaptation Control by Self-Explaining Roads (SPACE) deliverable 1, ERA-NET Road. Chia, S, Jurewicz, C & Turner, B 2013, ‘Staggered T rural intersections: investigation of safety effectiveness’, contract report, ARRB Group, Vermont South, Vic. Choueiri, EM, Lamm, R, Kloeckner, JH & Mailaender, T 1994, ‘Safety aspects of individual design elements and their interactions on two-lane highways: international perspective’, Transportation Research Record, no. 1445, pp. 34-46. Department of Transport and Main Roads 2002, Road planning and design manual: chapter 11: horizontal alignment, TMR, Brisbane, Qld. Derr, BR 2003, ‘Application of European 2+1 roadway designs’, NCHRP Research Results Digest Record, no. 275, 31 pp. Donnell, E, Harwood, D, Bauer, K, Mason, J & Pietrucha, M 2002, ‘Cross-median collisions on Pennsylvania interstates and expressways’, Transportation Research Record, no. 1784, pp.91-9.

Austroads 2015 | page 36

Road Geometry Study for Improved Rural Safety

Engelsman, JC & Uken, M 2007, ‘Turbo roundabouts as an alternative to two lane roundabouts’, Proceedings of the 26th Southern African Transport Conference (SATC 2007), Pretoria, South Africa, 9 pp. European Road Safety Commission 2014, Getting initial safety design principles right, ERSC, Brussels, Belgium, viewed 10 February 2015, . Federal Highway Administration 2014, Interactive highway safety design model (IHSDM): overview, FHWA, Washington, DC, USA, viewed 10 February 2015, . Fitzpatrick, K, Fambro, DB & Stoddard, AM 2000, ‘Safety effects of limited stopping sight distance on crest vertical curves’, Transportation Research Record, no. 1701, pp. 17-24. Fortuijn, L 2009, ‘Turbo roundabouts: design principles and safety performance’, Transportation Research Record, no. 2096, pp. 16-24. Gattis, J & Low, S 1998, ‘Intersection angle geometry and the driver’s field of view’, Transportation Research Record, no. 1612, pp. 10-6. Gross, F, Jagannathan, R & Hughes, W 2009, ‘Two low-cost safety concepts for two-way, stop-controlled intersections in rural areas’, Transportation Research Record, no. 2092, pp. 11-8. Gross, F, Jovanis, PP, Eccles, K & Chen, K-Y 2009, Safety evaluation of lane and shoulder width combinations on rural, two-lane, undivided roads, FHWA-HRT-09-031, Federal Highway Administration, McLean, Virginia, USA. Harwood, DW, Council, FM, Hauer, E, Hughes, WE & Vogt, A 2000, Prediction of the expected safety performance of rural two-lane rural highways, report FHWA-RD-99-207, Federal Highway Administration, McLean, Virginia, USA. Hallmark, SL, Peterson, E, Fitzsimmons, E, Hawkins, N, Resler, J & Welch, T 2007, Evaluation of gateway and low-cost traffic-calming treatments for major routes in small, rural communities, CTRE project 06-185, Centre for Transportation Research and Education, Iowa State University, Ames, IA, USA. Herrstedt, L 2007, Narrow cross sections without centreline markings: “2 minus 1” rural road: road user behaviour study: summary note, Trafitec, Denmark. Hughes, W, Jagannathan, R & Gross, F 2008, Two low-cost safety concepts for two-way STOP-controlled, rural intersections on high-speed two-lane two-way roadways: summary report, FHWA-HRT-08-063, Federal Highway Administration, McLean, Virginia, USA. Jagannathan, R, Gimbel, M, Bared, JG, Hughes, WE, Persaud, B & Lyon, C 2006, ‘Safety comparison of New Jersey jug handle intersections and conventional intersections’, Transportation Research Record, no. 1953, pp. 187-200. Jurewicz, C, Chau, T, Mihailidis, P & Bui, B 2014, ‘From research to practice: development of rural mass curve treatment program’, Australasian road safety research, policing and education conference, 2014, Australasian College of Road Safety, viewed 10 February 2015, . Jurewicz, C, Lim, A & Phillips, C 2013, ‘Route-based rural curve safety treatments, stage 2: development of parameters for a program business case’, VicRoads report Q23–03172, VicRoads, Kew, Vic. Leonard, J, Bilse, D & Recker, W 1994, Superelevation rates at rural highway intersections, UCI-ITS-RR-941, Institute of Transportation Studies, Irvine, California, USA.

Austroads 2015 | page 37

Road Geometry Study for Improved Rural Safety

Levett, S 2005, ‘The application of asymmetrical design principles to rural roads’, Australasian road safety research policing education conference, 2005, Wellington, New Zealand, Ministry of Transport, Wellington, New Zealand, 9 pp. Levett, S 2007, ‘Retro-fitting road safety to existing rural roads’, IPWEA NSW conference, 2007, Sydney, New South Wales, Australia, IPWEA, Sydney, NSW, 17 pp. Levett, S, Job, RFS & Tang, J 2009, ‘Centreline treatment countermeasures to address crossover crashes’, Australasian road safety research, policing and education conference, 2009, Sydney, NSW, Australia, Roads and Traffic Authority, Sydney, New South Wales, 14 pp. McGee, HW & Hanscom, FR 2006, Low cost treatments for horizontal curve safety, FHWA-SA-07-002, Federal Highway Administration, Washington, DC, USA. Makwasha, T & Turner, B 2013, ‘Evaluating the use of rural-urban gateway treatments in New Zealand’, Australasian road safety research policing education conference, 2013, Brisbane, Queensland, Australia, Australasian College of Road Safety, Mawson, ACT, 9 pp. Makwasha, T & Turner, B 2014, ‘Evaluating vehicle activated signs on rural roads’, ARRB conference, 26th, 2014, Sydney, New South Wales, ARRB Group, Vermont South, Vic, 15 pp. Neuman, TR, Pfefer, R, Slack, KL, McGee, H, Prothe, L, Eccles, K & Council, F 2003, Guidance for implementation of the AASHTO Strategic Highway Safety Plan: volume 4: a guide for addressing head-on collisions, NCHRP report 500, Transportation Research Board, Washington, DC, USA. NZ Transport Agency 2013, High risk intersections guide, NZTA, Wellington, New Zealand. Olson, PL, Cleveland, DE, Fancher, PS, Kostyniuk, LP & Schneider, LW 1984, Parameters affecting stopping sight distance, NCHRP report 270, Transportation Research Board, Washington, DC, USA. Oxley, J, Corben, B, Koppel, S, Fildes, B, Jacques, N, Symmons, M & Johnston, I 2004, Cost- effective infrastructure measures on rural roads, Monash University Accident Research Centre, report 217, Monash University, Clayton, Vic. Pyta, V & Tziotis, M 2011, ‘Heavy vehicle crashes in rural and remote Australia and New Zealand’, Australasian college of road safety conference, 2011, Melbourne, Victoria, Australia, Australasian College of Road Safety (ACRS), Pearce, ACT, 13 pp. Reekmans, S, Nuyts, E & Cuyvers, R 2004, Effectiveness of infrastructural road safety measures (in Dutch), RA-2004-39, Policy Research Centre for Traffic Safety, Belgium, Brussels. RoadSafe LLC 2012, Appendix B: engineer’s manual RSAP, RoadSafe LLC, Maine, USA, viewed 11 February 2015, . Schneider, WH, Savolainen, PT & Moore, DN 2010, ‘Effects of horizontal curvature on single-vehicle motorcycle crashes along rural two-lane highways’, Transportation Research Record, no. 2194, pp. 91-8. Stamatiadis, N, Pigman, J, Sacksteder, J, Ruff, W & Lord, D 2009, Impact of shoulder width and median width on safety, NCHRP report no. 633, Transportation Research Board, Washington, DC, USA. Standards Australia 1999, Road safety barrier systems, AS/NZS 3845, Standards Australia, North Sydney, NSW. Stein, WJ & Neuman, TR 2007, Mitigation strategies for design exceptions, Federal Highway Administration, Washington, DC, USA.

Austroads 2015 | page 38

Road Geometry Study for Improved Rural Safety

Sutton, M 2013, ‘Dutch roundabouts and eye-level signals on trial in UK’, BikeBiz, 30 April 2013, viewed 21 January 2015, . Theeuwes, J & Godthelp, H 1995, ‘Self-explaining roads’, Safety Science, vol. 19, pp. 217-25. Transport Research Laboratory 2015, Dutch style roundabouts, TRL, Crowthorne, UK, viewed 11 February 2015, . Turner, B, Tziotis, M, Cairney, P & Jurewicz, C 2009, Safe system infrastructure: national roundtable report, research report ARR 370, ARRB Group, Vermont South, Vic. Whittaker, A 2012, ‘The safety benefit of continuous narrow painted median strips’, Engineering Technology Forum, 2012, Transport and Main Roads, Brisbane, Qld, 7 pp.

Austroads 2015 | page 39

Road Geometry Study for Improved Rural Safety

Geometric Design Elements and Crash Risk This appendix provides detailed findings of the review of available research literature and existing design guidance relating to geometric design elements that may contribute to increased crash risk or crash severity in the rural road environment.

A.1

Horizontal Alignment

A.1.1

Curve Radius

Curve radius has a large impact on crash risk. In Australia and New Zealand, 40% of heavy vehicle casualty crashes on rural roads happen on curves (Pyta & Tziotis 2011). Levett (2007) analysed crash rates in New South Wales, which showed that the majority of injury crashes occurred on curves with a radius range between 150 and 450 m, as shown in Figure A 1. The majority of fatal curve crashes were distributed in a similar radius range of 250 to 550 m. Levett’s analysis represents both the inherent risk of sharper curves and their occurrence on the NSW rural road network. Figure A 1:

Crash numbers and curve radius

Source: Levett (2007).

Austroads (2010a) took this sort of analysis a step further, showing a strong relationship between curve radius and casualty crash rate per 100 million kilometres of travel (i.e. adjusted for exposure of vehicles to curves of different radius). The relationship is shown in Figure A 2 and shows that as the curves increase in radius, the relative casualty crash risk decreases. The crash risk for a very sharp curve (< 100 m radius) was found to be approximately 5.5 times that of a relatively straight section (> 1000 m radius).

Austroads 2015 | page 40

Road Geometry Study for Improved Rural Safety

Consideration of both Figure A 1 and Figure A 2 suggests that acceptance and proliferation of less risky curves in the 200–600 m range may actually deliver a greater road safety burden across the network than the relatively rare hair-pin bends of less than 100 m in radius. Figure A 2:

Crash rate vs curve radius

Source: Austroads (2010b).

The other component to examining crashes is the severity of the crash. This has also been examined in Austroads (2011a), which reports the effect of curve radius on crash severity, with an average cost used as the indicator. Figure A 3 shows that sharper curves had a lower average cost, i.e. severity, which increased for milder curves. This may be interpreted as the effect of speed. Curves with a radius under 400 m generally require drivers to slow down, even if marginally. This is likely to have a strong positive effect on the severity of outcomes in the case of a run-off-road event.

Austroads 2015 | page 41

Road Geometry Study for Improved Rural Safety

Figure A 3:

Curve radius and run-off road casualty crash cost

$600,000

Average ROR casualty crash cost

$500,000

$400,000

$300,000

$200,000

$100,000

$-

500

1,000

1,500

2,000

2,500

3,000

3,500

Curve radius (m)

Source: Austroads (2011a).

Austroads (2010e) reported on several research reports that indicated that the drivers’ selection of their speed was influenced by several factors, one of which was the road geometry and their perception of the ‘safe’ speed to travel along the road.

A.1.2

Curve Combinations

The frequency of curves and the curve radius may also increase the crash risk. When a sharp curve is located after a sequence of larger radii curves, the differences in the speed at which the curves can be negotiated can cause drivers to misjudge a safe speed for particular curves and result in increased crashes. Figure A 4 provides an indication of the range of appropriate curve radius combinations for curve sequences.

Austroads 2015 | page 42

Road Geometry Study for Improved Rural Safety

Figure A 4:

Turning radii in curve sequences

Source: European Road Safety Commission (2014).

Austroads (2010a) reports on an analysis of crash and traffic volume data on Victorian rural, undivided state roads that provided a relationship between the number of curves along a section of road, the road ‘curviness’, and the crash rate. The crash rates increased with the frequency of curves, as shown in Figure A 5. That is, as the number of curves increases, the crash rate increases. Curves were less than 1500 m radius and in the same direction of curvature.

Austroads 2015 | page 43

Road Geometry Study for Improved Rural Safety

Figure A 5:

Curve frequency and crash rates

Source: Austroads (2010a).

A.1.3

Superelevation

Superelevation is the development of pavement crossfall to counteract the effects of lateral forces as a vehicle travels around a curve. In conjunction with the curve radius, the effect of superelevating a curve has an influence factor on the crash risk. Where the superelevation is deficient, there is a greater lateral force on the vehicle as it travels around the curve, which can result in the vehicle leaving the travelling lane and running off the road or into the opposing traffic flow. A comparison of the lack of appropriate superelevation compared with curves that have appropriate superelevation is reported in Stein and Neuman (2007), and Table A 1 shows the relative risk (or accident modification factor (AMF) in the USA) corresponding to superelevation deficiency. Table A 1:

Accident modification factors for superelevation on rural two-lane highways

Superelevation deficiency

Accident modification factor

0.02

1.06

0.03

1.09

0.04

1.12

0.05

1.15

Source: Harwood et al. (2000).

Austroads 2015 | page 44

Road Geometry Study for Improved Rural Safety

A.2

Vertical Alignment

A.2.1

Vertical Grades

Austroads (2011a) crash rates analysis of rural undivided roads found that the vertical grade affects vehicle speed and control. Downhill grade was found to have a stronger effect on the risk of run-off-road (ROR) crashes than the uphill grade, as shown in Figure A 6. Run-off-road casualty crash risk was 4.5 times higher on a road with a downhill grade of more than 6%, than on a relatively flat road. A similar uphill grade was three times more risky than a flat road. Combination of downhill grade and sharp curves was found to be a particularly strong contributor to the risk of run-off-road casualty crashes on rural roads. Austroads (2011a) proposes a simple evaluation matrix quantifying this risk relationship. Austroads (2014a) developed this further into a framework for assessment of run-off-road risk in the Safe System context given different road design parameters. Figure A 6:

Grade and run-off-road crash likelihood

Source: Austroads (2011a).

Vertical alignment was also considered by Harwood et al. (2000) in a study on a two-lane road, which provided accident modification factors (all severities). Table A 2 shows the effect that grade can have on the safety of the road. These AMFs apply to the road section, irrespective of whether the grade is uphill or downhill.

Austroads 2015 | page 45

Road Geometry Study for Improved Rural Safety

Table A 2:

Accident modification factors for grade on rural two-lane highways

Grade (%)

Accident modification factor (AMF)

0

1.00

2

1.03

4

1.07

6

1.10

8

1.14

Source: Harwood et al. (2000).

Austroads (2010b) reports that a study by Choueiri et al. (1994) showed that crash rates increased only slightly up to a 6% grade, beyond which the rates increased sharply. This was contrary to the relationship in the Harwood report as indicated in Table A 2. Austroads (2010b) recommended relative risk ratios, based on the Harwood work, as shown in Table A 3. Note that for grades steeper than 6%, the relative risk has been significantly increased. Table A 3: Grade (%) Relative risk

Relative risk ratios for grade of road 0

2

4

6

8

1.00

1.03

1.05

1.10

1.16

Source: Austroads (2010a).

These studies did not provide any information on the whether there was any influence by other road elements, in particular road curvature, on their findings. This is an area that would require further research. Austroads (2011a) reports a relationship where the crash severity (as indicated by ROR crash cost) varies with the gradient of the road, as shown in Figure A 7. The severity, as measured by crash cost, decreases as the road grade increases to the 4 to 6% range, after which there is a significant increase in severity. Further analysis would be required to identify the causes for the significant change in crash severity on grades steeper than 6%.

Austroads 2015 | page 46

Road Geometry Study for Improved Rural Safety

Figure A 7:

Road grade and average ROR crash cost

Source: Austroads (2011a).

A.2.2

Vertical Curves

The standards for vertical curves are determined by the sight distance along the curve. Only limited research has been undertaken on the effects of sight distance on crash rates. Austroads (2010b) cites that studies have been undertaken by Choueiri et al. (1994), which indicates that crash rates are higher with low sight distances but do not have much change when the sight distance is greater than 150 to 200 m; and by Olson et al. (1984), which matched low sight distance crests (36 to 94 m) with nearby crests of high sight distance (greater than 215 m), with otherwise comparable geometry. The crash numbers on the low sight distance crests were 50% greater than those on the high sight distance crests. The design speed standard in the study by Olson et al. compared crests of design speed standard less than 70 km/h with crests of design speed standard greater than 110 km/h, with the actual operating speed probably between 90 and 100 km/h. Austroads (2010b) provided the relative risk ratios for crests shown Table A 4 below. Table A 4:

Relative risk ratios for sight distance deficiencies on crest curves

Sight distance deficiency Relative risk

< 40% of design value

> 40% of design value

1.1

1.4

Source: Austroads (2010b).

Austroads 2015 | page 47

Road Geometry Study for Improved Rural Safety

Oxley et al. (2004) also cites a study by Fitzpatrick et al. (2000), which examined the effects of sight distance on crashes on three rural roads that had different characteristics-multi-lane, two-lane with shoulders, and two-lane without shoulders. They found that the multi-lane road had the lowest crash rate and the two-lane road without shoulders had the highest crash rate, but only a small number of crashes on the roads with limited sight distance were attributable to the limited sight distance. Fitzpatrick et al. (2000) concluded that from the results of the study, crash rates on rural two-lane roads with limited stopping sight distance are similar to the crash rates on all two-lane roads.

A.2.3

Summary

On vertical grades, casualty crashes increase as the road grade increases, particularly on downhill grades; with a significant increase in the number of crashes and the severity of the crashes when the grade is greater than 6%. Sight distance along vertical curves affects the crash risk, and it has been found that where the curve is deficient by greater than 40% compared to the design value, the risk increases by 40%.

A.3

Cross-section

A.3.1

Pavement Widths

Austroads (2011a) indicates that the sealed pavement width has an influence on the run-off-road crash likelihood, and provides the crash rates shown in Table A 5. Table A 5:

Effect of sealed pavement width on ROR crash rates

Sealed pavement width (m)

ROR casualty crashes per 100M VKT

<6

11.0

6–7

6.9

7–8

6.4

8–9

4.6

9–10

4.1

This shows that sealed pavement widths less than 6 m have a much higher crash rate compared to sealed pavements greater than 6 m. A further marked decrease occurs when the pavements are greater than 8 m wide. Another issue with narrow pavements is shown in Figure A 8, which illustrates that the run-off-road crash rates associated with a particular sealed pavement width range decreases as the clear zone width is increased. Furthermore, the effect of increasing clear zone width on reducing ROR crash rate is relatively greater for the narrowest seal width range (< 6 m).

Austroads 2015 | page 48

Road Geometry Study for Improved Rural Safety

Figure A 8:

Effect of sealed pavement width and clear zone on ROR crash likelihood

25

<6m 6-7 m

ROR casualty crashes per 100M VKT

20

7-8 m 8-9 m 9-10 m 15

10

5

0 < 4.0 m

4-8m

>8m

Clear zone range

Source: Austroads (2011a).

A.3.2

Lane Widths

As part of the development of a crash prediction algorithm, Harwood et al. (2000) developed accident modification factors (AMFs) including the results of before and after crash investigations and consultation with an expert panel. These AMFs are shown in Figure A 9. The base width for the lanes is 3.6 m (12 feet), and as lane widths increase, the AMF values decrease. These factors for lane widths less than 3.6 m are constant for average daily traffic (ADT) volumes above 2000 vpd, but decrease for ADTs less than 2000 vpd.

Austroads 2015 | page 49

Road Geometry Study for Improved Rural Safety

Figure A 9:

Accident modification factors for lane width on rural two-lane highways

Source: Harwood et al. (2000).

Austroads (2010a) provides a relationship between lane width and crash rates, based on Queensland data. The rural results, particularly for the two-lane undivided major roads that have a speed limit of 100 km/h, shows that the narrow lanes (2.5 m wide) have a crash rate that is approximately 50% higher than the wider (3.5 m) lanes, as illustrated in Figure A 10.

Austroads 2015 | page 50

Road Geometry Study for Improved Rural Safety

Figure A 10:

Lane widths and crash rates

Source: Austroads (2010a).

A.3.3

Shoulder Width

Road shoulders are provided to support the structural function of the travelling lane and also to allow a better opportunity for a driver to recover and redirect their errant vehicle back onto the travelling lanes. Baldock et al. (2008) in a study of rural crashes in South Australia, reported that where the crash involved more than one vehicle, the most common crash type was where vehicles were making a right turn and headon crashes. Many of the head-on crashes were from a vehicle travelling on the unsealed shoulder and then moving back across the road in an out-of-control manner. The effects of sealing the road shoulder are shown in Figure A 11 (Austroads 2010a). There is a change of approximately 25% between the crash rates of a narrow sealed shoulder (0.5 m), compared with a wider sealed shoulder (2.5 m).

Austroads 2015 | page 51

Road Geometry Study for Improved Rural Safety

Figure A 11:

Crash rate and sealed shoulder width

Source: Austroads (2010a).

Shoulder sealing on straights also affects the crash risk. The wider the shoulder, the greater the safety benefit (Levett 2005). However, the relationship is not linear, with a law of diminishing returns, as shown in Table A 6. Table A 6:

Effect of sealing shoulder on crash rate

Sealed width (m)

Crash rate per 100million vehicle km

Reduction in crash rate (%)

0

42.5

n/a

0.5

31.8

25

1.0

26.4

38

1.5

24.5

42

Source: Levett (2005).

Due to this effect, Levett proposes that extra width be used to add a painted median rather than additional sealed shoulder width. Wide shoulders may also encourage higher speeds (Stamatiadis et al. 2009), as they provide the driver with a wider area to correct the vehicle. Charman et al. (2010) also reported that drivers may increase their speed and overtaking manoeuvres if the shoulders are very wide. Therefore, very wide shoulders may increase the potential for crashes by encouraging excessive speed and undesirable manoeuvres.

Austroads 2015 | page 52

Road Geometry Study for Improved Rural Safety

A.3.4

Lane and Shoulder Width

The effect of lane width and shoulder width also has an impact on the crash risk. Austroads (2010b) shows that the relative crash rate varies with differing lane and shoulder widths. The selection of these components of the road design can be undertaken separately, but the effects of each element combined has an influence on the relative crash rate, as can be seen in Figure A 12. Figure A 12:

Crash rate relative to a 3.6 m lane and a 3.0 m shoulder

Source: Austroads (2010b).

A.3.5

Lane Separation

Head-on crashes are a common type of crash on a rural road. This type of crash occurs most commonly on undivided roads, but can also occur on divided roads where the vehicle is not prevented from crossing the median.

Austroads 2015 | page 53

Road Geometry Study for Improved Rural Safety

Neuman et al. (2003) commented that head-on crashes were most likely due to drivers making unintentional manoeuvres, ‘wandering’ into the opposing traffic lanes due to inattention, or travelling too fast around a horizontal curve. Austroads (2010e) cited Oxley et al. (2004), stating that head-on crashes occur for four main reasons:

•

when the driver’s performance declines due to fatigue or inattention, and they allow the vehicle to wander into the path of the opposing traffic flows

•

when the driver wanders onto the road shoulder and over-corrects the movement, and then enters the opposing traffic flow

•

when a driver misjudges their overtaking manoeuvre

•

when the driver travels around a horizontal curve and enters the opposing flow lanes. This is often due to excessive speed.

A.3.6

Embankments

Austroads (2011a) investigated the effects of embankment (i.e. roadside slopes) and clear zone width on run-off-road crash likelihood. Based on limited availability of data, a relationship between ROR crash rates and the roadside slope was developed and is shown in Figure A 13. This shows that the crash likelihood on steep slopes or drop-offs (slopes steeper than 1:3.5 on the near-side or left-side of the road pavement) was much more than on a flat roadside (1:6 of flatter). Figure A 13:

Effect of near-side batter slope and clear zone on ROR crash likelihood

18

16 Steep and wall/drop off Flat roadside

ROR casualty crashes per 100M VKT

14

12

10

8

6

4

2

0 0 - 1m

1 - 2m

2 - 3m

3 - 4m

4 - 5m

5 - 6m

6 - 7m

7 - 8m

8 - 9m

9 - 10m 10 - 11m 11 - 12m 12 - 13m

> 13m

Clear zone

Source: Austroads (2011a).

Austroads 2015 | page 54

Road Geometry Study for Improved Rural Safety

A.3.7

Summary

Sealed pavement widths less than 6 m wide were found to have a casualty crash rate of 11.0 per 100 m VKTs being significantly more than sealed pavements of 6–7 m wide, which were found to have a crash rate of 6.9 per 100 mVKTs. This reduced further on wider pavements. ROR crashes increased significantly where a narrow pavement width (< 6 m) was at a location where the clear zone was less than 4 m. Narrow lanes (2.5 m wide) have a much higher crash rate than wider (3.5 m) lanes. Narrow shoulder widths of 0.5 m have been shown to have approximately 25% more crashes than wider 2.5 m shoulders.

A.4

Intersections

A.4.1

Type of Intersection

The form, layout and road geometry of an intersection can contribute to the safety performance of the intersection. The layout of the intersection can be complex with multiple approach legs, or skew angles of the intersecting approaches. Gattis and Low (1998) indicated that compared to 90° intersections, skew angle intersections may cause problems, including:

•

vehicles may require a longer time to cross, due to the increase in travel distance across the intersection. This may lead to a greater sight distance being required.

•

drivers making left or right turns may encroach into the opposing flow lanes

•

drivers undertaking a turn that only requires a small turning angle, may attempt the turn without slowing sufficiently

•

drivers making right or left turns may have more difficulty aligning their vehicles as they turn onto the cross-road

•

the driver’s sight distance may be obstructed by the body of some vehicles.

A study by Barua et al. (2010) on intersection crashes 3 found that there was an increase in crashes at offset intersections (i.e. where the opposing approaches are not aligned with each other and there is an offset between the approaches), at T-intersections on horizontal curves, and at intersections located on a sag curve or constant grade, as shown in Figure A 14. An offset intersection is one with a misalignment in the through movement, as shown in Figure A 15. (Note: the degree of misalignment was not nominated in the study).

3

This study was undertaken in Alberta Canada, and the crashes were found to be less severe on snow covered surfaces compared with dry surfaces.

Austroads 2015 | page 55

Road Geometry Study for Improved Rural Safety

Figure A 14:

Fatality risks at different types of intersections on rural undivided highways in Alberta Canada

Source: Barua et al. (2010).

Figure A 15:

Example of a misaligned intersection

Source: Barua et al. (2010).

Austroads 2015 | page 56

Road Geometry Study for Improved Rural Safety

The crash rates at the offset intersection apply to a situation where vehicles must turn onto and off the intersecting road, rather than making a straight crossing manoeuvre. Compared to cross-intersections, Barua et al. (2010) found that the offset intersection is almost four times the relative risk of a cross-intersection; and where horizontal curves were present, the relative crash risk was approximately 20% higher than the crossintersection; still well below the offset intersection. Intersections on horizontal curves have reduced sight distance due to the curve of the intersecting road, which may increase crash risk. Where an intersection is on a sag curve, the risk was also found to increase, and it was felt that this was due to the tendency for drivers to drive at higher speeds as they travel down the sag curve.

A.4.2

Superelevation through an Intersection

Superelevation may also impact crash rates, as superelevation may be reduced to accommodate the intersecting road, which may affect the lateral acceleration of the vehicle. Oxley et al. (2004) reported on the effect of reducing superelevation through an intersection to facilitate turning of entry vehicles from the minor road. It was considered that this practice may have reduced the safety for vehicles travelling along the major road, as the lateral acceleration forces change as the vehicle passes through the intersection. This was an issue particularly for trucks with their higher centre of gravity.

A.4.3

Intersection Sight Distance

Drivers need to observe that they are approaching an intersection, as other vehicles enter or leave the intersection, with enough time to respond, if necessary. If the available sight distance is limited, there is less time for the driver to respond to other vehicles. The restriction could be for the driver travelling along the main road, not being able to observe vehicles on the intersecting or side road; or for drivers on the side road, not being able to identify vehicles travelling along the main road. A study by Baldock et al. (2008) on rural crashes in South Australia, where an ambulance was called, found that crashes at uncontrolled intersections occurred mostly during daylight hours, which suggests that a likely factor was the failure of one driver to see the other. The most common factor identified to be contributing to this was the presence of vegetation or a crest. The presence of horizontal and vertical curves affects the available sight distance at an intersection. Whilst poor site distance is a crash risk (Pyta & Tziotis 2011), extensive sight distance can also be a crash risk as drivers anticipate a gap that might not be as big as it appears from a distance. A study by Charlton (2003) found that at a rural intersection in New Zealand, over a five-year period, of the 24 crashes that occurred, 23 were turning or crossing type crashes. Of the 23 crashes, 91% were from an approach that had an unrestricted view of on-coming traffic on the main road. It was hypothesised that drivers were anticipating the gap well before the intersection, as much as 100 m before.

A.4.4

Summary

The alignment of both through and turning lanes can make it difficult for drivers to align their vehicle to complete the manoeuvre, and results in drivers travelling into the opposing flow lanes. Misalignment of an intersection was found to be a significantly higher risk of crash compared to other types of intersections. The superelevation change through an intersection, to accommodate an intersecting road, can affect the lateral acceleration of a vehicle, particularly a heavy vehicle. The available sight distance at an intersection can be influenced by the form and layout of the intersection. Limited sight distance can be due to the presence of crests or horizontal curves. Vegetation can also contribute to the reduced sight distance.

Austroads 2015 | page 57

Road Geometry Study for Improved Rural Safety

A.5

Motorcycles and Heavy Vehicles

Some additional elements specific to motorcyclists and heavy vehicles are provided in the following sections.

A.5.1

Motorcyclists

In a study by Schneider et al. (2010) on the impacts of horizontal alignment on motorcyclists, four road design elements were identified that had a significant impact on motorcycle crashes:

•

length of the curve

•

curve radius

•

distance from the end of the curve

•

shoulder width.

There were over 30 000 curves within the study, ranging from 30 to 1750 m in radius, and 32 m to 386 m in length.

Length of curve Schneider. et al. (2010) reports that as the curve length increases, motorcyclists have an increasing risk, compared to shorter curves. A 1% change in curve length increased the crash frequency by 0.39%. Longer curve lengths are indicative of higher design speeds. This may indicate that the motorcycle riders may not be able to handle their motorcycle over these higher speeds. It may be, that when the motorcyclist travels around the curve, they tend to accelerate to a ‘high-speed instability’ compared to the straight sections.

Curve radius Crash frequency increases significantly the tighter the curve radius. Whilst this is an intuitive comment, it is supported by Austroads (2010e), which reports that generally, the majority of speed-related crashes occur on curved roads.

Distance from the end of the curve Whilst most crashes occur on curves, the curve appears to have an influence on the crash rates beyond the curve. For every 30 m beyond the curve, the motorcyclist crash rates decrease by 43%, up to a distance of 100 m.

Shoulder width The shoulder width was found to have an influence on the crash rates. Shoulder widths less than 1.80 m were found to have an increase in motorcyclist crash rates of 52%, over the wider shoulders.

A.5.2

Heavy Vehicles

Austroads (2009a) reports that a key issue that arose from a study of road crashes in rural and remote areas across Australia and New Zealand (Austroads 2005), was the need to improve heavy vehicle safety. It should be noted that heavy vehicles (rigid and articulated) have a fatal crash rate (fatalities per 100 million vehicle km travelled) of about three times that of all other vehicle types combined.

Austroads 2015 | page 58

Road Geometry Study for Improved Rural Safety

Forty-four sites were examined as part of the work to determine crash causation, and from this work, the geometric-related factors identified were: 1. at intersections:

–

poor sight distance

–

delineation either not provided or inadequate (i.e. line marking, raised reflective pavement markers (RRPMs), edge lines, and guide posts)

–

confusing road alignment (Australia)

–

pavement too narrow (New Zealand)

2. at mid-block sections:

–

unexpected transition between roads of varying standard (New Zealand)

–

poor sight distance for overtaking

–

confusing road alignment

–

steep embankments.

Lane widths For larger vehicles, A-triples and rigid plus three vehicles operating at 90 km/h, the 3.5 m wide lane was found to not be wide enough, allowing for tracking movements of the vehicle. These larger trucks need wider lanes, and this has been acknowledged in Austroads (2010d) which suggests that 3.7 m wide lanes be constructed to allow for the tracking of multi-combination trailers. It indicates that where there are even larger vehicles likely to use the road, even wider lanes be provided.

Sight distance Austroads (2010c) recognises the different performance characteristics of trucks over passenger vehicles and provides guidance on the stopping distances required for trucks. On horizontal curves, the higher driver eye height does not provide any advantage over the passenger vehicle.

Other projects A current Austroads project TS1607, Road Design for Heavy Vehicles; will also provide some guidance on actions to assist the design criteria for heavy vehicles.

A.5.3

Summary

For motorcyclists, it has been found that four road design elements had a significant influence on crashes – length of horizontal curve, curve radius, distance past the end of a curve, and shoulder width of the road. The elements that have been identified for heavy vehicles were sight distance at intersections or when overtaking, narrow pavements, confusing road alignment, and steep embankments.

Austroads 2015 | page 59

Road Geometry Study for Improved Rural Safety

New Road Design Elements A number of new standards and practices were identified for the road components. Some of these practices are considered to be in the early stages of their implementation and require further evaluation. The standards and practices identified are outlined in the following sections.

B.1

Horizontal Geometry

B.1.1

Curve Widening

The current design guide (Austroads 2010d) provides guidance on the application of curve widening as a means to maintain clearance between vehicles as they travel around a curve. The widening is provided to allow for vehicle overhang and the different tracking paths of the wheels.

Description The widening is most commonly provided on the inside of the curve, as shown in Figure B 1. Figure B 1:

Current curve widening on un-transitioned curves

Source: DTMR (2002).

An investigation by Levett (2007) on run-off-road crashes in New South Wales showed that 27% of those occurring on high-speed undivided roads were run-off-road to the left on right-turning bends, and 11% were run-off-road to the right on left-turning bends. Levett (2007), suggests that the difference between the crash rates of 16% could be due to the availability of the pavement and shoulder on the left-turning bends, for the driver to regain control on the opposing lane and shoulder, if they are not involved in a head-on crash first. Levett (2007) also reports that an analysis of fatal crashes for the period 2002 to 2004, indicated that 38% of off-road crashes (12% of the total off-road crashes) to the right on a right-turning curve, were initially a lossof-control on the left side of the curve with over-correction. Levett (2007), suggests that by providing treatment to the left side of the road, on the right-turning curve, the run-off-road crashes to the right would be expected to reduce.

Austroads 2015 | page 60

Road Geometry Study for Improved Rural Safety

Further, curves up to 600 m make up only 13% of the New South Wales state road network, and so make up a much larger proportion of the crashes compared with larger curves or straights. Crash reductions were found with widening on curves less than 1000 m radius. Levett (2007) suggests that reductions of up to 45% could be expected with the appropriate treatment to the left side of right-turning curves. Curves which have a radius of 200 and 600 m, have the majority of run-off-road crashes (see Figure B 2), and so it is logical that they should be considered for treatment ahead of the larger radius curves. Figure B 2:

Crash occurrences vs crash radius

Source: Levett (2007).

The sealing of shoulders provides safety benefits. Levett et al. (2009) indicates that any width of sealed shoulder will be beneficial compared to none. Previous work by Levett (2005) established that the most beneficial widening and shoulder sealing is on the outside of right-hand bend curves. Levett (2005) suggests that any curve with a radius less than 1500 m have a sealed shoulder of 2.5 m (at the curve apex) reducing to 1 m (at the ends of the curve). Figure B 3 and Figure B 4 show how the widening is applied.

Austroads 2015 | page 61

Road Geometry Study for Improved Rural Safety

Figure B 3:

Curve widening on a single curve

Source: Levett (2005).

Where there are a series of curves, Levett (2005) suggests that if the radii of the subsequent curves are less than 450 m, then they should also have 2.50 m wide shoulders. If there are just two bends in a reverse curve situation where both have a radius of less than 1500 m then they should both have sealed shoulders. Figure B 4:

Curve widening on multiple curves

Source: Levett (2005).

Effectiveness The reduction in crashes with the implementation of these treatments suggested by Levett (2007) would contribute to reducing the number and severity of run-off-road crashes by up to 45%.

Austroads 2015 | page 62

Road Geometry Study for Improved Rural Safety

B.1.2

Optical Speed Bars

Description Optical Speed Bars, see Figure B 5, are transverse stripes spaced at gradually decreasing distances. They are used to increase a driver’s perception of their speed, and so cause them to reduce speed. These transverse stripes have been installed 450 mm long and 300 mm wide. The spacing of the bars is based on the vehicle speed and the reduction trying to be obtained. The bars are spaced at four per second of travel speed. For example, commencing with a vehicle speed of 100 km/h and seeking a vehicle speed of 70 km/h, the bars start at four bars over 28 m, reducing to four bars over 19 m. The principle is that the spacing creates the perception that drivers have increased their speed and so they slow down. Figure B 5:

Optical speed bars

Source: McGee and Hanscom (2006).

A study by Arnold and Lantz (2007) placed optical bars at two separate locations, with the speed bars at one site similar to the layout shown in Figure B 5, and at a second site with the speed bars across the travelling lane as shown in Figure B 6. The installation similar to Figure B 5 was 1.05 miles long (1.68 km), and the second installation was one mile long (1.6 km). At both installations, the speed bars were 300 mm wide. Speeds were measured before, at the start, at the middle, and at the end of the bars. Speed reductions of 1.5 km/h to 5 km/h were achieved at the start of the bars, and vehicle speeds were found not to slow as they travelled across the bars. Arnold and Lantz (2007) concluded that the optical speed bars were effective in achieving small reductions in speeds. Placing the speed bars across the travel lane are more effective in reducing speeds than those that just extend a short distance from the centreline or edge line, see Figure B 6.

Austroads 2015 | page 63

Road Geometry Study for Improved Rural Safety

Figure B 6:

Optical speed bars across the travel lane

Source: Arnold and Lantz (2007).

Effectiveness The effectiveness of this type of treatment, in limited studies has found that there is some speed reduction of up to 8 km/h, but further investigation is needed regarding their long-term effectiveness. Optical speed bars are effective in reducing the speeds of vehicles approaching a hazardous roadway section, a reduced speed zone, or other roadway/travel change area. The reductions in speeds may be small. Optical speed bars that extend across the travel lane are more effective in reducing speeds than those that just extend a short distance from the centreline or edge line.

B.1.3

Summary

Widening of horizontal curves by 2.5 m, on the outer edge of the curve, to provide a surface for the driver that is travelling on the inside of the curve, provides significant safety improvements by reducing the number and severity of run-off-road crashes. Optical speed bars provide a perception of slowing by placing transverse bars at reducing distance/spacing along a section of road, and this treatment has been found to achieve small reductions in vehicle speed.

B.2

Intersection Treatments

B.2.1

Turbo Roundabouts

Turbo roundabouts are an alternative to a two-lane roundabout, which eliminates some of the conflict points on a normal two-lane roundabout. The turbo roundabout was developed by LGH Fortuijn in the Netherlands in the mid-1990s. A layout of a turbo roundabout is shown in Figure B 7. This type of roundabout was developed primarily for situations typically found on rural roads, i.e. unequal flows from the entering legs of the roundabout.

Austroads 2015 | page 64

Road Geometry Study for Improved Rural Safety

Figure B 7:

Turbo roundabout in The Netherlands

Source: Engelsman and Uken (2007).

An important element of the turbo roundabout is the spiral lane markings and raised separators that eliminate the possibility of weaving or lane changing within the circulating roadway. The lane configuration provides two entry lanes that are physically separated for the direction of travel, which then radiate away from the centre to become the lane for leaving the roundabout. With weaving being discouraged by providing raised lane dividers, traffic is kept in the lane and side-swipe type collisions that can occur when entering or leaving the a roundabout are prevented. Large vehicles can travel over the lane dividers if necessary, see Figure B 8. Note that the lane dividers have drainage gaps that would need to be carefully detailed and constructed to avoid vehicle instability or adverse driver response in correcting their vehicle off the divider. The size of a turbo roundabout is comparable to a two-lane roundabout, with a diameter of about 50 m. Figure B 8:

Large vehicle crossing the elevated lane divider

Source: Engelsman and Uken (2007).

Austroads 2015 | page 65

Road Geometry Study for Improved Rural Safety

Capacity To assess the capacity effects of a turbo roundabout, a quick scan model was developed, as described in Engelsman and Uken (2007) which showed the capacity of a turbo roundabout to be about 25% to 35% higher than the capacity of a two-lane roundabout, depending on the balance of the traffic flows on the approaches. The main reason for the higher capacity of the turbo roundabout is the reduction of conflict points for traffic entering and exiting the roundabout. Table B 1 shows the maximum entry capacity of a conventional twolane roundabout and turbo roundabout of two situations with different balances of traffic volumes on the approaches of the roundabout. Table B 1:

Maximum entry capacity of two-lane roundabouts and turbo roundabouts

Balance of traffic volumes on the approaches

Approximate maximum entry capacity** (in private car equivalents per hour) Two lane roundabout

Turbo roundabout

2300

3050

3200

4050

** As the results highly depend on the design of the roundabout and the driver behaviour factors used in the model, the results should only be interpreted as a comparison between the two-lane roundabout and turbo roundabout. Source: Engelsman and Uken (2007).

Safety Benefits Research in the Netherlands, as reported in Engelsman and Uken (2007) indicates that from a comparison between turbo roundabouts and traffic lights or yield intersections, a 70% reduction in collisions resulting in serious injuries can be expected. A turbo roundabout is an alternative form of roundabout that can provide additional benefits compared with a conventional two-lane roundabout, namely the reduction in conflict points from 16 to 10, and hence improved safety.

B.2.2

Jug Handle Intersection

A jug handle intersection is used to remove right turning vehicles from the major through traffic flows. Vehicles are diverted to the left prior to the intersection with the secondary or minor road, see Figure B 9.

Austroads 2015 | page 66

Road Geometry Study for Improved Rural Safety

Figure B 9:

Jug handle intersection layout

Source: Provided by VicRoads.

This type of intersection is used where there are a high number of right-turn crashes and/or the capacity of the intersection needs to be improved. It also improves the traffic flows along the major road by removing the right turning vehicles. The additional intersections need to be located beyond the queue length of the secondary or minor road, and so may require additional land to enable the slip lanes to be constructed. They also require additional pavements to be constructed and so there are ongoing maintenance costs to be considered. Property access is not possible along the major road, from the slip lane to the secondary road.

Safety Benefits A study by Jagannathan et al. (2006), over a 5 1/2 year period (1999 to 2004) in the USA, compared 44 jug handle intersections with 50 conventional intersections, and found that at jug handle intersections, the proportion of rear-end crashes increased from 42.7 to 56.6% of the total crashes; and left-turn crashes (equivalent to right-turn crashes in Australia and New Zealand) reduced from 11.8 to 3.7% of the total number of crashes. Further, head-on crashes decreased from 2.1 to 1.2% and fatal and injury crashes decreased from 33.8% to 30.2% of total crashes.

B.2.3

Intersection Narrowing

Amongst the causes of crashes, is the driver on the minor approach leg entering the intersection into a gap that is too short. The driver’s ability to judge and select a gap may be influenced by the approach speeds of the vehicles on the major road. A treatment to reduce the approach speeds along the major road is the placement of a painted centre median, to narrow the lanes. This is supplemented by rumble strips within this median and along the outside of the edge lines of the pavement. The major road lane width is reduced from 3.66 m to 2.75 m, and the median is typically 1.2 m to 1.8 m wide. A typical layout of this treatment is shown in Figure B 10.

Austroads 2015 | page 67

Road Geometry Study for Improved Rural Safety

Figure B 10:

Centre painted median with rumble strips

Source: Asokan and Bared (2009).

The rumble strips are applied on within the painted centre median and along the shoulder of the road. The lengths of the treatment vary with the posted speed limit and are shown in Table B 2. From Figure B 10, the first section A commences with the advance warning sign at its beginning; the second area, section B is the tapering of the lanes; and section C is the median at its full width. Typical lengths of the sections are shown in Table B 2. Table B 2:

Section lengths

Speed (km/h)

Section A (m)

Section B (m)

Section C (m)

70–85

30

60

45

100

45

60

45

The rumble strips, installed on the edge line commence 15 m prior to the end of the median to avoid turning traffic from travelling across the strips. On the edge lines, the rumble strips commence 15 m prior to the start of section B. This treatment has been applied at eight sites in the USA and it was found that the total number of crashes reduced by 32%, and fatal/injury crashes reduced by 34%. Whilst this treatment obtained these results, it still requires further assessment due to the limited number of sites that have been installed and examined.

B.2.4

Raised Platforms

As part of the Dutch ‘Sustainable Safety’ approach, platforms have been installed at intersections on 60 km/h local rural roads in order to reduce speeds to 50 km/h or less (the speed at which a side impact is likely to be survivable). These are installed to raise awareness, specifically at intersections that are on the boundary of different speed zones, and at intersections that are dangerous or potentially dangerous. An example of an intersection platform from a rural local road is shown in Figure B 11.

Austroads 2015 | page 68

Road Geometry Study for Improved Rural Safety

Figure B 11:

Rural intersection platform

Source: Turner et al. (2009).

This treatment has been placed on lower speed rural roads in The Netherlands and there has been no research on the effectiveness of the treatment for the rural application.

B.2.5

Reducing Sight Distance

Increasing sight distance is a common action undertaken to improve the safety and operation of an intersection. For some time, it has been hypothesised by some road design practitioners that the restriction on sight distance may also improve the safety of an intersection. Charlton (2003) provided an example of an intersection in rural New Zealand, where, on one approach to a staggered T-intersection, there was an unrestricted view of traffic travelling from one direction on the main route. Twenty-five crashes occurred at the intersection in a five-year period, with 24 of these involving crossing or turning movements. The majority of the crashes occurred in daylight and involved local drivers. It was thought that drivers were anticipating gaps in traffic at the intersection, as much as 100 m in advance of the intersection. This view was supported by the very short stop times at the intersection. Hessian screens were erected to restrict sight visibility at the intersection, beginning 125 m from the intersection and ending 25 m prior to the intersection (Figure B 12).

Austroads 2015 | page 69

Road Geometry Study for Improved Rural Safety

Figure B 12:

Reducing sight distance at an intersection

Source: Charlton (2003).

The approach speeds to the intersection, 25 m from the intersection, reduced from an average of 38 to 29 km/h, and the 85th percentile speed reduced from 50 to 39 km/h. Follow-up studies, 37 weeks after the installation, reported that crashes initially ceased at the site following installation of the screen. Whilst the treatment was felt to be effective, there were still plans to install a roundabout at the intersection. Austroads (2010e) further indicated that there had been three crashes at the site in the most recent five-year period from this approach, one of which involved a fatality. Charman et al. (2010) agreed that sight distances may be deliberately limited, but suggested retaining the normal, or design standard, visibility requirements on the approach to an intersection to increase uncertainty and slow the vehicles.

B.2.6

Summary

An alternative to a two-lane roundabout is a turbo roundabout. This type of roundabout incorporates spiral lane markings that radiate away from the centre leading to the departure lane, reducing the number of conflict points from 16 to 10. The capacity of the roundabout is greater than a conventional two-lane roundabout and a 70% reduction in collisions is expected. Jug handle intersections remove the right turning vehicles (in Australia and New Zealand) from the major road, through traffic flows, by directing them left to an intersection with a secondary road. This type of intersection improves the capacity along the major road and reduces the right turning against type crash, but the rear-end crashes increase. Reducing the approach speeds along the minor road approach to an intersection can be achieved by narrowing the lanes. This can be achieved by the installation of painted traffic islands, with rumble strips placed within the island. This type of treatment has resulted in a reduction in crashes of over 30%. Raised platforms have been installed on 60km/h rural roads, but their effectiveness is still to be determined. Increasing sight distance is a common treatment at an intersection; however, reducing the available sight distance has also been demonstrated to provide speed reduction benefits.

Austroads 2015 | page 70

Road Geometry Study for Improved Rural Safety

B.3

Cross-section

B.3.1

Single Lane Cross-section, 2 – 1 Roads

A two minus one road is typically a two-lane, two-way local road that has had the travelling lanes reduced from two lanes to one lane. This treatment was trialled in Denmark on a 6 km long section of a 5.8 m wide, local, two-lane road. The road carries 2500 vehicles per day, of which 8 to 10% are trucks. The pavement was marked with a 3.5 m wide single travelling lane, with 0.3 m wide intermittent edge lines, which left 0.85 m of pavement, on both sides retained for use by cyclists and to allow approaching vehicles to pass. Figure B 13 shows the lane arrangements. Figure B 13:

Two minus one road

Source: Herrstedt (2007).

The two minus one treatment was supplemented by 12 road narrowing points, which included edge islands and bollards, see Figure B 14. Figure B 14:

Road narrowing within the two minus one treatment

Source: Herrstedt (2007).

Austroads 2015 | page 71

Road Geometry Study for Improved Rural Safety

Within the length of this treatment, the posted speed limits were reduced to 40, 50 and 60 km/h. Within each speed zone, the edge markings were marked as follows:

•

60 km/h section: 2 m strip with a 2 m gap

•

50 km/h section: 1 m strip, with a 1 m gap

•

40 km/h section: 0.5 m strip, with a 0.5 m gap.

Effectiveness The evaluation of the speeds, as reported by Herrstedt (2007), indicated that speeds were as follows. Table B 3:

Speeds from 2 – 1 lane treatment

Speed Limit

Mean speed

85% speed

40 km/h

53–57 km/h

63–68 km/h

50 km/h

60–65 km/h

69–76 km/h

60 km/h

69–70 km/h

78–81 km/h

Herrstedt (2007) concluded that the treatment did not reduce speeds as was intended with the high speeds occurring on the road. (Note - no speeds prior to the installation were reported). It was also found that drivers behaved as intended when they met, both moving to one side of the road allowing each other to pass. At the narrowing points, there was some driver confusion as to which driver had priority, with some drivers just travelling through the narrowing, forcing the other driver to give way.

B.3.2

Lane and Shoulder Combinations

A study by Amjadi (2009) was undertaken in the USA to evaluate the relative safety benefits of different lane widths and shoulder widths for existing pavement width combinations on two-lane undivided rural roads. The study obtained data for six sections of road that had total pavement widths between 7.92 to 10.97 m (i.e. lane widths plus shoulder widths). The lane widths were 3.05 m, 3.35 and 3.66 m, and the resulting shoulder width varied from 0.3 m on the 7.92 m pavement width to 2.44 m on the 10.97 m pavement width. The study found that for:

•

9.75 m pavements, lane widths of 3.66 m (1.22 m shoulders) had a crash rate 3 to 6% lower than the lane widths of 3.05 m (1.82 m shoulders)

•

10.36 m pavements, lane widths of 3.35 m (1.83 m shoulders) had a 22% lower crash rate, compared with pavements with 3.05 m (2.13 m shoulders) lane widths, and lane widths of 3.66 m (1.52 m shoulders) had a crash rate 19% lower than the 3.05 m lane widths

•

10.97 m pavements, lane widths of 3.35 m or 3.66 m (2.14 and 1.82 m shoulders) were found to have a 5% lower crash rate compared with the 3.05 m lane width (2.42 m shoulders).

The study results were found to generally apply to roads with annual average daily traffic (AADT) volumes of more than 1000 and speed limits of greater than 40 km/h. The conclusion reached by Amjadi (2009) was that for existing pavement widths, the allocation of the lane width and shoulder width can be a cost-effective treatment to reduce crashes on two-lane undivided roads, with the wider lane having a small advantage over widening of the shoulder.

Austroads 2015 | page 72

Road Geometry Study for Improved Rural Safety

B.3.3

Medians

Due to the geographical constraints of rural roads, a physical median, or other physical means of lane separation, is not always possible. However, head-on and off-to-the-right are the most severe types of crashes on rural roads in New South Wales (Levett et al. 2009). In many cases, a solid painted centreline that prohibits overtaking is painted on the road surface. By having two centrelines and separating them, a painted median is created. The width of the median is often determined by the available width of pavement. By assessing the safety benefit of different widths of painted medians, Levett et al. (2009), determined that the wider medians had a lower crash risk and that painted medians should be at least 1 m wide. Levett et al. (2009) also suggests that the effectiveness of the median would be enhanced by using audio-tactile line markings.

Wide Painted Medians Levett et al. (2009) examined a number of measures to address two of the most severe types of crashes on the NSW road network, off-to-the-right crashes and head-on crashes, not overtaking crashes. The study examined five types of centrelines: standard centreline and barrier line (control sites), 0.5 m wide painted median, 1.0 m wide painted median, 0.5–1.0 m wide profile (audio-tactile profile) painted median, and 2.0 m wide wire rope median barrier. The 0.5 m wide painted median installation on the Pacific Highway had some effect on the total crashes, which fell from 41 to 25, but had no effect on crossover crashes (11 before treatment, and 10 after). Levett et al. (2009) provided a comment that the 0.5 wide median did not provide adequate separation of the opposing flows to allow for driver error. The 1.0 wide median installed on the Pacific Highway resulted in crashes reducing from 21 to 8, with the crash severity (i.e. fatal and injury crashes) also reducing from 8 to 4. Crossover crashes reduced from 9 to 4. The wider separation was considered to provide a positive effect on the crash rates and crash severity outcomes. The crash data at the 0.5 to 1.0 m wide profiled painted median installations showed a reduction in total crashes from 19 to 5, and crossover crashes from 7 to 1. The width along the sites varied between the 0.5 and 1.0 m width, and Levett et al. (2009) commented that varying of the width seemed to make little difference to the crash rates and that to maximise the benefit, the profile median should be at least 1.0 m wide. The 2.0 m wide median with a centre wire rope barrier achieved a reduction in total crashes from 46 to 33, and crossover crashes from 23 to 22. Prior to installation of the treatment, from the 23 crossover type crashes, 12 were fatal or injury and 11 only required the vehicle to be towed away. Following installation, of the 22 crossover type crashes, 2 of the crashes were fatal or injury crashes and 20 only required the vehicles to be towed away, with the vehicles crashing into the wire rope barrier. Levett et al. (2009), concluded that a painted median needs to be at least 1.0 m wide to maximise the benefits of reducing crossover crashes, and that this treatment could be enhanced by the addition of an audio-tactile profile in the line marking. A review of local and international literature by Austroads (2012b) showed a casualty crash reduction factor of 15% for rural and urban environment. A more recent, preliminary evaluation of Queensland 1 m rural road painted medians on Bruce Highway by Whittaker (2012) reported a reduction in severe hear-on and crossover centreline crashes of 75%. Run-off-road to the left and total severe crashes were reduced by 59%. The treatment was undergoing a full evaluation in 2015.

Austroads 2015 | page 73

Road Geometry Study for Improved Rural Safety

B.3.4

2 + 1 Roads

A 2 + 1 road is a two-lane road with a third lane alternating in flow direction, to allow vehicles to pass. This arrangement provides a safe overtaking lane on each direction of flow at various sections along a road. The opposing flows are separated by a safety barrier system. The 2 + 1 arrangement has been found, according to Derr (2003), to operate effectively when the traffic volumes are between 15 000 and 25 000 vehicles per day. In Sweden, which has led the installation of this treatment, the pavements typically are 13 m wide comprising three 3.25 m lanes, a 1.25 m wide median and 0.75 m shoulders with a median barrier. Sweden has installed 2100 km of this treatment, with an 80% reduction in fatalities (personal communication from Torsten Bergh, Swedish Transport Administration, 5 December 2011). Figure B 15:

2 + 1 road – Sweden

Source: Picture provided through personal communication with Torsten Bergh, Swedish Transport Administration (2011).

The passing lanes, in Sweden are typically 1 to 2.5 km long, see Figure B 16, and are spaced 4 to 10 km apart, see Figure B 17. Figure B 16:

2 + 1 road passing lane

Source: Personal communication Bergh presentation (2011).

Austroads 2015 | page 74

Road Geometry Study for Improved Rural Safety

Figure B 17:

2 + 1 road passing lane spacing

Source: Personal communication Bergh presentation (2011).

In Sweden, the 2 + 1 treatment has achieved an 80% reduction in fatal crashes using a wire rope barrier as the median barrier (personal communication from Torsten Bergh, Swedish Transport Administration, 5 December 2011).

B.3.5

Summary

Reducing a two-lane road to one lane, known as a 2–1 road, has been installed on a 6 km section of road, carrying 2500 vehicles per day. The treatment was aimed at reducing speeds, but speed reductions have not been obtained. The allocation of lane and shoulder width on existing pavements can influence the crash rates, and it has been found that providing wider lanes has a small advantage over providing wider shoulders. The provision of a wide painted median, installed at 0.5 m wide, 1.0 m wide and 2.0 m wide with a centre wire rope barrier has achieved crash reductions for all of these installations. The 0.5 m wide median reduced the number of crashes, but did not reduce the crossover type crashes. The 1.0 m wide median was found to reduce the number and severity of the crashes; and with the 2.0 m wide median with wire rope barrier, the number of crashes reduced but crossover crashes remained almost constant, however these crashes were predominantly property damage type crashes, with vehicles colliding with the barrier. A 2 + 1 road is a three-lane road with the provision of alternating passing lanes. The passing lanes in Sweden are 1 to 2.5 km long, spaced between 4 and 10 km apart, with the opposing flows separated by a barrier. This treatment provides opportunities for passing for each direction of flow. This treatment has been found to provide significant reductions in fatalities.

B.4

Gateway Treatments

B.4.1

Introduction

Gateway treatments are a single treatment or a combination of treatments, used to slow vehicle speeds as they enter a slower speed environment. They are usually found on the approaches to rural towns. Treatments to establish a gateway include the use of pavement and/or lane narrowing, traffic islands, a different coloured and/or textured pavement surface, roadside signing, and roadside planting to create a visual narrowing. An example is shown in Figure B 18.

Austroads 2015 | page 75

Road Geometry Study for Improved Rural Safety

Figure B 18:

Gateway treatment in the UK

Source: Austroads project ST1426.

Austroads (2010c) reported of treatments in the UK, at the entry points to a village to support the village speed limit of 30 mph (48 km/h) from an approach road speed of 60 mph (96 km/h). Here, visual treatments such as coloured road surfaces, visual narrowing and large roadside signs, reduced 85th percentile speeds by 11 km/h, and this reduction increased when physical restrictions were also installed, with speeds reducing by about 16 km/h. If a gateway treatment is not accompanied by changes further along the road, such as decreases in road width or speed maintenance countermeasures, speed reductions produced at the gateway may dissipate within 250 m (Austroads 2010c). Based on a survey of international researchers and practitioners, Charlton and Baas (2006) reported that:

•

for lower speed transitions, physical measures employing build-outs (curb extensions), speed tables (flat-topped road humps), and changes in road surface texture and colour were identified as most effective, sustainable, and suitable

•

for speed change thresholds at higher speed profiles, perceptual measures using edge lines, hatching, angle parking, and landscaped central islands received the highest ratings.

B.4.2

Pavement Marking

Pavement markings have been used as gateway treatments. The treatment includes the use of converging chevrons on pavement speed markings and lane narrowing.

Converging Chevrons The chevrons are placed with a decreasing space between them so that drivers obtain a perception of travelling faster than they actually are and therefore slow down.

Austroads 2015 | page 76

Road Geometry Study for Improved Rural Safety

Figure B 19:

Converging chevrons

Source: Hallmark et al. (2007).

Speed Marking and Lane Narrowing The placement of chevrons at regularly spaced intervals reminds drivers of the speed limit. The treatment shown in Figure B 20 also includes lane narrowing to create the perception of constricted space, and so drivers tend to reduce their speeds. Figure B 20:

Pavement speed marking and lane narrowing

Source: Hallmark et al. (2007).

Safety Benefits Hallmark et al. (2007) report that based on a study in Iowa, where chevrons were placed on the approach to a town with a speed limit of 55 mph (98 km/h), a reduction in the 85th percentile speed of 1 mph (1.6 km/h) was achieved.

Austroads 2015 | page 77

Road Geometry Study for Improved Rural Safety

B.4.3

Summary

Gateway treatments are typically employed where a speed zoning changes from a higher speed to a lower speed. The gateway can be a single treatment or a combination of treatments that can be established with the installation of an island (raised or painted), pavement markings to narrow the lane widths, or coloured pavement surfaces. These treatments have been found to reduce speed, but need to be accompanied by further changes along the road to retain the reduction.

B.5

‘Self-explaining’ Roads

The self-explaining roads concept is based on the driver’s perception of the road to influence their behaviour. By having a consistent application of standards, drivers respond and modify their behaviour to align with their perception of the road environment, which results in speed choices consistent with the safe speed for that function and design. Recognition of the current road function, and to predict road elements, requires the following three features:

•

clear marking and signing

•

recognisable road categories

•

design elements (e.g. cross-section width) for each road category that are uniform within the road category.

Having a limited number of distinct road classes and making these obvious to the driver by having markings, signs, and road or roadside elements specific to the type of road, is essential to enable road users recognise the class of road they are travelling on and to be aware of the speed limit on that road. Therefore, adoption of a self-explaining road approach typically involves assigning roads to a newly developed set of road categories, and then implementing changes in order to make these categories discrete but uniform. Not only should the road have features that clearly indicate to motorists the type of road they are on, but it should also act implicitly (or subconsciously) to control the behaviour of motorists. In terms of speed management, Charlton and Baas (2006) suggest that these features could include use of median and edge line treatments, access controls, road markings, pavement surfaces, and roadside furniture. The important characteristics to be achieved to support this concept are:

•

road functional classification

•

expectancy or predictability

•

road classifications need to be recognised with different road classifications having a different function that can be recognised by drivers.

The characteristics of self-explaining roads, as indicated by Theeuwes and Godthelp (1995) include:

•

unique road elements (homogeneous within one category and different from all other categories)

•

unique behaviour for a specific category (homogeneous within one category and different from all other categories)

•

unique behaviour should be linked to unique road elements

•

the layout of crossings, road sections and curves should be linked uniquely with the particular road category

•

where there is a transition in road category, the change should be clearly marked

•

the road design should expel speed differences and differences in direction of movement

•

the design standards of the road are aligned to the operating speed of the road

•

road marking and signing is to the same standard.

Austroads 2015 | page 78

Road Geometry Study for Improved Rural Safety

The road characteristics include the presence of medians, pavement markings (including edge lines), access controls from abutting properties, pavement surfaces, signing, and other street furniture.

Austroads 2015 | page 79

Road Geometry Study for Improved Rural Safety

The following DRAFT Commentary is intended to invite discussion and eventually capture the existing state of practice in Austroads guidance review. It is proposed that it may form future informative, high-level guidance for road experts new to the field of guideline revision. The Commentary discusses the general reasons and processes applicable to Austroads guidance review. It then discusses the different types of evidence which are used in reviewing specific guidance.

C1.1

Reasons for Road Design Guidelines Review

The need for road design guideline review is typically driven by one or more of the following factors.

Update The most common reason for reviewing road design guidelines is the feedback generated by road designers and, more indirectly, by road construction practitioners. This would involve updating of the guidelines to incorporate:

•

design/technical notes developed since the last revision

•

new research findings and methods

•

changes to standard drawings representing adopted practices and construction solutions

•

harmonisation of guidelines across jurisdictions

•

correction of identified inconsistencies.

Policy Change Many revisions of road design guidelines are initiated by changes in the scope of road agency business, or the way in which this business is conducted. These may include design guideline changes following from response to external (e.g. parliamentary) or internal reviews of practice. Examples of such include simplification of speed limit setting, extended design domain (EDD), or adoption of Safe System principles. Sometimes, a revision may be necessitated by a broader government policy, such greater inclusion of cyclists in the road transport system. This necessitated a gradual development of guidelines for provision of on-road bicycle lanes, and specialised management options at intersections and road crossing locations.

Input from Other Professions Another trigger for road design guidelines reviews comes from inputs referred to the road design policy makers by transport planners, network operators, road safety practitioners, legal professionals (e.g. coroner’s court, legal team), or police. In many cases, this occurs in the form of road designers’ response to emerging pressures such as legal liability risk management, provision of equitable access to the transport system, public transport integration, and removing undesirable design options. For this collection of reasons, road design guidelines should be revised periodically, e.g. every 3–5 years.

C1.2

Road Design Guidelines Review Process

In the majority of cases, changes in road design guidance is made by consensus of senior road design policy makers with inputs from practitioners and other stakeholders, e.g. asset managers and road safety experts.

Austroads 2015 | page 80

Road Geometry Study for Improved Rural Safety

In some cases, decisions may be made by consensus of a small group of relevant managers reviewing available evidence. In other cases, there may be a much greater focus on consultation with internal and external practitioners during the review process. This promotes ownership of the proposed changes in the industry. Either way, reaching consensus on the net benefit of road design guidelines revisions is a crucial part of the process. The process generally followed in revision of a typical Austroads guide involves:

•

Establishment of an advisory group, typically consisting of senior design practitioners and policy makers.

•

Collation and review of evidence, e.g.:

–

jurisdictional supplements

–

feedback on the existing guide from industry practitioners

–

findings from recent Austroads and other relevant research (e.g. internal reports, international publications).

•

Advisory group discussion of new evidence and implications for the guide content.

•

Preparation of a working paper listing recommended changes to the guide based on the outcomes of evidence review and discussions. Identifying knowledge gaps for future research.

•

Preparation of a draft update to the guide.

•

Obtaining comments of the advisory group and finalising the guide revision.

It is important that evidence for a review is sourced using more than one method. Using multiple sources of evidence on any one subject will provide a stronger and more defensible basis for adopted changes. Choice of methods depends on the subject of a review. Each method has its own strengths and limitations. The individual methods are discussed in Section C1.2.

C1.3

Evidence for Review of Specific Road Design Guidance

This part of the discussion deals with different methods of generating evidence to consider in revision of specific road design guidance. The subject of such guidance may be a given design element (e.g. selection of design speed, or curve transitions), contents of a design table/graph (e.g. recommended sealed shoulder widths), or even a specific value (e.g. a standard size of a speed limit sign). There is sometimes difficulty in obtaining such evidence. As explained in Section C1.2, it is important that the reviewers source evidence using more than one method. An outline and discussion of these methods is contained in the following sections.

Reviews of Other Guidelines One of the most common inputs into revision of specific design guidance is to refer to other existing guidelines, supplements or technical notes for alternatives. In many cases, this is done as part of harmonisation of design guidance across jurisdictions. Sometimes, a cost-effective way of managing an emerging design issue in one road agency is through adoption of established design guidance from another. One way to approach the task is to collect several of the most relevant guidelines dealing with the subject (e.g. from Australia, New Zealand, USA, and UK) and extract the relevant sections for analysis and further discussion. The reviewers may look for:

•

suitability of available guidance in their own context, e.g. complexity, ease of application, compatibility with other existing guidance, local availability of supply

•

evidence of successful application, e.g. cost-effectiveness in addressing the issue at hand, no adverse effects on safety or traffic flow performance

•

harmonisation with other jurisdictions, e.g. using the same sign face designs.

Austroads 2015 | page 81

Road Geometry Study for Improved Rural Safety

A significant advantage of this method is when effectiveness of the guidance was formally evaluated in one or more jurisdictions. This provides a degree of confidence that the approach may bring the desired outcomes in other jurisdictions. Industry practitioner feedback is also valuable in this method. Feedback draws attention to aspects of current guidance which may cause application issues, and may sway the choice of the adopted design guidance. A drawback of only reviewing other guidelines may be entrenchment of consensus on design guidance with outdated or minimal evaluation evidence. This occurs when circular references make the existing design guidance self-affirming. An example of this is when one road agency sets a design value based on accepted practice/opinion, a second agency adopts it in a later review, and then the first agency references the second agency’s guidance assuming that it was based on a recent and thorough revision. The key to addressing this issue rests in seeking out the primary sources of the evidence behind any guidance and in using multiple revision methods, e.g. review of research.

Audit Feedback Road safety audits, design compliance audits, and operational audits provide a potential source of information on where road design fails to meet its objectives. They may highlight practice areas where the specific design guidance is not practicable, or needs strengthening. Examples of this may include consistent recurrence of road safety audit recommendations relating a combination of several design values/solutions and its impact on crash risk. Audit data availability and analysis are the keys to making such audit information useful as evidence in revision of guidance. Collating audit recommendations into a database would enable numerical analysis of trends and priorities. To date, there have been limited opportunities to do so. Use of experienced auditor focus groups and drawing on the collective experience may provide an alternative solution.

Review of Published Research A major input into revision of specific design guidance is a review of the available research on the subject. The reviewers of design guidance need to collate the available research and review it for relevance to the subject. The most common process of doing this involves a researcher identifying relevant studies and extracting the research evidence needed to answer a specific question based on the design guidance being revised (e.g. ‘what is the safety-optimal lane width on high-speed roads?’). Often multiple questions need to be answered across several research fields in order to make a balanced decision. Austroads research and technical project reports can provide a compendium of recent findings relevant to the revision. Such projects are typically generated from the previously identified knowledge gaps. Literature search of locally and internationally published research frequently provides valuable insight into practice in other countries. Overseas research can sometimes provide the depth of investigation not easily attainable in Australia and New Zealand (e.g. large sample sizes). Unpublished studies and trials carried out by individual road agencies can often provide valuable insight into specific issues. Industry practitioner feedback is valuable in this method, as new published and unpublished studies and trials emerge frequently and may not always be captured in literature searches.

Austroads 2015 | page 82

Road Geometry Study for Improved Rural Safety

It is important to question relevance, applicability and robustness of findings when reviewing research literature with a view to extract evidence about specific design guidance. Overseas evidence relating to effectiveness of a specific design may or may not be relevant to local conditions. An example may be a finding of a study on barrier effectiveness from a jurisdiction with low seat belt wearing rates. Such findings would have limited relevance in Australia where seat belt wearing rates are high. Applicability of a European semi-flexible barrier crash test may be questionable if such a product is not likely to be available in Australia and New Zealand. Robustness of findings needs to be questioned with each reviewed study. Typically, standard error, confidence intervals, p-value, and sample sizes are a good indicator of how confident one can be in the findings. Studies which do not provide such information should be treated with caution. The listed or implied study limitations can tell us a lot about the relevance and applicability of the findings to local conditions.

Before and After Studies Revision of specific design guidance may be informed by results of a before and after study. Such a study needs to have a clearly stated research question based on the design guidance being revised. It needs to have at least one quantifiable indicator of performance, e.g. crash numbers, percentage of day with traffic congestion, or unit maintenance cost. There are many ways in which a before and after study can be designed. Austroads (2012a) provides a good introduction to different methodologies from Empirical Bayes (highly regarded) to naïve (not recommended). Before and after studies test for the change in a given indicator caused by the treatment, e.g. a specific design element or its value. The basic before and after study methodology involves comparing indicator data from a set of locations for a time period before a treatment was introduced with a period after its introduction. It is important to control the change in indicator data for any broader effects, e.g. increased enforcement or season. To do this, a number of control locations need to be selected. These locations have to be similar to the treatment sites. The indicator data needs to be collected in four categories, as shown in Table C1 1. Depending on the indicator type, the before and after data periods may range from several weeks (speeds, traffic volumes) to several years (crashes). If the before and after periods are of different length, the results can be compared as rates (e.g. crashes per annum). Table C1 1: Data collection needs for a before and after study Before

After

Treatment locations

XX

AA

Control locations

YY

BB

In some types of studies, the indicator data may come entirely from historical records held by the road agency, e.g. crashes, operating speeds. Effectiveness of a treatment is typically calculated as the ratio of the indicator value after the treatment, to the expected value of the indicator if the treatment was not applied. Austroads (2012a) provides more detail on this calculation, together with some case studies. Before and after studies are useful in revision of design guidance because they can demonstrate if a given solution is effective in the road agency in which it will be applied. The studies can be very specific, e.g. intersection approach speed reductions in the presence of vehicle-activated signs or effect of shoulder sealing on casualty crashes. A robust finding may lead to adoption, rejection or further refinement of such treatment as part of specific design guidance. This can be an advantage over less specific or less robust evidence from international research.

Austroads 2015 | page 83

Road Geometry Study for Improved Rural Safety

Before and after studies offer an explanation of causality, i.e. a different design value would cause the desired change. This is an important advantage over other methods, and robust studies provide confidence of repeatability of results in similar conditions. On the other hand, the specific nature of such studies is also one of their limitations. A study carried out in one jurisdiction may need to be thoroughly scrutinised for relevance and applicability in another, e.g. due to different climatic and geographic conditions. Another aspect relevant to revision of road design guidance is that many before and after studies are based on sites where a treatment was applied for a reason, e.g. poor crash history. This is referred to as selection bias. It means that we cannot be certain that the results would be the same if the treatment was applied as a design variable across the entire road network. For example, it is unlikely that provision of a fully-controlled right-turn at an average traffic signal site would result in the same crash reduction as at targeted black spot sites which made up the study sample. Another issue with before and after studies arises from the treatment application. Rarely is a single road design element changed without consequentially affecting other elements. Typically several road design variables are modified. When sealing shoulders, for example, pavement quality, line marking and culvert headwalls are upgraded, but not equally at all treated locations. This weakens the relevance of findings about the selected design element, as the evaluation includes a random selection of other associated treatments. As shown by this study, availability of treatment data can be a serious issue when carrying out before and after studies. Road agencies generally do not have a central depository of completed project data, such as as-built drawings. Details of treatment at each location need to be retrieved from project files and design archives in scattered offices. Insufficient number of identified sites can challenge robustness of results or even the viability of a study. Also, many lesser treatments are not subject to detailed designs, e.g. line marking changes, safety barrier installation. This further complicates identification of similar treatment locations for a study.

Frequency Analysis Some road design guidance revision may be informed by frequency analysis of one or two selected performance indicators. There are many variations of this approach, and some examples include analysis of:

•

distribution of crashes vs a crash data attribute, or road attribute. This may draw attention to crashes being over-represented on parts of the road network with a given characteristic where road design may need to be strengthened

•

observed speeds for a given road type or road feature. This may lead to revision of such design parameters as design speed, or an operating speed model (e.g. Austroads 2013a)

•

distribution of driver eye height among the driver population. Such analysis may inform an update of the procedure for sight distance measurement at intersections

•

costs, e.g. which design scenarios for a given feature regularly result in high maintenance costs.

An example of such analysis is shown in Figure B 20 based on Donnell et al. (2002) who reported on an investigation of cross-median crashes on interstate motorways in Pennsylvania, USA. Some 57% of the crashes were found to have occurred at locations where the median was wider than the 15 m AASHTO warrant at the time. This analysis led to questioning the existing design guidance on median widths, and the eventual development of new guidance for application of median barriers.

Austroads 2015 | page 84

Road Geometry Study for Improved Rural Safety

Figure C1 1:

Distribution of cross-median crashes by median width and ADT

Source: Donnell et al. (2002).

Figure C1 2 shows a graphed curve departure operating speed vs. curve radius, given a 100 km/h curve approach operating speed. The figure is an example of analysis of measured field data parameter (operating, or 85th percentile speed) used to inform revision of current guidance, in this case the operating speed model (Austroads 2013a). The analysis shows that the observed performance of the road network (red line) differed from that in the past (dashed blue line), and the specific design guidance required an update. Figure C1 2:

105

Revised and existing operating speed models (100 km/h approach speed)

Small radius

Medium radius

Large radius

Curve departure speed (km/h)

100 95 90 Curve midpoint speed (all cars)

85

Curve midpoint speed (banded)

80

Revised deceleration on horizontal curves model

75

Existing deceleration on horizontal curves model

70

50

100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Radius (m)

Source: Austroads (2013a).

Austroads 2015 | page 85

Road Geometry Study for Improved Rural Safety

One of the main advantages of using the frequency analysis method is that in many cases, the data is either available or can be easily collected from the road network. The method does not require expert skills and equipment. Interpretation of findings can be done by most experienced practitioners and policy makers. A serious limitation of this method rests in hidden biases and correlations between the main variable explaining the performance indicator and other variables which are not considered. Hence, this approach should be used with caution, and where possible, the analysed data should be controlled for exposure, speed limit, road type and other key variables to reduce this limitation. The main way of overcoming the bias and correlation problem is through statistical regression modelling the interaction between the performance indicator with various road design elements. This is described in Section C1.3.6.

Statistical Regression Models This way of generating input into specific road design guidance involves analysing the behaviour of an important road performance indicator 4 in the presence of multiple road design and operational variables 5. It shows the effect of changing the value of a specific design variable, and provides a measure of robustness of this evidence (typically the p-value). Increased availability of road data and statistical software means that modelling is emerging as a more common form of input into revision of road design guidance. There are many mathematical methods which can be applied and depend on the type and availability of data, the level of statistical modelling skills, and the available software tools. Typically, the main steps involve:

•

gathering of road performance indicator data, e.g. crash frequency, crash severity, speed, delay, congestion

•

gathering road design, operational and other variables, e.g. speed limit, AADT, road curvature, pavement width, pavement friction, intersection design parameters, signal phasing, etc.

•

associating the information with individual road segments, lanes or intersections, often by application of geocoding techniques

•

selection of model variables – this typically involves testing for individual impact of each variable on the indicator (the weakest are removed), and a check for cross-correlation between variables (correlated variables are removed)

•

testing the statistical distributions of the remaining design and operational variables. Often this suggests aggregation of values or conversion into ranges in order to discover a meaningful relationship

•

choosing a statistical modelling technique depending on the distribution of the performance indicator; for crash studies, this is typically a Poisson or negative-binomial model

•

running the model, often involving several revisions

•

interpretation of the final results from their mathematical format (e.g. model exponents) to design variables (e.g. widths, distances) and interpretation of the meaning with respect to design guidance.

The tabulated results for a given road design variable can be interpreted as the progressive effect of changing that variable on the main indicator (e.g. crash modification factors), all other variables being held equal. Also, the effect can be quantified, e.g. as expected crash numbers per unit length of road. If the unit cost of changing the variable is known (e.g. providing an extra metre of pavement width), there is good basis for analysis of economic impacts of a change in design guidance. Warrants for application of the revised guidance can be derived from the economic viability of desirable change at different AADT values.

4

Called the ‘dependant variable’ in statistics.

5

Called ‘independent variables’.

Austroads 2015 | page 86

Road Geometry Study for Improved Rural Safety

Robustness of the findings is immediately clear as statistical software produces a range of relevant parameters, such as p-values or confidence intervals. The main advantage of this method is that jurisdictions can use their own road system data to generate information about its improvement. National models can be prepared based on pooled data from several jurisdictions. Typically, the method relies on existing road information rather than collection of new data. Hence, this method can be faster and more cost-effective than before and after studies. The main criticism of using a statistical regression method to inform specific road design guidance revision is that it highlights association between variables, but does not imply causality. Hence, it cannot be said that ‘a different design value would cause the desired change’. Rather, it can be said that ‘it is observed that a desired change occurs in the presence of a different design value’. Thus, interpretation of findings requires a degree of experience, and factors that could not be modelled need to be considered qualitatively (e.g. future changes in the vehicle technology, different levels of enforcement, etc.). Another limitation of this method is that it uses a cross-sectional approach to analysis, i.e. the results represent performance of the road system from which the data came, relevant to the known time period. Relevance of the findings to other road systems, even within the same jurisdiction, needs to be carefully assessed. Notably, the same limitations on relevance apply to before and after studies. Applicability is limited to design solutions already in place for which there is data that can be collected and regressed. Extending the relationships beyond the range of the available data should be done with caution, and only if there is additional supporting evidence. For example, it is unlikely that making shoulders progressively wider and wider would continue to result in further crash reductions.

Theoretical Models Where no road data exists because there are no examples of a given design treatment (e.g. a new solution), or such data is scarce, analysis of design performance may be sought from theoretical models. Such analysis may then inform development of the specific design guidance. There are many different methods which can be employed for this, but most fall into these three groups:

•

Gap acceptance and reliability theory (probability of failure) studies. These are based on the relationships between variables based on observed data and assumed mathematical relationships. Such studies can provide an indication of crash risk, delay, or level of service under untested design scenarios over a range of performance conditions. These types of studies often use iterative methods such as Monte Carlo analysis. Some past examples used in road design guidance include development of minimum lengths of overtaking lanes (Llorca et al. 2014).

•

Modern decision support tools combine many empirically-developed relationships to generate practical advice on a range of real-world technical situations. Some are in the form of documented methodologies, others in various forms of software. They are sometimes used to provide input into design guidance based on economic viability or crash reductions expected from establishing new or changing existing specific guidance. There are many examples of such support tools, e.g.:

•

–

Highway Safety Manual (AASHTO 2010)

–

Interactive Highway Safety Design Model (IHSDM) (Federal Highway Administration 2014)

–

Roadside Safety Assessment Program (RSAP) v3 (Roadsafe LLC 2012)

–

Australian Level Crossing Assessment Model (ALCAM)

–

ANRAM (Austroads 2014b).

Finite element analysis can be used to predict mechanical responses to typical use or impact, providing initial guidance on some design parameters (e.g. safety barrier design).

Austroads 2015 | page 87

Road Geometry Study for Improved Rural Safety

Figure C1 3 shows part of the analysis done in Florida using RSAP v3 to revise design guidance on application of guardrail to shield open canals. The program was used to calculate the minimum required length of canal next to the road before guardrail installation would be expected to return a BCR ≥ 1, based on crash reduction. Figure C1 3: Example of using the US RSAP v3 to model the minimum required length of water body (canal) before a barrier can be installed

Source: Personal communication from Using the Roadside Safety Analysis Program: Part 2, workshop January 2014.

The advantages of using theoretical models include flexibility and transparency of model design. Internal relationships can be added, changed, or bypassed in order to carry out sensitivity testing. Given, that there are numerous professional tools already available, some of them free, the cost of analysis may be low when compared to data-demanding alternatives such as before and after studies, or statistical regression. On the other hand, theoretical modelling requires a substantial investment in scientific expertise, and thus, it has been a domain of academics. Many of the decision support models are more accessible but still require specialist training. Also, the application of models developed in one country may provide only indicative or relative findings in another, e.g. due to differences in vehicle fleet, road rules, or crash reporting.

Design Testing Testing new designs under laboratory conditions has been long established in the areas of pavement and structural design. Other examples of testing that can have input into revision of specific road design guidance include:

•

Crash testing of safety barriers and end treatments under controlled laboratory conditions is an accepted step in seeking adoption of a new product design by road agencies (Standards Australia 1999).

•

Test-track trials are sometimes conducted in order to measure road user response data for innovative designs, as shown by the example in Figure C1 4 (Transport Research Laboratory 2015, Sutton 2013).

Austroads 2015 | page 88

Road Geometry Study for Improved Rural Safety

•

Driving simulators are often used to collect road user reaction and performance data in the presence of new road design, vehicle and other environmental inputs. These laboratory studies use a speciallyconstructed vehicle cabin fitted with screens displaying the road environment and providing dynamic feedback to the test driver (e.g. tilting of the seat, audio and tactile response). Multiple drivers statistically matched to the driving population are invited to take part. These studies typically focus on researching new or unusual design standards before they are trialled in the field (e.g. sign colour or symbols).

Figure C1 4:

Trials of the Dutch-style roundabout in the UK

Source: Sutton (2013).

The main advantage of this approach is the ability to perform analysis of a completely new design (e.g. the trial shown in Figure C1 4), or a very different application of an existing design. While the test methods vary, they are generally robust enough to provide an indication of the key performance indicators. The disadvantages of testing include high cost of such studies, limited relevance in the context of the road network in which the design will eventually perform, and a limited sample size. [Back to body text]

Austroads 2015 | page 89

View publication stats