The Manual for Streets: evidence andresearch

Prepared for Traffic Management Division, Departmentfor Transport

I York, A Bradbury, S Reid, T Ewings and R Paradise

TRL Report TRL661

ii

First Published 2007ISSN 0968-4107ISBN 1-84608-660-4Copyright Transport Research Laboratory, 2007.

This report has been produced by TRL Limited, under/as part ofa contract placed by the Department for Transport. Any viewsexpressed in it are not necessarily those of the Department.

TRL is committed to optimising energy efficiency, reducingwaste and promoting recycling and re-use. In support of theseenvironmental goals, this report has been printed on recycledpaper, comprising 100% post-consumer waste, manufacturedusing a TCF (totally chlorine free) process.

iii

CONTENTS

Page

Executive Summary 1

Acronyms 3

1 Introduction 5

1.1 Manual for Streets 51.2 Design Bulletin 32 61.3 Underlying research 61.4 Report structure 6

2 Review of existing literature 7

3 Site selection and measurement 9

3.1 Site selection 93.2 CAD measurements 93.3 Site surveys 11

4 Speeds and geometry data site ranges 11

4.1 Outliers 124.2 Variation within the data 13

5 Speed adaptation 15

5.1 Link speeds 155.2 Junction speeds 16

6 Modelled safety impacts 18

6.1 Braking modelling 186.2 Stopping distances on links 186.3 Stopping distances at junctions 196.4 Implications of modelled situations 20

7 Observed safety 21

7.1 Belgravia 217.2 Accidents at junctions 227.3 Accidents on links 23

8 Household survey 24

8.1 Sampling 248.2 Sample composition 25

iv

Page

9 Residents opinions 26

9.1 Streetscape 269.2 Parking 27

9.2.1 Car use and off-street parking 279.2.2 Parking problems 289.2.3 Parked vehicles 289.2.4 Respondents issues with parking in their street 29

9.3 Main safety concerns 309.4 Road safety 31

9.4.1 Walking and cycling safety 329.4.2 Safety of children 329.4.3 Improving road safety in residential streets 33

9.5 Accidents 339.6 Non-motorised vs. Motorised users: Access 349.7 Footways 349.8 Summary of household survey findings 34

10 Testing of network layout using SafeNet 35

10.1 Junction spacing 3510.1.1 Analysis 3610.1.2 Overall results 38

10.2 Crossroads analysis 38

11 Conclusions 38

12 Acknowledgements 40

13 References 40

Appendix A: Literature review 43

Appendix B: Case study sites 61

Appendix C: Braking distance matrix 82

Appendix D: Household survey questionnaire 85

Abstract 89

Related publications 89

1Executive Summary

! Conflicting movements at junctions result in a highernumber of accidents, but geometry can lower speedswhich reduce both the likelihood and severity ofaccidents.

! Stopping distances on links and at junctions have amargin of safety down to a visibility of around 20 m inthe environments studied, unless other speed reductionfeatures are incorporated.

! The sites included roads with a range of surface types,varying use of speed restriction measures, differentlevels of on-street parking and a range of forwardvisibilities. The results are consequently applicable to awide range of developments throughout the UK.

! Parking was found to reduce speeds on links and atjunctions by 2 to 5 mph. That is, drivers react to theperceived danger by reducing their speed. The effect ofthis on safety is unclear. Reducing speed increasesrelative safety, but parked vehicles reduce lines of sightand can consequently obscure (crossing) pedestrians.There was no clear indication that this resulted in highernumbers of casualties from the accident statisticsanalysis. However, many of the reported accidents fromthe household survey were related to parked vehicles.

! The largest effect on speeds was found to be associatedwith reducing lines of sight. A reduction from 120 to 20metres reduced approach speeds by approximately 20 mphon links and 11 mph at junctions. Modelling has shownthe reduction in approach speed should result in sightdistances of 40 metres being safe, i.e. there is anacceptable safety margin to stop should a danger presentitself. However, the margin of safety becomes rapidlysmaller below 40 metres.

In addition, a household survey was undertaken toobtain the residents opinions of their streets at the twentycase study sites. This was to determine user satisfactionof a variety of residential street layouts, and to considerresidents transport needs alongside their perceptions ofsafety and sustainability of their streets. Three hundredhousehold questionnaires were returned for analysis toexplore the relationship between resident perceptions ofroad safety and highway geometries.

With respect to the perceptions of residents surveyed,the following can be concluded:! Across the sites there were mixed reactions to whether

personal, or road, safety issues were of most concern.Residents at DB32 compliant sites considered personalsafety (in relation to crime) to be of the greatest concern,but this was not the case at other sites. It is unclearwhether this was owing to higher crime rates at theDB32 sites, the perception of road safety at other sites,or a combination of both these factors. However, overallnearly half the respondents considered road safety to bethe main issue, compared with nearly 30% whoconsidered personal safety to be the highest concern.

Demand for the Manual for Streets emerged from theGovernment research report Better Streets, Better Places(ODPM, 2003), which was commissioned to establishwhether there are any problems over the adoption of newhighways meeting the requirements of Planning PolicyGuidance Note 3, Housing (PPG3). This document focusedon new residential streets and identified highway standardsas a barrier to placemaking in the UK. The reportconcluded with a recommendation for development of aManual for Streets to replace Design Bulletin 32 (DB32)with an updated set of design guidelines for local roads toprovide a catalyst for innovative design that emphasisesplace over movement.

The resulting Manual for Streets is a guide to thedesign, construction, adoption and maintenance of newstreets whose aim is to deliver streets that help strengthencommunities, are pleasant and attractive, are cost-effective to construct and maintain, and are safe. TheManual for Streets has updated geometric guidelines forlow trafficked residential streets, examined the effect ofthe environment on road user behaviour, and drawn onpractice in other countries.

This research undertaken by TRL provides the evidencebase upon which the revised geometric guidelines in theManual for Streets are based, including:! Link widths.! Forward visibility.! Visibility splays.! Junction spacing.

In order to obtain primary data for examining therelationships between geometry, the environment, speed,and casualties, twenty survey sites were selectedthroughout the UK comprising a mixture of new build,Commission for Architecture and the Built Environment(CABE) good practice, DB32 compliant and historic(pre-war) street layouts, to produce a wide range ofdevelopment type to ensure the results were applicable tomany developments within the UK. Methods adopted tocollect data included measurement of X- and Y-distancesat junctions, visibility on links, road width, manual andautomated speed data readings and observations onparking, signing, lining, and traffic calming. 190 linksand 77 junctions were included in the research.

The headline findings from the site surveys can besummarised as follows:! Lower vehicle speeds are associated with reduced road

width and reduced visibility, both on links and atjunctions.

! Site type (for example historic, new build, DB32compliant etc) is not a significant determinant of speed.Junction and link geometries are the important variables.

! Speed is known to be a key factor for road safety. Thefindings of this research are consistent with this fact,indicating that higher speeds on links increase thelikelihood of injury and its severity.

2These results have been integrated into the Manual forStreets in the form of appropriate standards for residentialstreet design, and will become the focus for Governmentguidance on new residential streets.

3Acronyms

ATC Automatic Traffic Count

CABE Commission for Architecture and the Built Environment

CAD Computer Aided Design

DB32 Design Bulletin 32

DCLG Department for Communities and Local Government

DEFRA Department for the Environment, Food and Rural Affairs

DETR Department for the Environment, Transport and the Regions

DfT Department for TransportDPH Dwellings per hectare

HA Hectare

MfS Manual for streetsMPH Miles per hour

ODPM Office of the Deputy Prime Minister

PPG Planning Policy Guidance

SafeNet Software for Accident Frequency Estimation for Networks

4

51 IntroductionThe Manual for Streets (MfS) is intended to consolidate thenecessary components for effective street design into asingle source of information. The MfS builds upon andupdates the guidance contained in Design Bulletin 32(DB32) and its companion guide Places Streets andMovement: A Companion Guide To Design Bulletin 32,Residential Roads and Footpaths. Its aim is to provideguidance for practitioners who will shape the developmentsof the future. It is therefore intended for:! Developers.! Local highway authorities.! Local planning authorities.! The emergency services.! Utility and drainage companies.! Access officers.! Public transport providers.! Architects.! Highway engineers.! Landscape architects. ! Town planners.! Transport planners.! Urban designers.

1.1 Manual for StreetsThe Manual for Streets has been designed to recognise thefull range of design criteria necessary for the delivery ofmulti-functional streets, assisting practitioners in makinginformed decisions relating to appropriate street design. TheManual will initially cover the design considerations forresidential streets and other lightly trafficked local roads.

The Manual deals with underlying values that can becreatively deployed by practitioners in order to pursue theGovernments placemaking agenda of individuallydistinctive localities, while ensuring streets remainfunctional and safe. The Manual for Streets was preparedagainst a backdrop of sustainable development guidanceand initiatives, including the Department for Communitiesand Local Governments Communities Plan SustainableCommunities: Building for the Future (ODPM, 2003b) toensure that it facilitates the long-term sustainability ofstreets, and contributes to an enhanced sense of place.

The Manual for Streets supports the objectives of theGovernments commitment to sustainable development asexpressed in A Better Quality of Life: A Strategy forSustainable Development in the United Kingdom (DETR,1999) and in the latest document on delivering the UKssustainable development strategy Securing the RegionsFutures: Strengthening Delivery of SustainableDevelopment in the English Regions (DEFRA, 2006).This will ensure that residential streets meet the needs ofall street users, not just motorised vehicles.

1.2 Design Bulletin 32The document DB32 was used to assist in designing newhousing developments. It was created to remove the

restrictive criteria imposed in the post-war period thatresulted in a high degree of conformity between estateswithin the UK. Its purpose was to permit a more flexibleapproach to design that enabled developments to be bettertailored to the requirements of residents, for exampleHome Zones in which a variety of techniques (speedreductions and surface treatments) are used to create agreater impression of shared space.

However, a number of requirements are included toensure safety of pedestrians and road users within theestate. These include minimum sight distances in order thatvehicles travelling at a design speed are able to react to adanger and safely stop. The sight distances are specifiedfor an observers eye being between 1.05 and 2 metresabove ground level and in the case of a junction, the carbeing 4.5 metres from the stop line. The required visibilitydistances are summarised in Table 1.1, and Figure 1.1.

However, according to Places Streets and Movement:A Companion Guide To Design Bulletin 32, ResidentialRoads and Footpaths these should be extended to 45 min a 20 mph zone and 90 m in a 30 mph zone to allowfor drivers exceeding the speed limit by up to 10 kph(Table 1.2). Also the Y-distance should be measured forvehicles at the following distances from the stop line onthe minor arm of the junction:! 9.0 m: The normal requirement for major new junctions

and for the improvement of existing junctionsbetween access roads and district or local distributorroads - for instances where the minor road is busy.

! 4.5 m: For less busy minor roads and busy privateaccess points.

! 2.4 m: The minimum necessary for junctions withindevelopment to enable a driver who has stoppedat a junction to see down the major road withoutencroaching onto it.

! 2.0 m: For single dwellings or small groups of up to halfa dozen dwellings or thereabouts.

Table 1.1 DB32 visibility (Y) distances for differentdesign speeds

Speed ( mph) 5 10 15 20 25 30Speed (kph) 8 16 24 32 40 48Distance (metres) 6 14 23 33 45 60

0

10

20

30

40

50

60

70

0 5 10 15 20 25 30 35

Speed (mph)

Y D

ista

nce

(metr

es)

Figure 1.1 DB32 visibility (Y) distances for differentdesign speeds

6Consequently, consider a residential area with a 30 mphlimit. A driver on a minor road approaching a junctionshould be able to see vehicles at a distance of 60 to 90metres from the junction on the major road depending onwhether the drivers on the major road remain within thespeed limit. Further, they should have this field of view fora distance of 4.5 metres before the junction if showingcaution, or 2.4 metres if they are stopping at the junction.

These calculations assume a design speed and thestandard stopping model of a driver when presented with adanger: i.e. permitting a reaction time and then assumingthe driver will apply a constant braking force. However,the situation can be considered from the opposite direction.

If sight lines are reduced below the recommendation inthese guidelines, do drivers react to the lack of visibility?Suppose drivers reduce their speed when encounteringreduced lines of sight and therefore increased risk. Thiscould, in effect, result in the housing development being assafe as one with greater lines of sight and promote lowerspeeds. That is, the reduction in speed could still permitthem to stop the same distance before a hazard eventhough they see it when it is closer to them.

1.3 Underlying researchTRL has performed research into identifying andinvestigating design elements whose impact was not fullyunderstood, and in particular those not previously based onrigorous research. The study initially performed a literaturereview of local authority design guides: 32 were included.These indicated designs were constrained by the followingcritical aspects of highway geometry:! Link widths.! Forward visibility.! Visibility splays.! Junction spacing.

Little robust research supporting the DB32 standardswas found in the review. Furthermore, the identifiedresearch did not explore detailed design elements, such asgeometric dimensions.

The review revealed the majority of the localauthorities complied with, and recommended, the samestandards as DB32. Where standards differed, they weregenerally more stringent than those in DB32. Asexpected, road safety was the most significant barrier tothe adoption of standards with relaxed values of widthand visibility.

This research therefore aimed to assist in setting thedesign standards for MfS and to inform its developmentwith respect to road widths, visual splays, parking, andremoval of road markings. In particular, it aimed to:

! Examins the relationship between driver behaviour andhighway geometry.

! Establish the safety of roads which do not meet DB32standards in terms of casualty numbers, driver behaviourand resident perceptions.

! Investigate highway layouts to determine whether morepermeable layouts are associated with higher levels ofcasualties than spine and cul-de-sac layouts.

Three strands of research were included:1 The first consisted of observations, conducted at twenty

residential developments, to examine links and junctionsthat were either at, or below, the limits specified in DB32.

Observations included detailed information on thegeometry and layout of each link and junction,observations of speed and obtaining accident statistics.Statistical analysis and predictive models based uponthese observations indicated whether relaxed geometricand visibility values could be incorporated into the MfS.The features also varied in relation to:! Road width.! Whether speed humps were present.! Whether parking was permitted.! Whether lines were present to indicate priorities e.g.

give way line.! The type of road surfacing used.

2 A household postal survey was conducted at the sametwenty sites. This explored residents opinions on anumber of aspects of their living environment. However,the emphasis of the questionnaire was to considerwhether they had any perceived safety or personalsecurity concerns. A comparison across the sites couldtherefore investigate if junction and road geometries thatdo not meet DB32 standards are perceived by residentsto be as safe as those that conformed to the standards.

3 Lastly, SafeNet, which can model a road network andconsider the effect of changes on safety, was used tostudy the effect of junction spacing on casualty rates,and the effect of changing the characteristics of aresidential area from a cul-de-sac approach to a moreconnected layout.

Within this report the following terminology is used:! A site is a housing development that has been surveyed

in this study.! A feature is defined as being either a link or junction

within the site.

1.4 Report structureSection 2 discusses the research methodology used, includinga rationale for selecting the research study sites and thevariety of methods used to collect and analyse the field data.

Section 3 provides a literature review and gap analysis thatsupports the development of the primary researchmethodology.

Table 1.2 Companion guide Y-distances

Speed (mph) 20 30 40 50 60 70Speed (kph) 32 48 64 80 97 113Distance (metres) 45 90 120 160 215 295

allowing speedingDistance (metres) 33 60 120 160 215 295

7Section 4 discusses the site surveys conducted, including adescription of the different sites and the results andanalysis of the surveys, particularly with regard tovisibility on links and at junctions in relation to trafficspeed and accident risk.

Section 5 describes the models used to consider whetherany observed speed reductions are sufficient for a link orjunction with limited visibility to be safe.Section 6 considers the predicted effect of limited visibilityon speeds, and models a number of situations to ascertainwhether the speed reductions compensate for the lack ofvisibility.

Section 7 analyses road accidents statistics (STATS19data) for the research sites and the relationship betweenspeed and geometries on accidents.1

Sections 8 and 9 give an overview of results from thehousehold survey intended to assess residents opinions ofthe streets where they live and to identify particular issuesthat residents have about the design and layout of theirstreet and built environment.

Section 10 presents the results and analysis of the junctionspacing research using the Software for AccidentFrequency Estimation for Networks (SafeNet) to comparehypothetical networks based on DB32 compliant andorganic street layouts.

Section 11 concludes with an overall summary of theresearch and a discussion about the proposed standardsthat will be included in the Manual for Streets.

The report concludes with a summary of results from theresearch, and how this affects the standards for residentialstreet design, to be revised for the MfS.

2 Review of existing literatureA literature review (see Appendix A) was performed as aninitial element of this research to collect togetherinformation relevant to the Manual for Streets. Its purposewas to perform a gap analysis. That is, to ascertain whereresearch could underpin the DB32 standards so they couldbe directly included in the Manual for Streets, and toidentify elements within the current guidance based uponlimited research (either needing validating or furtherresearch). Reports and articles on the following subjectswere sought:! Policy, legal and technical frameworks.! Objectives for streets.! Sustainable communities.! Quality places.! Movement.! Access.! Parking.

! Materials, street furniture and planting.! Lighting.! Services and drainage works.! Maintenance and management.

The literature evidence relating to each element is thenlisted, so that it may be cross-referenced to Appendix A.This gap analysis is shown in Table 2.1.

The review found few references concerned with theeffect of geometric dimensions within the contexts required.However, references were located on the theoretical aspectsof creating liveable streets. The majority of elements wereassessed as partially covered by current knowledge butneeding validation before being used.

One element covered within a number of researchreports was road or carriageway width. However, furtherresearch was necessary to determine suitable carriagewaywidths within residential areas: i.e. those promoting lowtraffic speeds, create a safe environment and ensureadequate access.

Traffic calming measures also have adequate coveragewithin the literature, again with the objective of slowingspeeds and creating safer places. Whilst the majority of therobust evidence is concerned with physical traffic calmingmeasures (humps, pinch points etc), there is recentresearch examining psychological traffic calmingmeasures. These measures use the surroundings toinfluence driver behaviour (e.g. width of road, colouredsurfaces, location and height of buildings close to thecarriageway edge). However, this research was primarilyconducted in rural areas and hence its applicability toresidential areas requires further research.

There are a number of publications on Home Zones, orthe Dutch Woonerven. Research studies have also beenundertaken focusing on traffic volumes, accident levelsbefore and after implementation and the views of residents.However, there are some research gaps within the area ofHome Zones, such as the inclusion of disabled people(research has recently been commissioned on this topic)and robust research focusing on the safety implications ofHome Zones (as only limited after data was analysed)and associated social impacts of schemes.

Parking research was fairly extensive, but not consideredrobust. Generally, research suggests parking should beincorporated within the design of residential streets as it canact as a traffic calming measure. However, its inclusion inresidential streets should not create danger for playingchildren or crossing residents, especially parking atjunctions, which may obstruct the vision of drivers.

Overall the research review highlighted a lack of robustresearch supporting DB32 standards and the information tobe incorporated in the Manual for Streets. Where researchhas been undertaken, detailed design elements, such asgeometric dimensions, have been neglected. Consequently,the following design aspects were investigated further:! Road widths which widths result in low speeds whilst

maintaining safety, access and ease of traffic flow?! Visual splays which sight lines result in preferred

driver behaviours, whilst maintaining safety?1 In this report, accidents are used to denote personal injury incidents.

8Table 2.1 Gap analysis of research literature

Supported

Chapter content specifications Fully Partially Not Evidence source (see Appendix A for full details)

Street networks and types

Hierarchies of traffic and place functions "

Travel demand by mode "

Grids vs cul-de-sac "

Block dimensions find / course grain "

Achieving appropriate speeds throughSee Achieving appropriate speeds below

network / environmental effects

Public transport, walk and cycle networks "

Integration / segregation of cars / cycles / people "

Guard railing " English partnerships and LlewellynDavis (2002)

Mixed use streets "

Shared space "

Shared surfaces " Polus and Craus (1996)Home Zones " Barrel and Whitehouse (2004); Tilly et al.

(2005); Layfield et al. (2005);Webster et al. (2005)

Rural lanes / Quiet lanes " DfT (2004); Kennedy et al. (2004a and b)

Street dimensions

Design vehicles dimensions, dynamic envelopes "

Widths carriageways, cycleways, footways shared areas " Burrow (1977) Daisa & Peers (1997);Gibbard et al. (2004); Oxley (2002);Lawton et al. (2003)

Capacity for vehicle movement "

Street in cross-section kerb height / crossfalls "

Absence of centre-line markings " Countryside Agency (2005)

Where streets meet

Place importance of junctions / squares "Visibility splays "

Unmarked junctions "Junction spacing "

X junctions "T junctions "Roundabouts " Lawton et al. (2003)Signals "

Informal squares "

Footway crossings "

Pedestrian crossings- signal / zebra / refuge / courtesy "

Continued ....

9! Parking How can parking be best incorporated intoresidential design? Can it be successfully used as atraffic calming measure? What are the impacts onsafety? Can the quality of the area be retained?

! Removal of road markings What effect does theremoval of road markings have on driver behaviour?

3 Site selection and measurement

3.1 Site selectionTwenty survey sites were selected throughout the UK; tenof the sites were case study and new build areasselected by CABE. The remaining 10 sites were a mix ofhistoric (pre-War), DB32 compliant and new build sitesselected to produce a wide range of development type to

ensure the results were applicable to many developmentswithin the UK. The full list of sites is shown in Table 3.1,and their distribution within the UK is shown in Figure 3.1(see Appendix B for a summary of each study site).

The sites were initially evaluated from CAD drawingsand then assessed in detail during a site visit.

3.2 CAD measurementsDetailed site characteristics were measured from site plans:technical drawings of the selected sites depicting allstructures in the area. Using AutoCAD it was possible totake accurate measurements of sight lines, as the drawingswere detailed, accurate and all obscuring features wererecorded (see Figure 3.2).

The X-distance was set at 2.4, 4.5 and 9 m from thejunction measured down the centre line of the road, in

Table 2.1 (Continued) Gap analysis of research literature

Supported

Chapter content specifications Fully Partially Not Evidence source (see Appendix A for full details)

Street alignments

Gradients "

Curve radii, horizontal and vertical "

Forward visibility " Kennedy et al. (1998); Layfield et al.(1996); Summersgill and Layfield(1996); Taylor et al. (1996)

Achieving appropriate speeds

Ideally through network (and natural traffic calming) " Scottish Executive (1999); Grayling et al.(2002); Vis et al. (1990) Kennedy et al.(2005); Hardy (2004); Elliott et al. (2003)

Traffic calming as fallback " Engel and Thomsen (1992)Integration of TC with environment "

Access

Access requirements to buildings people and vehicles "

Emergency access. References to building regulations "

Servicing: refuse collection, deliveries, removals "

DDA/disabled requirements into buildings, along streets " Oxley (2002)Need to balance the perceived conflict between accessibility " ODPM (2004)and crime permeability versus security

Parking

Layouts / design of on/off street parking " Noble and Jenks (1996); TRL (1992);Noble et al. (1987);Westdijk (2001);Scottish Executive (2005)

Relationship with capacity and safety "

Motorcycle / cycle parking "

10

Table 3.1 Research study sites

Rural / Housing Land NetworkCharacteristic Town Ward Region urban period use Density type

Historic (pre-war) Reading New Town South East Urban Victorian Mixed High GridLavenham Suffolk South East Rural Medieval Residential Low OrganicOxford Jericho South East Urban Victorian Residential High GridBloxham Village Oxfordshire South East Rural Victorian Residential Low OrganicChichester West Sussex South East Urban Medieval Mixed High OrganicLondon Belgravia South East Urban Victorian Mixed High Grid

Case study Charlton Down West Dorset South West Rural Post 90s Residential High OrganicLichfield Darwin Park West Midlands Urban Post 90s Residential High OrganicEastleigh Former Pirelli site South East Urban Post 90s Residential High Atypical gridNewhall East Harlow East of England Suburban Post 90z Residential High OrganicGuildford Queens Park South East Urban Post 90s Residential Mid OrganicLondon Tower Hamlets South East Urban Post 90s Residential High GridGlasgow Crown St. Scotland Urban Post 90s Residential High OrganicChelmsford Windley Tye East of England Suburban Post 90s Residential Low Court layoutChelmsford Beaulieu Park East of England Urban Post 90s Residential Low GridManchester Hulme North West Urban 1990s Residential Low Grid

New build Ipswich Rapier St. South East Suburban Post 90s Residential High Atypical gridPortishead Port Marine South West Suburban Post 90s Residential Mid Organic

DB32 Compliant Leicester Syston East Midlands Urban 1980> Residential Mid Cul-de-sac with spineReading Lower Earley South East Urban 1980> Residential Mid Cul-de-sac with spine

Figure 3.1 Distribution of sites

1: New Town, Reading2: Lavenham, Suffolk3: Jericho, Oxford4: Bloxham Village, Oxfordshire5: Chichester, West Sussex6: Belgravia, London7: Charlton Down, West Dorset8: Darwin Park, Lichfield9: Former Pirelli Site, Eastleigh

10: Newhall, East Harlow11: Queens Park, Guildford12: Tower Hamlets, London13: Gorbals, Glasgow14: Windley Tye, Chelmsford15: Beaulieu Park, Chelmsford16: Hulme, Manchester17: Rapier Street, Ipswich18: Portmarine, Portishead19: Syston, Leicester20: Lower Earley, Reading

11

accordance with generally accepted practice. From thesethree distances the line of sight (Y-distance) is measuredleft and right looking out of the junction. The Y-distance isthe furthest point of visibility on the kerb (the blue line),taking into account any obscuring features, and ismeasured from the centre line of the side road.

Lines of sight on links were measured using a similarmethod. The drivers (direct) sight line was used to findthe shortest distance ahead of the vehicle that is visiblewhen on the link. In addition, the width of the road, bothwith and without footways, was also measured from theCAD drawings.

This information gave a first indication of the range ofgeometries available across the chosen sites. However, thiscould not take account of all complications on site. Thereforethis information was validated and enhanced by site surveys.

3.3 Site surveysAll site surveys were conducted from January to March2006. Each survey comprised measurements of speed, roadgeometry and visibility, at a minimum of 10 junctions and2 links within the site, and speed readings at a minimum of5 junctions and 2 links. The purpose of the surveys was tovalidate the measurements taken from the CAD drawingsand obtain a sample of speeds from a wide range ofjunctions and links throughout the sites. The followingprocedures were undertaken at each junction and link:

Junction visibility (Y Distance)1 The centre line of the main road was ascertained.2 The three X values were measured and marked off using

chalk on the road.3 Y-distances were measured at a height of 1.2 m from

each X chalk mark to the left and right.4 Where parking was explicitly marked on the road, a

second measurement, assuming a parked vehicle waspresent, was also recorded. Lines of sight also took intoaccount cars parked on the road, as these were assumed tobe the general conditions that drivers would encounter.

5 Any visible obstructions were noted down, theseincluded buildings, brow of a hill, phone boxes, hedges,parked cars etc.

Link forward visibility1 A pre-defined place in the road calculated to have

minimum visibility from the CAD drawings was located.2 All obstructions, including parking, were recorded.3 Taking the obstructions into account the pre-defined

point of minimum visibility was confirmed as correct, oradjusted and recorded.

4 The distance between the correct position and thefurthest point of visibility was measured: along the roadif considered safe, or along the kerb, and the road widthwas recorded.

Manual speed measurements (speed gun)! Recorded vehicle speeds for 20 minutes at a sub-sample

of links and junctions.

Sample sizesEstimates from CAD drawings and measurements fromsite surveys resulted in geometric information beingcollected for:! 190 junctions, and! 77 links.

A speed gun was used to record individual vehicleapproach speeds for approximately twenty minutes at asub-sample of the surveyed features. Exact speeds (to thenearest mph) were recorded above 10 mph, however,manually collected speeds at, or below, 10 mph wereclassified into one category. These observations werecollated into an analysable form for! 57 junctions, and! 23 links.

Also, automatic traffic counters (ATCs) were installedfor one week on:! 18 links.! The approach to 10 junctions.

The speed gun survey provided a snapshot of theapproach speeds of drivers across a wide range of differentjunctions and links, and therefore a wide range ofgeometries. In contrast, the ATC data was restricted to asmaller number of features, and therefore geometries, butgave a robust picture of the approach speeds used.

The purpose of the next section is to examine how thespeeds varied across the features studied and identify thecharacteristics that influenced those speeds.

4 Speeds and geometry data site rangesA general understanding of the effect of road layout, andthe resulting visibilities, on drivers speeds can only be

x

y

Obscuring Feature

Centre Lines

Figure 3.2 Y-distance at junctions

12

200150100500Y distance measured at 4.5 metres from junction

(metres)

60

40

20

0

Frequ

ency

Mean = 30.97 Std. Dev. = 27.717

N =194

Figure 4.3 Range of observed junction visibilities

achieved if a wide range of different site conditions arestudied. The selected 20 sites were consequently chosen tobe representative of developments throughout the country.In terms of the type of sites studied there were:! 6 historic (pre-war) sites.! 6 sites conforming to CABE good practice guidelines.! 2 sites conforming to DB32 specifications.! 6 other (case study) sites.

Road surfacing within the sites varied from tarmac(12 sites), through to a mixture of tarmac and block paving(3 sites) to wide-scale use of block paving (5 sites). Halfthe sites had no speed restricting measures present (e.g.road humps or horizontal deflections), whilst some form ofspeed restricting measures were present in the others. Theextent of on-street parking also varied across the sites with7 sites having no parking near to junctions in evidenceduring the site visits, whilst some vehicles parked close tothe junctions at other sites. Also, parking on the linksvaried between no observed parked vehicles, to parking onone side of the road on some links and parking on bothsides of the road on others.

Forward visibilities on links, and visibility at junctions(Y-distance), were estimated from plans of thedevelopments and measured during the site visits. Often,the observed visibilities were less than those estimatedbefore the site visit (using AutoCAD). This was owing tothe presence of street furniture and other obstructions (forexample planting) that limited visibility. This study isconsidering the effect of actual visibility on driversspeeds, therefore the visibilities measured on site havebeen used in assessing driver adaptive behaviour. Theobserved ranges of road width, forward visibility on linksand visibility at junctions are shown in Figures 4.1 to 4.3.

Road widths (excluding footways) generally varied from4 to 10 metres across the studied sites. Visibilities on links,and at junctions, varied from below 10 metres toapproximately 100 metres. The sites can therefore beconsidered to be representative of the majority of situationsoccurring within residential developments in the UK. 4.1 Outliers

Sites with abnormal characteristics can distort the findingsof any statistical analysis. Such sites can produceobservations that significantly differ from the trendspresent within, and across, other typical sites: suchobservations are referred to as outliers. These observationscan therefore affect the statistical models fitted to the data,referred to as points of high influence.

The accepted approach in these situations is to excludethese abnormalities from the analysis and produce modelsthat reflect the trends within the majority of situations. Theprevious section shows that nearly all the sites have roadwidths less than 10 metres (excluding footways) and linesof sight of less than 100 metres on links and at junctions.In addition, the average daily flows and average linkspeeds at each of the sites studied using ATCs aresummarised in Table 4.1.

Observations from Belgravia were anomalous. The roadwidth where the ATC was situated was 11.4 metres wide(excluding footways), and the average for all linkssurveyed in Belgravia was 10.5 metres. In addition, the

12.010.08.06.04.02.0Road width without footways (metres)

40

30

20

10

0

Fre

quen

cy

Mean =6.285 Std. Dev. =1.641

N =145

Figure 4.1 Range of observed road widths

450400350300250200150100500

Forward visibility (metres)

60

50

40

30

20

10

0

Frequ

ency

Mean = 57.187 Std. Dev. = 69.7081

N = 99

Figure 4.2 Range of observed link forward visibilities

13

forward visibility on the link where an ATC was installedwas 446 metres and generally there was good visibility onthe links throughout the Belgravia site. These conditionsresulted in the highest average speed in any site. Inaddition, the daily vehicle flows were almost 30% higherthan on any other site. Thus the Belgravia observations aregenerally excluded from the analysis performed. As aconsequence, the results can be considered to be relevantfor developments such as shown in Figure 4.4 with roadwidths up to 10 metres and visibilities on links (and atjunctions) of up to 100 metres, and not for those with gridlayouts such as Belgravia (see Figure 4.5) with wide roadsand larger visibilities.

The site with the next highest average link speed wasLeicester where the forward visibility was 65 metres, andthe lowest average link speeds were at Beaulieu Park inChelmsford, where the forward visibility was 10 metres.The following sections consider the effect of forwardvisibility on links (and visibility at junctions) and otherrelevant influencing factors on speed.

4.2 Variation within the dataRanges in geometry, parking, signing and speed reductionmethods present within the 20 studied sites have beenexplored. The sites (excluding Belgravia) provide a goodvariation in all these site dependent factors that mayinfluence speeds. However, it is important to be able toensure that these are the only factors affecting differencesobserved between the sites and features studied, in orderthat the results are not confounded by other differences.Other possible influences at a given feature include:! Time of day, for example night driving.! Day of week, weekend driving compared with weekday

driving.! Weather conditions.! Age of driver.! Driving style.! Other vehicles present.

It is not possible to take account of all these factors withinthis type of analysis. However, the best consistency wassought between the manual speeds and loop speeds. Allmanual speeds were collected on weekdays in the daytime.Consequently, the speed loop analysis was also restricted toobservations on a weekday between 0700 and 1900.

Table 4.1 Average flows and speeds

Site Average daily flow Average speed (mph)

Lower Earley 70.4 11.3Guildford 481.9 18.2New Town, Reading 242.6 14.4Chichester 1372.8 19.4Eastleigh 427.7 17.3Belgravia 2029.5 25.7Tower Hamlets 627.0 19.1Ipswich 121.6 19.0Lavenham 221.4 11.7Newhall 482.9 15.6Windley Tye 294.9 16.9Beaulieu 83.8 10.2Bloxham 112.3 12.5Portishead 1161.1 15.8Leicester 528.9 23.5Manchester 1060.6 19.3Lichfield 362.6 16.7Glasgow 1575.4 10.3

A 461

27

19

127

86.0m

45

119a

89.3m

51

The

24

33

LEAMONSLEY

WALSALL

ROAD

VICTORIA

12

14

2

32

8

117

129

Vicarage

127a

15

23

121

40

44

40

38

House

36

32

26

4

11

2119

2

57

23

7

9

Court

Leomansley Court

37

53

27

22

15

CHATTER

TON AVE

7

189

Easter Hill

97

145

104

178

98

139

147

174

179

99

101

SAXO

N WA

LK

LEOM

ANSL

EY VI

EW

Rookery

10

159

The Old

19

Vicarage

LANE

168a

25

105

9

5

168

2

BARDELL

CLOSE

4187

El Sub Sta

192

25

21

52

16

2

ORMOND

S CLOSE

6

1

3

1

6

4

2

8

5

ALESMOR

E MEADOW

11

59

17

194

CLOSE

30

1

5

12

3

10

7

158

34

11

1

35

42

WALN

UT W

ALK

34

38

32

25

29

21

Headland

Crown Copyright 2005. All rights reserved. Licence number 100020449

92

1

146

138

5

106

CHRIS

TCHU

RCH

103

92.0m

165

14

POOLFIELD

ROAD

4

2

177

14

12

10

170

WALSALL R

OAD

191

BM 93.77

m

= ATC Location

= Junctions Measured

= ATC I.D. = Links Measured

= Manual Speed Reading Locations

Lichfield

1

23

5

6

7

8

910

11

1

2

3

4

5

25

26

1

23

5

6

7

8

910

11

1

2

3

4

5

25

26This organic layout is characterised by low link visibilities (see link 2 image). Jinction sight lines are also shorter compared to the grid layout of Belgravia.

23

22

25

213

0

3 4

3 2

Figure 4.4 Lichfield non-grid layout (limited visibilities)

14

ECCLESTON

STREET

100

Bolivian

106

107

108

109

Embassy

8

6.9m

LB

7.8m

TCB

BM

7.87m

Eaton

Squa

re Ga

rdens

6

24

SQUA

RE

26

7

36

11

32

103

102

EATO

N

95

98

Belgian Embassy

103

= ATC Location

= Junctions Measured

= ATC I.D.

= Links Measured

= Manual Speed Reading Locations

Belgravia - London

The grid layout of Belgravia gives rise to high visibility levels at the majority of junctions, as can be seen in the image of junction 4. There were also clear views down the lengths of all measured links (link 1 = 122m, link 2 = 89m, link 3 = 194m, link 4 = 183m). This can be seen in the image of link 4.

467

5

LB

8

1 to 4

LB

14

12a

12

7.1m

EATON SQUARE

EATO

N SQU

ARE

18

EATO

N SQU

ARE

Eaton

Squa

re Ga

rdens

EATO

N SQ

UARE

117

118 7.5m

12

UPPER BELGRAVE

STREET

KNIG

HTSBR

IDGE

AND

BELGR

AVIA W

ARD

13

15

14

2

2

118

11

KNIG

HTS

BRID

GE

AND

BELG

RAV

IA W

ARD

115

6

8

10

22

12

111

109

7.9m

22

21

23

LB

6.9m

Tennis

Court

98.0m

LB

4

11

1

BM

8.12m

17

31

5

36

24

EATO

N SQ

UARE

ECCLESTON

STREET

6.9m

100

Bolivian

106

107

108

109

31

33

34

Embassy

8

Eaton

Squa

re

Garde

ns

6.9m

LB

7.8m

TCB

BM

7.87m

Eaton

Squa

re

Garde

ns

LB

34

6

93

BELGRAVE PLACE

ECCL

ESTO

N ME

WS

24

SQUA

RE

26

7

36

11

32

107

103

44 to 47

40

CITIES OF LONDON AND WESTMINSTER BORO CONST

38

39

41 to 43

EATO

N SQ

UARE

90

46

102

EATO

N

95

98

32

19

Nuffield Club

Belgian Embassy

103

LONDON EER

19

43

37

35b

35

ItalianEmbassy

Emba

ssy o

f the

Hung

arian

Peop

le's Re

public

16

EATO

N MEW

S NOR

TH

54

56

14

15

86

38

83

LYALL STREET

84

BM

7.00m

LB

69

BM

7.75m

EATO

N PLA

CE

68

8

57

45

CITY OF WESTMINSTER LB

WEST CENTRAL GL ASLY CONST

Club

82

The Irish

82

7

The Irish

Club

EATO

N SQU

ARE

77

80

44

LYALL STREET

8

82

9

2

84

64

46

EATO

N MEW

S NOR

TH

48

47

62

6

48

EATO

N SQU

ARE

51

EATO

N SQU

ARE

57 to 60

53 to

56

6.9m

Eaton

Squa

re Ga

rdens

The F

ive Fi

elds

61

67

72

6189

91

55

EATO

N ME

WS NO

RTH

73

53

108

57

58

103

93

EATON PLACE

101

102

104

99

7.3m

71

94

86

90

87

79

7.4m

62

2

LB

EATON SQUARE

66

66a

63

Crown Copyright 2006. All rights reserved. Licence number 100020449

LBPH

35

12

7

2

34

1

123

4

5

6

15

9

10

11

823

Figure 4.5 Belgravia grid layout (large visibilities)

It was also possible that differences could be influencedby the type of site. That is speeds could depend on theintrinsic design present within a historical site comparedwith one conforming to the DB32 standards: similar to theeffect of the grid layout in Belgravia. If such variations werepresent the different type of sites would result in distinctdata clusters. The resulting ATC link speeds according tothe measured visibility are shown in Figure 4.6.

Observed speeds, and average speeds, on links withinthe sites increased with forward visibility. The type of siteappeared to have no effect. Therefore, all sites and featureswithin them were considered together and the effect of

geometries, speed restriction measures and surface typesexplored across all non-anomalous data. Figure 4.7 showsthe same link speeds according to forward visibility andthe sites on which they were measured.

80.070.060.050.040.030.020.010.0

Measured forward visibility

70

60

50

40

30

20

10

Obs

erve

d sp

eed

(km/h)

Site 85th percentileSite meanOtherHistoricDB32 compliant

Type of site

Figure 4.6 Speed ATC data for links (by site type)

80.070.060.050.040.030.020.010.0

Measured forward visibility

70

60

50

40

30

20

10

Obs

erve

d sp

eed

(km/h)

LichfieldManchesterLeicesterPortisheadBloxhamBeaulieuWindley TyeLavenhamIpswichTower HamletsEastleighChichesterNew Town, ReadingGuildfordLower Earley

Site Number

Figure 4.7 Speed ATC data for links

Even restricting observations to weekdays during thedaytime, it is clear there was a large variation in speeds ateach site and on any given link. The variation will havebeen affected by factors that cannot be accounted forwithin this study, for example, variations in driving styleand individual circumstances occurring on the link.

15

Interestingly, some drivers were observed at approximately10 mph at nearly all the sites though this could be owing toparticular circumstances encountered by the drivers, forexample, slowing to permit a vehicle through from theother direction.

Increasing variation with increases in an independentvariable can be accounted for by transforming the dependentvariable. However, even with this approach the size of theinherent variation at any one site and feature will ensure thatthe percentage of variation accounted for by a model will below. The statistical models therefore investigatedexplanatory variables whose coefficient was significant, andhence captured a significant trend within the data, withoutplacing any criteria on the overall models fit, therefore theregression (R2) value could be low.

5 Speed adaptationIt is probable that drivers adapt their speed according tothe perceived danger on the road. Initial investigation inthe previous sections supports this as the observed averagespeed on a link decreased with forward visibility. The aimof the analysis was therefore to find significant trends inspeeds explainable by the differences between thejunction, and link, layouts.

Speed is an indicator of safety on a road. If the averagespeed is lower, then arguably the road is safer and lessintimidating to vulnerable road users. There are two effectsthat make the road safer. Firstly, the stopping distance ofvehicles is less and therefore the probability that a driverwill be able to stop and avoid an accident is higher underthe same conditions. Secondly, should an accident occur,its severity could be less owing to reduced impact speed.

Reductions occurring through non-geometric treatments(including the type of road surface, speed humps andsigning) would be expected to increase safety comparedwith a similar road without the same measures. However,the situation is more complicated with respect to geometricdifferences where lines of sight are affected. On the onehand drivers may reduce their speeds owing to theperceived danger, but there is a potential increase indanger through drivers being unable to see hazards untilthey are closer to them. These two effects counteract each

other. So, this research considers whether any observedspeed reductions (through perceived danger) are sufficientfor a link, or junction, with limited visibility to be as(actually) safe as one with good visibility.

5.1 Link speedsDrivers within a development generally negotiate anumber of links and junctions during their journey. Theywould be expected to have a highly variable drive cycle asthey can be stationary at junctions, and possibly on links,to give way to other traffic. They would be expected toadapt their speed to perceived dangers such as parking andhorizontal deflections that reduce their forward visibility.However, drivers will typically try to maintain themaximum speed to minimise their journey time.

Higher speeds, with greater variation, would be expectedon links as vehicles would not be expected to show thesame caution as at junctions. Therefore lines of sight areimportant on links because if lines of sight are reduced andspeeds remain high, an unexpected occurrence such as apedestrian stepping into the road, is more likely to result ina serious accident. For comparison the observed speeds onthe approach to a junction and on a link in Tower Hamletsare shown in Figure 5.1.

Two (multi-linear) regression models were fitted to theavailable data. One to the manual speed data from 23 linksand the other was fitted to the ATC data from 18 links. Theforms of the regressions were:

In( ) ( ) ( )speed a b road width c forward visibility= + + +( ) ( )d parking e surface type+

The natural logarithm of speed (km/h) was found toproduce the best fit to the data. The results of the models areshown in Figures 5.2 and 5.3 respectively, and in Table 5.1.

The regressions explained only between 20 and 22% ofthe total variation in the data. However, all the variableswere significantly different from zero at the 95%confidence level.

The models imply that either permitting parking on alink, or the use of block paving, can reduce link speeds by2 to 5 mph which could improve safety, though there areclearly issues with pedestrians being obscured by parked

70605040302010

Observed speed (km/h)

400

300

200

100

0

Frequ

ency

Mean = 26.0118 Std. Dev. = 7.69201

N = 4,400

70605040302010

Observed speed (km/h)

400

300

200

100

0

Frequ

ency

Mean = 30.8089 Std. Dev. = 8.47612

N = 4,389

Figure 5.1a Junction speeds Figure 5.1b Link speeds

16

vehicles. Similarly, reducing road width also has the effect ofreducing vehicle speeds on the link. However, the two modelsdisagree about the extent of the reduction. The manual speeddata implies that a reduction from 10 to 5 metres could reducethe link speeds by between 9 to 20 mph, depending on theforward visibility on the link. In contrast, the ATC dataimplies the reduction is between 2 and 4 mph. It is theconservative estimate from the ATC data that is most likely tobe accurate given that the model included nearly as manysites and also given that the average number of manualobservations was 32 within any one site.

Both models agree that drivers do adapt their speedaccording to the forward visibility on the link. Accordingto the model based on the ATC data an average driver

Table 5.1 Predicted changes across roads with 5 metrewidths and forward visibilities between 5 and110 metres

Decrease ( mph)

No parking to parkingon both sides of link Tarmac to block paving

Link model Minimum Maximum Minimum Maximum

Manual speeds with 2.8 4.8 2.5 4.4 average site featuresLoop speeds 1.5 3.5 1.3 3.2

reduces their speeds by approximately 20 mph if the forwardvisibility on the link is reduced from 110 to 20 metres.

5.2 Junction speedsApproach speeds at junctions were generally found to beless than link speeds, as drivers slowed and showed cautionon the approach in case they needed to give way to anotherdriver with priority. The requirement to assess a number offactors during their approach seems to have both heightenedawareness and also increased workload on the driver.Overall there appeared to be more accidents on the sites atjunctions than on links: 110 accidents at 187 junctionscompared with 21 accidents on 74 links. Hence reducingspeeds and accidents at junctions is clearly important.

Information was available on the visibility (Y-distance)at each junction for a vehicle positioned at 2.4, 4.5 and9 metres before the junction. Initial tests considered whichvisibility was the best predictor for the observed approachspeeds. This model indicated that a visibility of 4.5 metresshould be used within the regression modelling.

Regression models were formed on both the manuallycollected speed data and on the ATC data. In addition,whilst the ATC data recorded the speeds of all approachingvehicles, the manual data also classified the vehicles as towhether they turned at the junction. Therefore a separatemodel was also formed for all vehicles that did not turn atthe junction for this data set. A summary of the models isshown in Figures 5.4 to 5.6 and Table 5.2.

0

5

10

15

20

25

30

35

40

45

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7 Road Width = 8mRoad Width = 9mRoad Width = 10m

0

10

20

30

40

60

50

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7 Road Width = 8mRoad Width = 9mRoad Width = 10m

Figure 5.2 Link model manual speeds with average site features

0

5

10

15

20

25

30

35

40

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

0

5

10

15

20

25

30

35

40

45

50

0 20 40 60 80 100 120Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

Figure 5.3 Link model ATC speeds

17

0

5

10

15

20

25

30

0 20 40 60 80 100 120Y Distance (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

0

5

10

15

20

25

30

35

40

45

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

Figure 5.5 Junction model manual speeds with average site features (non-turners only)

0

5

10

15

20

25

30

0 20 40 60 80 100 120

Y Distance (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

0

5

10

15

20

25

30

35

40

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

Figure 5.4 Junction model manual speeds with average site features (all vehicles)

Table 5.2 Predicted changes across roads with 5 metre widths and Y-distances between 5 and 110 metres

Decrease (mph)

No parking No junctionto parking Tarmac to markings to

near junction block paving junction markings

Junction model Min Max Min Max Min Max

Manual speeds with average site features 1.5 2.1 3.7 5.2 -1.1 -0.8Manual speeds with average site features (non-turners only) 1.8 2.5 3.9 5.5 -1.0 -0.7ATC speeds 3.5 6.4 2.9 5.4 3.2 5.9

0

5

10

15

20

25

30

35

40

45

0 20 40 60 80 100 120Y Distance (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

0

10

20

30

40

50

60

0 20 40 60 80 100 120

Forward visibility (m)

Spee

d (m

ph)

Road Width = 5mRoad Width = 6mRoad Width = 7mRoad Width = 8mRoad Width = 9mRoad Width = 10m

Figure 5.6 Junction model ATC speeds

18

Parking and block paving were found to reduce vehiclespeeds by approximately 2 to 5 mph, though there is anindication from the ATC data that parking near junctionscould have a slightly stronger effect on speeds thanparking on a link. The models disagree as to the effect ofjunction markings. The manual data implies that withoutjunction markings, speeds were slightly greater. However,the ATC data indicates drivers reduced their speeds bybetween 3 and 6 mph when markings were absent.

The models agree on the order of magnitude of speedreductions through width reductions. For example at ajunction with Y-distance at 4.5 metres of 40 metres, areduction of road width from 10 to 5 metres would beexpected to reduce approach speeds by between 6 and 9 mph.

Both models also agree that drivers reduced theirapproach speed if the visibility at the junction (Y-distance)was less. The manual data predicts a reduction ofapproximately 5 mph if the Y-distance is reduced from110 to 20 metres at a junction where the road width is5 metres. However, the ATC data considers that drivers aremore sensitive to a lack of visibility. Under the sameconditions the resulting model predicts that drivers wouldreduce their speed by 11 mph.

6 Modelled safety impactsDrivers have been shown to alter their speed according tothe conditions on the road. Reductions through the type ofroad surface and speed reduction measures (for examplespeed humps) almost certainly improve safety at a site.However, the implication of reduced speeds owing toreductions in forward visibility on links and visibility atjunctions is less clear. The reduced speed results in asmaller stopping distance requirement, but less distancewill generally be available for stopping when a hazardbecomes visible. So, it is a question of whether driversslow enough to make the junctions and links with limitedvisibility as safe as those with greater visibility.

This section considers the predicted effect of limitedvisibility on speeds, and models a number of criticalsituations to ascertain whether the speed reductionscompensate for the lack of visibility. The average speedspredicted by the regression models formed on the ATCdata are assumed within this modelling as they are basedupon the largest data set, and the predictions are inapproximate agreement with those formed on the manuallycollected data.

6.1 Braking modellingIt is possible to model relative safety of the schemes byconsidering the ability of the drivers to stop under differentconditions. In order to create these braking models thefollowing has been assumed (see Appendix C):! The average perception-reaction time of a driver is 1.4

seconds when stopping in response to a hazard. This is aconservative estimate for the average driver, for exampleOlson (1997) reviewed 27 driver perception-reaction timestudies and concludes a great deal of data suggest thatmost drivers (i.e. about 85%) should begin to respond by

about 1.5 seconds after first possible visibility of theobject or condition of concern. Guidance in DB32 isbased on an assumed time of 2 seconds.

! The average deceleration rate of drivers stopping is4.5 ms-2. This is approximately half the maximumdecelleration that can be achieved by cars under favourableconditions, and is consistent with firm braking (see, forexample, Auto Express, 2005). Guidance in DB32 is basedon an assumed rate of 2.5 ms-2, approximately equivalent tostopping on snow without skidding.

For example, consider a driver travelling at 30 mph(13.4 ms-1). If the driver reacts to a danger by stoppingthen they take 13.4 1.4 = 18.8 metres to react to thedanger. Using standard equations of motion the drivertakes (13.4 13.4)/(2 4.5) = 20 metres to becomestationary when travelling at a constant rate ofdeceleration. That is, it takes a driver a total distance of38.8 metres to stop from a speed of 30 mph.

6.2 Stopping distances on linksA range of links have been examined in which theforward visibility generally varied from less than 10 metresto approximately 100 metres. It was found that theaverage speed of drivers reduced with forward visibility.In addition, block paving and parking on the link werefound to affect speeds. To remove this complication,within this modelling it is assumed that the link has atarmac surface and no parking is permitted on the link.Furthermore, it was found that link width also influencesspeeds. It is therefore assumed that the link is either 5 or 9metres wide, i.e. a narrow or wide road.

The situation considered is if an event occurs, forexample a pedestrian stepping into the road at the limit ofthe drivers forward visibility. It is assumed that the driverwill react as fast as possible and apply a fairly high averagedeceleration to stop their vehicle, as discussed in Section 6.1above. The distance required to stop is compared with thedistance available for road widths of 5 and 9 metres inFigures 6.1 and 6.2 respectively, and Table 6.1. Further, thedistance required to stop assuming drivers had not alteredtheir speeds owing to the reduced forward visibility isincluded for reference.

Initially it was assumed that drivers did not adapt theirspeeds as forward visibility reduced, i.e. that they did notperceive limited visibility as a danger and react to it. Thenthe model predicts that a collision would occur if visibilitywas less than 40 to 50 metres depending on the width ofthe road.

According to the regression modelling, drivers adaptedtheir speeds. However, the reduction does not fullycompensate for the reduction in forward visibility. That isthe margin for error is reduced as visibility falls. Forexample, with a forward visibility of 100 metres, should apedestrian appear at the limit of their vision a driver ismodelled as being able to stop at a distance ofapproximately 60 metres before the pedestrian. However,the model predicts that the driver would be able to stopwith just over 20 metres to spare under the sameconditions if the forward visibility was 40 metres. This

19

reduces to only 5 metres to spare with a forward visibilityof 20 metres.

Given the driver behaviour observed and the modellingassumptions, it would be expected that schemes in whichforward visibility is less than 40 metres could bereasonably safe for average drivers under these conditionsowing to the reduction in drivers approach speed.However, the margin for error decreases continually withforward visibility, and schemes with forward visibility onlinks of less than 20 metres are relatively unsafe, in theenvironments studied, unless other speed reductionfeatures are incorporated.

6.3 Stopping distances at junctionsA range of junctions has been examined in which theY-distance generally varied from less than 10 metres toapproximately 100 metres. It was found that the averagespeed of drivers reduced with Y-distance. Also, blockpaving, parking and the presence of lines to indicatepriority at the junction were found to affect speeds. Withinthis modelling it is assumed that the junction has a tarmacsurface, there is no parking at the junctions and no lines toindicate priority.

0

20

40

60

80

100

120

0 20 40 60 80 100 120

Forward visibility (metres)

Tota

l sto

ppin

g di

stan

ce (m

etres

)

Total stopping distance (m)Stopping distance if speeds not altered (m)Available distance (m)

Figure 6.1 Modelled stopping distances on a link of width 5 metres

Table 6.1 Modelled stopping distances for links ofvarying width

Stoppingdistance Estimated

Forward Road if speeds stopping Distancevisibility width not altered distance remaining(m) (m) (m) (m) (m)

20 5 41.1 13.2 6.87 44.0 14.1 5.99 47.1 14.9 5.1

40 5 41.1 16.8 23.27 44.0 17.9 22.19 47.1 19.0 21.0

60 5 41.1 21.5 38.57 44.0 22.9 37.19 47.1 24.4 35.6

80 5 41.1 27.7 52.37 44.0 29.6 50.49 47.1 31.6 48.4

100 5 41.1 35.9 64.17 44.0 38.4 61.69 47.1 41.1 58.9

0

20

40

60

80

100

120

0 20 40 60 80 100 120Forward visibility (metres)

Tota

l sto

ppin

g di

stan

ce (m

etres

)

Total stopping distance (m)Stopping distance if speeds not altered (m)Available distance (m)

Figure 6.2 Modelled stopping distances on a link of width 9 metres

20

The situation considered is when two vehicles approach ajunction. The first vehicle is on a minor arm of the junctionand the driver is assumed to show caution. They thereforedecrease speed from the modelled (regression) approachspeed (to 5 mph) as they approach over the 50 metres beforethe junction. The other vehicle has priority and the driverwill approach at the modelled approach speed.

Modelling has been used to consider the effect of thefirst driver seeing the second vehicle on its approach. Itconsiders the distance before the stop line at which theyneed to see the vehicle in order that the driver can react(taking 1.4 seconds) and then apply a higher deceleration(4.5 m/s-2) and be stationary at the stop line. These havebeen calculated using standard equations of motion, andare shown in Figure 6.3.

The driver approaching on the minor arm of the junctionand showing caution needs to see the vehicle approachingon the major arm when they are more than approximately4 and 6 metres from the junction in order to stop. Oncecloser to the junction they are committed to entering it.That is, drivers make a decision at or before 4 metres fromthe junction as to whether to stop or continue. This is inagreement with the regression modelling findings. Theregression models found that the visibilities measured at4.5 metres were the best predictors of approach speed, sodrivers appear to be making a decision on whether to stop

or continue at the junction based upon the information theyhave available at this distance.

Modelling has also been used to consider the effect ofthe second driver seeing the first vehicle on its approach. Itconsiders the deceleration required for the vehicle to stopbefore the junction if the driver sees the first vehicle whenat the Y-distance before the junction and then deceleratesafter the reaction time of 1.4 seconds, Figure 6.4.

If the visibility at the junction is greater than 40 metresa driver travelling at 30 mph should easily be able tostop. Drivers have been shown to reduce their approachspeed as the Y-distance reduces. However, it isinsufficient to fully compensate for the potential reducedstopping distance available.

The predicted reduction in approach speed results indrivers being able to stop using reasonable deceleration rates(less than 4.5 ms-2) under the modelled conditions if theY-distance is between 20 and 40 metres. Below 20 metresthe model implies it is unlikely that the driver will be ableto stop before the junction, and an accident would be morelikely to occur.

6.4 Implications of modelled situationsDB32 guidelines indicate that the Y-distance at a junctionwhere the speeds are 25 to 30 mph should not be less than45 to 60 metres. Regression modelling on a road with a

0

1

2

3

4

5

6

7

0 20 40 60 80 100 120

Y Distance (metres)

Dis

tanc

e fro

m s

top

line

(metr

es)

Figure 6.3 Distance at which first vehicle must see second vehicle in order to stop

0

1

2

3

4

5

6

0 20 40 60 80 100 120

Y Distance = Distance when driver decides to stop (metres)

Dec

eler

atio

n re

quire

d (m

/s/s)

Figure 6.4 Deceleration rate required by second vehicle to stop if seeing first vehicle at Y-distance metres from junction

21

width of 9 metres indicates that the average approachspeed of a vehicle travelling towards a junction with suchlines of sight would be 27 to 29 mph.

Modelling indicates that a vehicle approaching from aminor road has a decision point at approximately 4 to 6metres from the junction after which they are committedto entering the junction. However, a vehicle on the majorarm of the junction with a Y-distance of 45 to 60 metresshould be able to comfortably stop should such acircumstance occur.

With smaller Y-distances, down to approximately20 metres, the situation also appears relatively safe foraverage drivers. The extra caution shown by drivers on theminor road, and therefore the reduced approach speeds,result in them needing to decelerate at up to 3ms-2 in orderto stop safely: less than the usually acceptable limit.

Below Y-distances of 20 metres, the modelling indicatesthat the reduction in speed observed amongst drivers isinsufficient for the junction to be safe.

Observed standard deviations in approach speed implythat the 85th percentile of approach speeds could be4 to 6 mph greater than the average assuming a normaldistribution. Modelling implies drivers at these higherspeeds would be able to safely stop if the Y-distance is30 to 35 metres. Consequently, Y-distances between 20and 40 metres, and slightly below the recommended valuesmay be possible, but caution needs to be shown as thepercentage of drivers approaching at speeds that requirelarge decelerations if a conflict occurs quickly increases asY-distance decreases.

Considering links with a forward visibility of over40 metres, modelling indicates average drivers would beable to react to and stop safely before reaching a stationaryobject (say a pedestrian) appearing at the limit of theirvisibility. In fact the braking model indicates that anaverage driver would be able to stop with a margin forerror (i.e. expected distance between the resultingstationary vehicle and stationary object) of over 20 metres.

However, this margin of error reduces to approximately5 metres if the forward visibility reduces to 20 metres eventaking into account the reduction in speed observedamongst drivers in these situations. That is, the extent thatdrivers slow down as forward visibility reduces isinsufficient to result in geometries with a forward visibilityof less than 20 metres being safe. This modelling wasbased upon a tarmac surface and no parking on the link:both of which reduced link speeds. Therefore acombination of speed reducing measures may furtherreduce speeds to safely permit lower forward visibilities.

Observed standard deviations in link speeds indicate thatthe 85th percentile of link speeds could be 3 to 8 mphgreater than the average. Modelling implies drivers at thesehigher speeds would be able to stop with a margin of errorof 8 metres when the forward visibility is 40 metres and4 metres if the forward visibility was 35 metres.Consequently, forward visibilities of between 20 and 40metres might be considered, but again caution needs to beshown as the percentage of drivers approaching at speedswhere the margin of error is small quickly increase.

7 Observed safetyRoad safety can be explored by observing behaviour andanalysing trends in accident statistics. An investigation intoobserved behaviour showed that drivers adapted theirspeed on links where forward visibility was low and inresponse to road width and other speed limiting factors.The same was also seen at junctions where visibility, theY-distance, was low. However, although these modelledresponses indicated relatively small visibilities could beconsidered, drivers adapted insufficiently for visibilitiesbelow 20 metres (and possibly higher) to be as safe aslinks and junctions with higher visibilities.

The actual effect on the number of accidents was alsoexplored by collecting information on the number ofaccidents each year between 1995 and 2005 on all the 20sites (excluding Ipswich). However, where sites were built,or modified, between these dates, only data from the yearssince the site was in its current form were considered.Within the 19 sites, 261 links and junctions were studied,and at these features there was an average of 8.6 years ofaccident data at the 187 junctions and 8.3 years of data onthe 74 links.

Over all junctions there was an average of 0.07 accidentsper year, whilst at links the average was 0.03 accidents peryear. Consequently, it would appear that accidents are moreprevalent at junctions than on links. For this reason,junctions can be considered the most important feature withrespect to the number of accidents occurring on theresidential sites studied.

The STATS19 database classifies all accidents accordingto the severity of the injury to each of the casualties. Thetotal number of accidents for all ten years at the 19 sites (i.e.including dates before the sites were inhabited) shows that98% of the casualties at junctions were slightly (and 2%were fatally) injured, but just over half the casualties onlinks were seriously injured and the others were slightlyinjured. It is possible that this difference in severity is due tothe lower speeds of vehicles approaching junctions. Overallonly 27 casualties were seriously, or fatally, injured. Henceall accidents are considered together.

7.1 BelgraviaSection 4.1 showed that Belgravia was anomalous. Thedevelopment had a grid layout which resulted in largevisibilities on the links and at junctions. This combinedwith wide roads resulted in high vehicle speeds. Inaddition, the vehicle flows within this site weresignificantly higher than those at the other sites.

Considering all these factors, it is not surprising thatBelgravia also has anomalous accident data (in that it hadmore accidents than the other sites). Of the 131 accidentsoccurring on all 19 sites, 65 occurred within Belgravia. Thehigh incidence of accidents is shown in Figure 7.1 (note thatthe stars indicate the location of the accidents, some ofwhich denote multiple accidents at the same location).

Excluding Belgravia, over all junctions there was anaverage of 0.034 accidents per year, whilst on links theaverage was 0.028 accidents per year. Consequently, onthe remaining sites there appears to be an approximatelyequal likelihood of an accident at a junction, or on a link.

22

Table 7.1 Accidents with respect to visibility

Total Average numberY Distance Number of accidents of accidents per(metres) junctions per year junction per year

0 to 25 107 2.7 0.0325 to 50 48 2.6 0.0550 to 80 14 0.3 0.02Over 80 15 0.3 0.02

467

5

LB

8

1 to

4

LB

14

12a

12

7.1m

EATON SQUARE

EATO

N SQU

ARE

18

EATO

N SQU

ARE

Eaton

Squa

re

Garde

ns

EATO

N SQU

ARE

117

118 7.5m

12

UPPER BELGRAVE

STREET

KNIG

HTSBRIDGE AND

BELGRAVIA W

ARD

13

15

14

2

2

118

11

KNIG

HTSB

RIDG

E AN

D BE

LGRA

VIA

WAR

D

115

6

8

10

22

12

111

109

7.9m

22

21

23

LB

6.9m

Tennis

Court

98.0m

LB

4

11

1

BM

8.12m

17

31

5

36

24

EATO

N SQ

UARE

ECCLESTON

STREET

6.9m

100

Bolivian

106

107

108

109

31

33

34

Embassy

8

Eaton

Squa

re Ga

rdens

6.9m

LB

7.8m

TCB

BM

7.87m

Eaton

Squa

re

Garde

ns

LB

34

6

93

BELGRAVE PLACE

ECCL

ESTO

N MEW

S

24

SQUA

RE

26

7

36

11

32

107

103

44 to 47

40

CITIES OF LONDON AND WESTMINSTER BORO CONST

38

39

41 to

43

EATO

N SQ

UARE

90

46

102

EATO

N

95

98

32

19Nuffield Club

Belgian Embassy

103

LONDON EER

19

43

37

35b

35

ItalianEmbassy

Emba

ssy of

the

Hung

arian

Peop

le's Re

public

16

EATO

N MEW

S NOR

TH

54

56

14

15

86

38

83

LYALL STREET

84

BM 7.

00m

LB

69

BM

7.75m

EATO

N PL

ACE

68

8

57

45

CITY OF WESTMINSTER LB

WEST CENTRAL GL ASLY CONST

Club

82

The Irish

82

7

The Irish

Club

EATO

N SQU

ARE

77

80

44

LYALL STREET

8

82

9

2

84

64

46

EATO

N MEW

S NOR

TH

48

47

62

6

48

EATO

N SQU

ARE

51

EATO

N SQ

UARE

57 to 60

53 to

56

6.9m

Eaton

Squa

re

Garde

ns

The

Five

Fields

61

67

72

61

89

91

55

EATO

N ME

WS

NORT

H

73

53

108

57

58

103

93

EATON PLACE

101

102

104

99

7.3m

71

94

86

90

87

79

7.4m

62

2

LB

EATON SQUARE

66

66a

63

Crown Copyright 2006. All rights reserved. Licence number 100020449

LBPH

35Accident locations

Figure 7.1 Accidents in Belgravia

ST ANDREWS WARD

13

1

3

WEST CHELMSFORD CO CONSTHouse

1 to 6

2a

32.3m

40

36

2b

Upleatham

10

1

2

8

2

8

31

Path (um)

32.6m

191 to 2

3

BM 32.56m

11

23

11

17

19

8

24

FITZ

WAL

TER

PLAC

E

SACKVILLE CLOSE

WINDLEY TYE

9

Foremans

12

13 to 14

El S

ub S

ta

14

18

13

2931

EASTERN EER

22 20

CHELMSFORD DISTRICTROXWELL ROAD

ESSEX COUNTY

Crown Copyright 2006. All rights reserved. Licence number 100020449

54

Accident locations

Figure 7.3 Accidents in Windley Tye7.2 Accidents at junctionsThere are distinct variations in the number of accidentsoccurring at junctions in the sites, even when excludingBelgravia. For example, the number occurring in Glasgowis relatively high and the number in Windley Tye is low,see Figures 7.2 and 7.3 respectively.

Previous sections have shown that drivers alter their speedaccording to the junctions geometry. An investigation wastherefore conducted into whether junction layout affectedthe number of accidents. Visibility would be expected to beone of the most important variables with respect to accidentsat junctions. The average number of accidents per year atsites with different Y-distances at 4.5 metres from thejunction is shown in Table 7.1.

Figure 7.4 Junction accidents according to Y-distance

1501251007550250Y distance at 4.5 metres

0.40

0.20

0.00

Acci

dent

s pe

r year

(at ju

nctio

ns)

There appears to be little correlation between thenumber of accidents and visibility. Figure 7.4 shows thehigh variability present in the data.

Although a trend in the number of accidents appear tobe present, the variation and sample size preclude anypossibility of these being significant. A number of testswere performed. These included examining the percentageof junctions where accidents were observed according todifferent ranges of Y-distances. The conclusion was thatthe observed accident trends could have occurred as part ofnatural variation and it is possible that the actual numberof accidents is independent of the junction geometry. This

258

159 to

167

PINE

PLAC

E

146 to 150

130 to 134

136 to 144

19 to 27

1115

17

7

175

169

45

45 to

49

16

2435 to 43

28

to 32

ERROL GARDENS

GLASGOW PER 29 to

33

19 to 27GLASGOW

36

to 40

34

8

238

147 to

151

BENN

Y LY

NCH

COUR

T

16

11

ST

NINI

AN

TERR

ACE 36

56

40

7

137 to

143OLD

RUTHERGLEN ROAD

125 to 129

131 to

135

113 to

117

119 to

123

20

1517

6

PO

12 16

114 to 118

120 to 128

110

167

161

155

180-

154

61

PINE

PLAC

E

55

32

235

48

42

to 46

34

227

229

46

44

221 to

225

39

192190

184 to 1

88

215 to 219

CROW

N ST

REET

ERROL GARDENS

22

to 26

18

181

207 to 211

197 to

201

191 to

195

27

911

15

203 to

207

166 to 1

70

4-8

5-7

197

209 to 213

Supermarket

50

CROW

N ST

REET

9.1m

101

CAM

DEN

TERR

ACE

CUMBERLAND STREET

164

10.0m

10-14

201 to

205

221 to

225

16-18

5557

45-49

42

9 to 11

235

3739

227 to

233

170SAN

DIEF

IELD

ROAD

GLASGOW CITY

Supermarket

GORBALS

Play Area

Play Area

SAND

IEFI

ELD

ROAD

SCOTLAND EER

El Sub Sta

170

200

El SubSta

154

146 to 150

126 to 132

7 to

9

Play Area

NABURN GATE

KIDSTON

PLACE

23

15 to 19

CRESCENT

230

224

19

KIDSTON TERRACE

15

21 to 2328

22 26

11

ALEXANDER

GLASG

OW

SHETTLESTON P CO

NST

15

29

719 to 23

33

HAND

EL

PLAC

E

SOUTHSIDE CRESCENT

18

7

11

3

5 to

9

11 to

17

19 to

23

TCB

Crown Copyright 2006. All rights reserved. Licence number 100020449

GLASG

OW

HUTCHESONTO

WN

WAR

D

6 to

10

THIS

TLE

TERR

ACE

SOUTHSIDE CRESCENT

ALEXANDER

CRESCENT

25 to

27

11 to

1719

to 23

GO

RBALS

15

MAL

TA

TER

RAC

E

14

4

2

7

11

16

30

1921

12

to

16

18

to

22

24

to

28

3

Accident locations

Figure 7.2 Accidents in Glasgow

23

could have occurred because of the behaviouralmodifications explored in previous sections; alternatively,it is possible that the sample size was insufficient given thesize of the effect compared with the variation.

It would also be expected that the number of accidentsat a feature would be dependent on the traffic flow. Theobserved link flows were used to calculate the averagenumber of accidents per year for 1000 vehicles, these areshown in Figure 7.5.

Figure 7.5 Junction accidents weighted by flow accordingto Y-distance

1501251007550250Y distance at 4.5 metres

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0

Acci

dent

s pe

r year

per 1

000

vehi

cles

(at ju

nctio

ns)

GlasgowLichfieldManchesterLeicesterPortisheadBloxhamJerichoBeaulieuWindley TyeNewhallLavenhamTower HamletsEastleighChichesterNew Town, ReadingGuildfordLower EarleySite Number

1 to

3

129

8.6m

2022

124 to 134112