Research for MfS High Risk Collision Sites and Y Distance Visibility Date: September Vanguard Centre, University of Warwick Science Park, Sir William Lyons Road, Coventry CV EZ Tel: + () Fax: + () Email: info@tmsconsultancy.co.uk Web: www.tmsconsultancy.co.uk
Contents EXECUTIVE SUMARY.. INTRODUCTION METHODOLOGY... RESULTS CONCLUSIONS AND IMPLICATIONS FOR DESIGN. ADDENDUM ADDITIONAL RESEARCH ISSUES.... REFERENCES. AKNOWLEDGEMENTS. TMS Consultancy, September Page
Research for MfS High Risk Collision Sites and Y Distance Visibility Executive Summary. Introduction The approach to visibility at priority junctions has been to provide a minimum stopping sight distance value (SSD) appropriate to a particular design speed, for example m SSD for kph. The assumption made by some designers and road safety auditors is that this value provides a minimum road safety requirement, and that collision risk will increase if the SSD is not achieved. The purpose of this research was to examine whether a direct relationship can be established between variations in Y distance SSD and collision frequency at priority junctions.. Methodology Site selection A series of high risk priority junctions was identified as the basis for research. Uncontrolled X and T junctions were selected for all classes of road throughout all, and mph speed limits in Nottinghamshire County, Sandwell Metropolitan Borough, Lambeth, and Glasgow City council areas. For each area a list of all nonpedestrian collisions was ranked in descending order of collision total for a recent year period, with over collisions listed in total. Each location was then analysed in detail to identify specific collision characteristics. Collision analysis involving vehicles emerging from junctions into the path of main road vehicles, together with nose to tail shunts on the minor road were identified as the type of incident that could have been caused by poor visibility. The locations were then ranked in descending order of these types of crashes, and site visits were carried out at the worst sites. In addition to the potential poor visibility collisions, a record was made of collisions involving main road shunts, collisions involving main road PSV passengers, collisions involving main road large goods vehicles, and collisions involving main road two wheeled vehicles. There is a concern that these types of collisions could be over-represented at locations with poor visibility. Site visits Two investigators visited each location, and measured visibility to the left and to the right on the main road, from.m back from the main road channel. Visibility was measured from a height of.m, to a point at the kerb edge and a second point m out from the kerb edge, where observations showed that visibility increased. TMS Consultancy, September Page
. Summary of findings locations had three or more potential poor visibility collisions high risk sites in a five year period. of these were on mph roads, with three on mph roads. At of the locations the worst case visibility (either to the left or right) was restricted to less than m. Thus in relation to the total number of uncontrolled junctions that exist, the proportion of high risk sites where visibility is less than that recommended for kph in DMRB is likely to be very low. It is possible that some former high risk priority junctions have been converted to other forms of junction control; In two thirds of the cases where visibility was less than m, the restriction was due to parked vehicles or street furniture. It is not possible to determine whether the parking was present at the time of the collision; Linear regression to compare potential poor visibility collisions with Y distance shows the variation in collision frequency was explained by factors other than Y distance visibility, for a large number of different situations. Y distance can not be seen as a single deterministic factor at these high risk collision locations. For example, visibility to the right vs collision frequency; <m visibility to right from kerb vs "visi" collisions y =.x +. R =. Visibility A series of collision types at high risk locations where Y distance was less than m were compared with locations with more than m visibility. There were no statistically significant differences between the two sets of data. The data analysed included main road bus and large goods vehicle collisions; Collision type No & % in sites <m vis No & % in sites > m vis Potential vis collisions in dark (.%) (.%) Main road shunts (.%) (.%) Bus passenger (.%) (.%) Main road HGV (.%) (.%) Main road W (.%) (.%) TMS Consultancy, September Page
This study has been unable to demonstrate that road safety concerns regarding reduced Y distance are directly associated with increased collision risk at highrisk urban sites; Previous research for MfS demonstrated that main road speed is influenced by road width and forward visibility. Many of the locations in this study were straight roads with good forward visibility. The ability of the driver to stop is likely to be affected by more than just what is happening in the side road and an understanding of the factors influencing main road speed is important when assessing visibility requirements. TMS Consultancy, September Page
Introduction The literature review carried out by TRL for MfS highlighted a lack of robust evidence supporting DB standards, including visibility splays. TRL went on to examine the relationship between observed collision records and Y distance visibility at locations on lightly trafficked residential streets, concluding that the observed accident trends could have occurred as part of natural variation and it is possible that the actual number of accidents is independent of the junction geometry. This research formed part of the rationale for recommending within MfS that designers take a more flexible approach to Y distance visibility. The use of the SSD formula was recommended, resulting in the production of a table suggesting that for a kph design speed, an appropriate Y distance would be m. Many designers, and some safety auditors, have road safety concerns regarding a potential reduction in SSD at major/ minor junctions, which would result if MfS was adopted across a wider urban road network. These concerns are based on an intuitive approach to road safety, and an analysis of the constituent parts of the SSD formula. They include: The potential for left turn, right turn and straight ahead pull out from side road into path of main road collisions. As Y distance visibility reduces the driver on the main road does not have enough time to stop prior to a collision occurring; The potential for minor road shunts. The second vehicle runs into the rear of the lead vehicle, whose driver having started to emerge stops suddenly on seeing a main road vehicle (late) within a shortened visibility splay; The potential for main road shunts. The side road vehicle emerges within a short splay, causing the main road driver to stop suddenly, leading to a second main road vehicle running into the first; The potential for injury to main road PSV passengers, as the PSV driver brakes suddenly when a side road vehicle pulls out. There is concern that a PSV braking at.g (deceleration quoted in the MfS SSD formula) would lead to standing passengers falling over; Over-involvement of main road large goods vehicles and two wheelers. Again, concern has been expressed in relation to whether large goods vehicles can brake at.g. The research for MfS looked at cars only, and both goods vehicle and motor cycle braking systems are different to cars. Furthermore both motor cycle and pedal cycle riders could be lost behind point visibility restrictions within a splay. This research has been undertaken in an attempt to determine the extent to which these intuitive concerns about road safety at junctions are borne out in reported injury collisions. Collision occurrence is an important part of defining risk at junctions. Local highway authorities have for many years identified locations with high numbers of collisions of a similar type, and then successfully implemented site specific remedial treatments designed to remove or reduce the collision risk. intuitive approach based on a rational process of thinking through the likely collision scenarios if something changes in a particular way TMS Consultancy, September Page
. Methodology. Site selection Four British local highway authorities were contacted and each agreed to take part in the study by supplying collision data for a number of junctions within their areas. The four areas were Sandwell MBC (West Midlands), The London Borough of Lambeth, Nottinghamshire County Council (East Midlands), and Glasgow City Council (Scotland). Sandwell, Lambeth and Glasgow are all authorities with mainly continuous urban development, whereas Nottinghamshire is a predominantly rural county with a number of discrete urban areas within it. The objective of the analysis was to derive a series of sites with high collision records, where the type of collision occurring might be caused by poor visibility, and then to compare collisions with observed visibility. Give way or uncontrolled cross road and T-junctions were selected for all classes of road in, and mph speed limits. For each local authority a list of all nonpedestrian collision locations was ranked in descending order of collision total for a recent year period. Each location was then analysed in detail to identify specific collision characteristics.. Collision analysis In the residential collision analysis carried out for MfS, the roads were comparatively lightly trafficked. As a result the number of collisions at each location was very low, and TRL did not therefore distinguish between collision types. In this study the junctions are much busier, and involve a greater variety of traffic, so the number of collisions is understandably higher. It is therefore possible to disaggregate the collisions in order to determine whether those that appear intuitively to be related to visibility, can be shown to be so, from collision and site evidence. Collision diagrams were produced for every location from the ranked list of all nonpedestrian collisions at cross road and T junctions. The following information was recorded from the collision diagram analysis: The vehicle turning movement pattern; The collision severity; Daylight or darkness; Type of vehicles on side road and main road; In the first instance pull-out collisions and minor road shunts were summarised together as the most likely collision types to have been caused by poor visibility, and a ranked list of potential poor visibility collisions was produced. This provided the basis for locations to be examined in more detail through a site visit. Some main road shunts and PSV passenger collisions could be affected by poor visibility, as suggested earlier. However, the side road vehicle is often a ghost vehicle in these cases, and therefore not recorded on the Stats- form. It was NB this does NOT include the City of Nottingham, which is a Unitary Authority pedestrian collisions were not investigated in this study as the main concern expressed by designers and safety auditors related to vehicle collisions (see list in section ) left turn, right turn and straight across into path of main road. TMS Consultancy, September Page
decided to return to the main road shunts, PSV passengers, and main road HGV and two wheeler collisions once the visibility had been established.. Site visits Two investigators visited each location, and measured visibility to the left and to the right on the main road, from.m back from the main road channel. Visibility was measured from a height of.m, to a point at the kerb edge and a second point m from the kerb edge. A record of any point restrictions on visibility was also made. In addition, the following information was collected: Confirmation of speed limit; Type of carriageway and number of traffic lanes; Bus and/ or cycle lane present; A qualitative assessment of horizontal and vertical alignment of main and side roads; Any special site features. Site visits, together with photos for each site, were undertaken during December and January. the observers were checking whether an approaching car could be blocked so intervisibility at.m was the critical issue TMS Consultancy, September Page
. RESULTS. Individual local authorities Table summarises the number of locations examined as part of the collision study. The cross road junctions were split down into individual approaches for the purposes of the collision and subsequent site visit analyses. Table : locations examined in collision study Local authority No. T junctions Potential poor vis collisions Total collisions Sandwell./yr/site./yr/site Notts./yr/site./yr/site Lambeth./yr/site Glasgow./yr/site./yr/site./yr/site No. X road approaches Potential poor vis collisions./yr/site./yr/site./yr/site./yr/site Total collisions./yr/site./yr/site./yr/site./yr/site In Sandwell, Nottinghamshire and particularly in Lambeth, T junctions formed the main proportion of the study sites. The reverse was the case in Glasgow.. Combined collision data Tables & summarise the collision analysis for all four study areas. Table : potential poor visibility collisions T n= X n= Comb n= junc road No. % Ave/j No. % Ave/j No. % Ave/j RT.%..%..%. LT.%..%..%. St on.%..%..%. Minor shunt.%..%..%. Fatal.%..%..%. Ser %..%. %. Slight.%..%..%. Dark.%. %..%. Total %. %. %. average per junction for year period TMS Consultancy, September Page
Table : other collisions of interest T n= X n= Comb n= junc road No. % Ave/j No. % Ave/j No. % Ave/j Main.%..%..%. shunt PSV.%..%..%. pass HGV.%..%..%. main w main.%..%. %. Total %. %. %. Pot. visib colls.%..%. %.. Findings.. General Overall, a total of non-pedestrian collisions were investigated at,, and mph cross-roads and T junctions in four different local situations; (.%) of the collisions in the study as a whole involved a shunt collision on the main road; (%) of the collisions in the study as a whole involved a single vehicle bus collision in which a passenger was injured; (.%) of the collisions in the study as a whole involved a large goods vehicle on the main road either in a main road shunt or a collision with an emerging vehicle; (%) of the collisions in the study as a whole involved a two wheeled vehicle on the main road either in a main road shunt or a collision with an emerging vehicle; (%) of the collisions in the study as a whole involved a potential visibility related collision ; of the collisions in the study as a whole involved a potential visibility related collision occurring in hours of darkness. This represents.% of the total, which compares with an expected proportion of %. The difference between the study and control is indicative at the % level; the total exceeds the sum of the rows above, as there are significant numbers of other collisions which are of no particular interest to this study single vehicle bus collision no other vehicles involved, injury to bus passenger assumed to be due to braking and passenger falling over right, left turn or straight ahead pulling out from minor road vs main road, or shunt on minor road approach control data from RCGB collisions on built up roads TMS Consultancy, September Page
A total of fatal, serious and slight injury potential visibility collisions were recorded. The KSI proportion is.%, which compares with an expected proportion of.%... Cross- roads and T junctions (%) of the collisions took place at T junctions, and (%) took place at cross road approaches; there were T junctions, with an average of. collisions per junction per year; there were cross road approaches, with an average of. collisions per approach per year; (.%) of the collisions at T junctions involved a potential visibility related collision, whereas (.%) of the collisions at cross road approaches involved this type of collision; (.%) of the potential visibility collisions at T junctions involved a right turn, whereas (.%) of the potential visibility collisions at cross road approaches involved this type of collision; (.%) of the potential visibility collisions at T junctions involved a left turn, whereas (.%) of the potential visibility collisions at cross road approaches involved this type of collision; (.%) of the potential visibility collisions at T junctions were recorded as straight ahead movements, whereas (.%) of the potential visibility collisions at cross road approaches involved this type of collision; (.%) of the potential visibility collisions at T junctions involved a shunt on the minor road approach, whereas (.%) of the potential visibility collisions at cross road approaches involved this type of collision.; Figures & illustrate the potential visibility collision types for T junctions and cross road approaches. control data from RCGB non pedestrian collisions on built up roads cross road approach - one of the two minor road arms of the cross road junction TMS Consultancy, September Page
Figure : potential visibility collisions at T junctions All T junction sites - potential visibility collisions % % % % right turn left turn st on min shunt Figure : potential visibility collisions at cross road approaches All X road sites - potential visibility collisions % % % right turn left turn st on min shunt % (%) of the potential visibility collisions at T junctions involved a collision occurring in darkness, whereas (%) of the potential visibility collisions at cross road approaches involved this type of collision as shown in Figure. TMS Consultancy, September Page
Figure : potential visibility collisions in darkness Percentage potential visibility collisions in dark hours. t junction x roads combined. Site data The objective of the study was to determine the relationship between collisions and actual visibility measured on site, by analysing those locations where high numbers of collisions that could be caused by poor visibility have occurred. The locations generated a total of sites with between and potential visibility collisions in the year period, as shown in Table. Table : locations with potential visibility collisions No. T junctions No. X rd approaches Sandwell Nottinghamshire Lambeth Glasgow Total Visibility was measured at locations and all Y distances under m were recorded. Five cases were considered unsuitable due to recent changes to layout (eg the installation of a mini-roundabout), and one location on a mph road had been incorrectly recorded as mph on the Stats- records. The locations included cross road approaches and T junctions. There were no mph locations, mph locations, and just three mph locations, two of which were dual carriageways... What are the restrictions on visibility? A total of locations had visibility to the right to the kerb edge of less than m, whilst locations had visibility to the left restricted to less than m, shown in Table. TMS Consultancy, September Page
Table : restrictions to visibility Restriction Vis to left Vis to right Parking Street furniture Trees/vegetation Wall/ building Alignment Total The issue of parking is somewhat different to the other visibility restrictions. The table above records those cases in which parking was present at the time of the site visit. This does not necessarily mean that parking was present at the time of the collision this level of detail is not available within Stats- collision record analysis. The urban nature of the environment in the majority of cases would suggest that parking is common place, and therefore could have been present at the time of collisions at these locations... What other information was recorded? Just over % of the road types were two lane single carriageways, with a further % either or lane single. A further % were dual carriageway. A qualitative assessment of alignment was made. % of the locations had straight (as opposed to bendy) main road approaches, with % of main road approaches being flat. % of side roads approached the junction on an uphill gradient. of the locations had bus lanes ( of which were in Lambeth). In cases a pedestrian crossing was located on the main road within m of the junction.. Collision and visibility comparisons A series of graphs and tables have been prepared to compare the potential visibility collisions with observed visibility. The graphs examine those cases where visibility was measured (ie those locations with less than m Y distance visibility). In each case an R coefficient of determination has been obtained. An R of, say., would indicate that % of the variation in collisions could be explained by visibility, and that % could be explained by other variables... Visibility to the right Visibility to the right was measured to the kerb, and to m out from the kerb. In each case the variation in collision frequency was explained by factors other than Y distance visibility. The results are shown in Figures &. TMS Consultancy, September Page
Figure : visibility to the right at the kerb <m visibility to right from kerb vs "visi" collisions y =.x +. R =. Visibility Figure : visibility to the right m out from the kerb <m visibility to right from m out vs "visi" collisions y =.x +. R =. Visibility Table shows the number of potential visibility collisions per site per year for differing visibility ranges. in many of theses ranges the small number of sites provides insufficient data to demonstrate representative collision averages, so the data should be treated with caution TMS Consultancy, September Page
Table : Range of visibility to right and collision frequency Vis to right to kerb m range Vis range No. of sites No. collisions per year per site per year -.. -.. -.. -.. -.. -.. +.. Vis to right to kerb No. of No. Ave no. Ave no. Vis range sites collisions colls/site colls/site/yr -.. -.. +.. -.. -.. -.... Visibility to the left Visibility to the left was measured to the kerb, and to m out from the kerb. In each case the variation in collision frequency was explained by factors other than Y distance visibility. The results are shown in Figures &. Figure : visibility to the left at the kerb <m visibility to left from kerb vs "visi" collisions y = -.x +. R =. Visibility TMS Consultancy, September Page
Figure : visibility to the left m out from the kerb <m visibility to left from m out vs "visi" collisions y = -.x +. R =. Visibility Table shows the number of potential visibility collisions per site per year for differing visibility ranges. Table : Range of visibility to left and collision frequency Vis to left from kerb m range Vis Range No. of sites No.collisions per year per site per year -.. -.. -.. -.. -.. -.. +.. Vis to left to kerb Vis Range No. of Sites No.Collsions Ave no colls/site Ave no colls/site/yr -.. -.. +.. -.. -.. -.. in many of theses ranges the small number of sites provides insufficient data to demonstrate representative collision averages, so the data should be treated with caution TMS Consultancy, September Page
.. The effect of measuring m from the kerb The following observations were obtained by measuring to m from the kerb, compared to the kerb itself. Visibility to the right: average increase in visibility of.m range of m reduction in visibility to a.m increase Visibility to the left: average increase in visibility of.m range of.m reduction in visibility to a.m increase.. Cross road and T junctions T junctions were compared with cross road approaches. In each case the variation in collision frequency was explained by factors other than Y distance visibility. The results are shown in Figures. Figure : T junction visibility to right at kerb 'T' Junction - visibility to right from kerb vs "visi" collisions y =.x +. R =. Visibility TMS Consultancy, September Page
Figure : T junction visibility to left at kerb 'T' Junction - visibility to left from kerb vs "visi" collisions y =.x +. R =. Visibility Figure : cross road approach visibility to right at kerb 'X' roads approaches - visibility to right from kerb vs "visi" collisions y =.x +. R =. Visibility Figure : cross road approach visibility to left at kerb 'X' roads approaches - visibility to left from kerb vs "visi" collisions y = -.x +. R =. Visibility TMS Consultancy, September Page
.. Worst case visibility scenario The worst case (kerb or m out) visibility was compared with potential visibility collisions. The results are shown in Figures. In each case the variation in collision frequency was explained by factors other than Y distance visibility. The first graph looks at collisions vs worst case left or right visibility. Figure : worst case visibility to right or left <m to left or right - "worst case" visibility (kerb and m out) vs "visi" collisions y =.x +. R =. Visibility Figure : worst case visibility to right <m to right - "worst case" visibility (kerb and m out) vs "visi" collisions y =.x +. R =. Visibility TMS Consultancy, September Page
Figure : worst case visibility to left <m to left - "worst case" visibility (kerb and m out) vs "visi" collisions y = -.x +. R =. Visibility.. The locations with visibility under m The study identified sites where the worst case visibility was less than m. Ten of these were in Lambeth. A total of non-pedestrian collisions took place at these locations, of which (.%) were potential visibility collisions. In addition to pull-out collisions and minor road shunts, there are concerns that poor visibility could lead to main road shunts and main road bus passenger collisions, in a scenario where the main road vehicle stops suddenly to avoid a collision with an emerging vehicle. There are also concerns that main road HGVs and two wheeler vehicles could be over-represented in collisions at locations with poor visibility. Finally, there is a concern that collisions in dark conditions could be over-represented at these locations as dark conditions could compound other visibility issues. Table compares the proportion of collision types at sites with less than m visibility, with their proportions in the remainder of the collision sites visited. Table : collision types at sites with <m visibity Collision type No & % in sites <m vis No & % in sites > m vis Potential visibility collisions in dark (.%) (.%) Main road shunts (.%) (.%) Bus passenger (.%) (.%) Main road HGV (.%) (.%) Main road W (.%) (.%) Potential visibility collisions in the dark and bus passenger collisions occur at higher proportions at the poor visibility sites, but the differences are not statistically significant. The other factors are all under-represented compared to the sites with higher visibility. (Again, differences are not statistically significant.) TMS Consultancy, September Page
Figure compares the worst visibility at each site with two wheeler collisions on the main road. As with previous situations the variation in collision frequency was explained by factors other than Y distance visibility. Figure : main road two wheeler collisions at worst visibility sites "Worst visibility" at each site vs W collisions on main road W collisions main road y = -.x +. R =. visibility (m) TMS Consultancy, September Page
. Conclusions and implications for design A high risk collision site can be defined as one with multiple collisions of a similar type within a fixed time period. The total number of uncontrolled junctions within the four areas studied is unknown, but is likely to be several thousand in number. This study has revealed a total of locations with three or more potential visibility collisions high risk sites in a five year period. This is likely to be a small proportion of the total number of uncontrolled junctions throughout the study. In of the cases with potential visibility collisions the worst case visibility was restricted to less than m. In relation to the number of uncontrolled junctions that exist, the proportion of high risk sites where visibility is less than that recommended for kph in DMRB is likely to be low. Y distance visibility is mainly restricted by street furniture and parking rather than buildings or alignment. Visibility to the right increased by an average of.m, and to the left by an average of.m, when measured to a point m from the kerb edge as opposed to the kerb itself. Linear regression to compare collisions with Y distance visibility shows the variation in collision frequency was explained by factors other than Y distance visibility, for a large number of different situations. Y distance can not be seen as a single deterministic factor at these high risk collision locations. A set of collision locations with Y distance visibility under m was examined. These cases did not appear to have abnormally high proportions of main road shunt/ bus passenger/ main road HGV or two wheeler/ dark collisions. This study has been unable to demonstrate that the road safety concerns expressed regarding reduced Y distance visibility are directly associated with increased collision risk at high-risk sites. Collision risk is one aspect of road safety perception of risk and driver comfort are also important aspects. Previous research for MfS demonstrates that main road speed is influenced by road width and forward visibility. Many of the locations in this study were on straight roads with good forward visibility. The ability of the driver to stop is likely to be affected by more than just what is happening in the side road and an understanding of the factors influencing main road speed is important when assessing visibility requirements. TMS Consultancy, September Page
Addendum - additional research issues During the course of the research, a number of questions were raised by consultees at a workshop convened to discuss progress on MfS. This led to additional work being undertaken in an attempt to answer some of the questions. The issues, together with the findings, are set out below.. The influence of traffic flow The main study did not include an analysis of traffic flows on any of the routes. A collection of traffic flow and speed data for each of the locations would have been beyond the budget for the project which concentrated solely on the relationship between collision frequency and Y distance visibility. However, the routes that were studied are mainly busy radial roads, many of which are bus routes. Concern has been expressed that a reduction in visibility, whilst not necessarily compromising collision frequency, could lead to an increase in delay as drivers queue back on the minor road behind a lead driver taking more time to emerge. Some consultees felt that there could be a relationship between potential visibility collisions and traffic flow. For example, such collisions may be more frequent in congested periods of the day due to impatience on the part of emerging drivers, alternatively they may be more frequent in low flows as emerging drivers take risks of there being no opposing traffic on the main road. In order to contribute to the discussion on this point an analysis was undertaken of potential visibility collisions vs non-visibility collisions by time of day. The study involved examining collision data at the high risk sites visited during the main study. The results for all four study areas combined are shown in Figure. Figure : potential visibility collisions and other collisions by time of day Time of day comparison.... no. collisions : -.... : -. : - :........... : - :.. : - : : - : : - : : - :........ : - : : - : : - : : - :...... : - : : - : : - : time of day................ : - : : - : : - : : - : : - : : - : : - : : - : : - :..... Potential Visibility collisions Non - Visibility % of Total for Hour TMS Consultancy, September Page
The results show that the daily pattern of potential visibility collisions is similar to that for non-visibility collisions. There are statistically significant over-representations of potential visibility collisions between - and -hrs, and statistically significant over-representations of non-visibility collisions between - hrs. Each of these periods is outside of the morning and evening peak, at times of relatively free flow. The peak hour (congested) periods of -hrs and - hrs show very similar proportions of potential visibility collisions compared to non-visibility collisions. This analysis suggests that time of day does not have a major influence in terms of discriminating between potential visibility and non-visibility collisions.. The influence of parking Of the instances in which visibility was restricted to less than m, (%) restrictions were caused by parking at the time of the site visit. As noted in the main study, it is not possible to state with certainty that this restriction was present at the time of the collision. Indeed, there may well be cases where visibility was measured on site at a particular value, but that at the time of the collision it was further restricted by a temporarily parked, stopped or slow moving vehicle. In order to investigate the parking issue in more detail, an analysis was made of potential visibility collisions vs Y distance visibility at sites where parking was the main restriction, compared to sites where the restriction was some other feature. The results are shown in Figures &. Figure : potential visibility collisions at locations restricted by parking Locations where visibility restricted by parking No of collisions y = -.x +. R =. Visibility (m) TMS Consultancy, September Page
Figure : potential visibility collisions at locations restricted by other features Locations where visibility restricted by other features No of collisions y =.x +. R =. Visibility (m) The R values for both linear regressions show that less than % of the variation in collision frequency was explained by visibility at the parking sites, and that less than % of the variation in collision frequency was explained by visibility at the nonparking sites. The inference from this is that the main study findings are not unduly biased by the parking sites.. The issue of mis-matched visibility Concern was expressed that collisions might occur more frequently in situations where there was a mis-match in visibility, for example where visibility to the left is very poor, but visibility to the right is very good. Drivers emerging from the side road may over-concentrate on the poor visibility side, and then pull out in conflict with traffic from the good side. In order to test this, a graph was drawn of potential visibility collisions vs difference between visibility to the right and the left. The results are shown in Figure. TMS Consultancy, September Page
Figure : potential visibility collisions at sites with variance in visibility left to right Difference in visibility from right to left No of collisions y =.x +. R =. Visibility Difference (m) The very low R value obtained suggests that the issue of mis-matched visibility does not explain the variation in collision frequency.. The issue of choosing only high risk collision sites The main study selected those sites where a discernable pattern of a least potential visibility collisions had occurred in a year period. These locations were obtained from a more general high risk site list that had been subject to detailed collision analysis. Concern was expressed that a relationship might exist between collision frequency and visibility at a lower level than the cut-off used in the study. In order to investigate this further, visits were made to sites in Sandwell and Nottinghamshire where,, and potential visibility collisions had been identified from the data for the year period. A graph was then drawn for all sites (- collisions) for the two authorities, shown in Figure. Figure : Notts and Sandwell all sites vs worst case visibility Nottinghamshire & Sandwell <m Worst Case Visibility All Sites y =.x +. R =. Visibility (m) TMS Consultancy, September Page
The very low R value obtained suggests that the results from the main study for high risk sites are repeated here for all sites, and that, at least as far as Sandwell and Nottinghamshire are concerned, the variation in collision frequency at all locations is not explained by the variation in Y distance visibility. A sample of Sandwell locations not part of the main study was visited and Y distance visibility was measured. As the sites had not been previously identified it was known that each of them had no potential visibility collisions within the five year period. A comparison was made of the visibility at the new Sandwell sites, with visibility obtained from measurements at the previously visited collision sites. The results are shown in Table. Table : visibility range at new sites compared to main study sites (Sandwell) Vis range No. of new sites New sites % No. of main study sites m+.%.% -m.%.% <m.%.% Main study sites % It appears that there is very little difference in the range of visibilities measured at the potential visibility collision sites compared to the other sites. There is therefore no apparent bias towards a particular visibility range grouping in either the collision or non-collision sites. Concern has also been expressed that the worst visibility sites that resulted in collisions occurring have all been treated either with visibility improvements or with the implementation of traffic signals, and that there are therefore very few true poor visibility sites left. It was not possible to test this theory within the existing study parameters.. The issue of high speed roads Finally, questions were raised regarding whether the results from this study can be applied to higher speed roads and mph. It was beyond the parameters of this study to undertake a full comparison of collisions and visibility at a sample of sites on high speed roads. An examination of the main study sites on dual carriageways (where free flow speeds might be higher than on single carriageways), produced an R value of.. This implies that the variation in collision frequency was explained by factors other than Y distance visibility, in line with other findings throughout the study. visibility measured to the right at the kerb TMS Consultancy, September Page
Data was obtained from Nottinghamshire County Council for sites subject to a or mph speed limit, and the worst locations for potential visibility collisions were derived using similar techniques to those described in section.. The results are shown in Tables &. Table : potential visibility collisions on - mph roads in Nottinghamshire compared with those for main study Notts - Main study No. % No. % RT.%.% LT.%.% St on.%.% Minor shunt.%.% Fatal %.% Ser.% % Slight.%.% Total % % Despite the differences in proportions shown in the table above, for example with respect to right and left turners, the only statistically significant difference is in severity. There is a % confidence that the KSI proportion at the Nottinghamshire -mph sites is higher than the KSI proportion for the main study. Table : other collisions of interest on - mph roads in Nottinghamshire compared with those for main study Notts - Main study No. % No. % Main shunt.%.% PSV pass %.% HGV main.%.% w main.% % Total % % Potential visib colls.% % The only statistically significant difference here is with respect to the proportion of potential visibility collisions within the total number of collisions. There is a.% the total exceeds the sum of the rows above, as there are large numbers of other collisions which are of no particular interest to this study TMS Consultancy, September Page
confidence of a difference between the Nottinghamshire -mph sites and those in the main study. More work should be carried out on high speed roads to compare collision frequency with Y distance visibility. TMS Consultancy, September Page
. References Manual for Streets, Department for Transport, published by Thomas Telford, March The Manual for Streets: evidence and research, TRL Report, Road Casualties Great Britain, Annual Report, Department for Transport Geometric design of major/minor priority junctions, Highways Agency Design Manual for Roads and Bridges, TD /. Acknowledgements The study author would like to thank Gareth Coles (Nottinghamshire County Council), Abu Barkatoolah (London Borough of Lambeth), Keith Sansom (Sandwell MBC) and Tim Clarke (Glasgow City Council) for supplying collision data for this study. Steve Proctor TMS Consultancy, September TMS Consultancy, September Page