The identification of mistakes in road accident records: Part 1, locational variables
-
Upload
kevin-austin -
Category
Documents
-
view
213 -
download
0
Transcript of The identification of mistakes in road accident records: Part 1, locational variables
Pergamon Accid. Anal. and Rev., Vol. 27, No. 2, pp. 261-276, 1995
Copyright 0 1995 Elsevier Science Ltd Printed in the USA. All rights reserved
OCO14575/95 $6.00 + .OO
oool-4575(94)ooo65-4
THE IDENTIFICATION OF MISTAKES IN ROAD ACCIDENT RECORDS: PART 1,
LOCATIONAL VARIABLES
KEVIN AUSTIN*
Institute for Transport Studies, University of Leeds, Leeds, LS2 9JT United Kingdom
(Accepted 6 June 1994)
Abstract-The current method of checking police-reported road accident data involves a rigorous process of manual and computer validation, with the objective of removing all the errors that exist on the accident report forms. This paper shows how a geographic information system (GIS) can be used to identify mistakes that exist in several locational variables once this process has been undertaken. It compares items contained on the accident report form with accurate highway feature information obtained from other sources. There were less than 10% of mistakes for the variables of district, speed limit, road class, and road number; less than 20% for junction control, junction detail, and pedestrian crossing facilities; and over 20% for carriageway type. If highway data were routinely entered onto a GIS by all British highway authorities, the above variables might not need to be entered by the police on their accident report forms.
Keywords-Geographic information system, Mistakes, Road accident, Highway inventory data
1. INTRODUCTION
The ultimate objective of road safety engineering and education is to provide for the safe movement of people and goods throughout the highway network. The information from personal injury accidents col- lected by the police is used to identify locations and population groups that have higher levels of accidents than the norm. To be confident that the correct remedial schemes are undertaken, the acci- dent data need to be of good quality, that is, the items are recorded correctly with few missing values. If this is not the case, then assessments of safety prob- lems relating to certain highway features or locations could be inaccurate, leading to an inefficient alloca- tion of resources.
Several investigations have been undertaken to assess the validity of accident records. In the United States, Shinar, Trent, and McDonald (1983) com- pared the information on 124 police records with that collected by multi-disciplinary accident investi- gation (MDAI) teams. They found the most inaccu- rate highway feature data were gradient, speed limit, surface composition, and curvature. In South Aus-
*The author’s current address is Halcrow Fox and Associ- ates Ltd., Vineyard House, 44 Brook Green, London W6 7BY, United Kingdom.
261
tralia, Howard, Young, and Ellis (1979) found that controls upon the road, intersection type, traffic con- ditions, and gradient were the variables most often incorrectly recorded. Questionnaire studies of U.K. local authorities (Ibrahim and Silcock 1992; Austin 1993), though, have shown the location of the acci- dent to be the most inaccurately defined variable.
The mistakes made by the police must be ex- pected given the situation they face at the scene of a road traffic accident, but provided a validation system can identify all of them, then these mistakes will not affect the overall quality of the data, al- though any that remain can have a substantial impact on safety investigation. Those accidents that are coded incorrectly will alter the total number of acci- dents relating to a certain feature, whilst those that are wrongly located will alter the number of acci- dents at certain sites, reducing the validity of site- investigation studies.
A comprehensive computerised safety rec- ordkeeping system (Chatfield et al. 1985) has been suggested (which includes linking accident and high- way inventory files) as a way of identifying and moni- toring the high accident frequency sites and features. The adoption of this system to identify mistakes in the accident records was not proposed, possibly because the non-spatially represented data files
262 K. AUSTIN
make it difficult to identify errors. The utilization of a geographic information system (GIS) should resolve this problem, and the objective of this paper is to show how it can be used to identify mistakes in several locational variables that remain after the current validation process has been completed. A GIS is a computer program that identifies the loca- tion of an object and provides information about it, enabling spatial and statistical analyses to be per- formed. This project used PC ARC/INFO (Environ- mental Systems Research Institute 1990), which is the most popular GIS used by local authorities in Great Britain with a market share of around 22%, almost double that of the next most popular package, produced by Alper Systems (Campbell and Masser 1991). Digitised road centre-lines were used to repre- sent the map information in this study and consisted of Ordnance Survey Centre-line Alignment of Roads (OSCAR) data for a five-kilometre square of north Hull. Between 1987 and 1991 a total of 1,884 acci- dents occurred within this area. Certain items coded onto the accident records were compared with the same items taken from road network data, so that mistakes in the coding of these items and in the locating of accidents could be identified.
2. DATA COLLECTION AND VALIDATION PROCEDURE
In the United Kingdom, the responsibility for collecting road-accident data lies with the police. Information is collected either at the scene of an accident, or later, when an involved party reports the incident at a police station. There are no specific personnel whose duty it is to collect this information, and so all uniformed officers may be called upon to undertake this task. The Road Traffic Act (Depart- ment of Transport 1988) makes certain stipulations upon drivers of motor vehicles involved in road acci- dents, stating that:
If any other person is injured . . . you must report the accident to the police as soon as possible, and in any case within 24 hours.
For these accidents, the Department of Trans- port requires a form (STATS 19) to be completed (see Appendix A). This form is divided into the sec- tions of Accident Circumstances, Drivers and Vehi- cles, and Casualties. The information contained within the Accident Circumstances section includes highway features, environmental conditions, and time-related data. The Drivers and Vehicles section relates primarily to the vehicle movement and crash impact, although summary details of driver age, gen-
der, and alcohol impairment are included. The casu- alty section is predominately concerned with the movement of any pedestrians involved or the posi- tion on or inside the vehicle of other injured parties. Only one value can be entered for each item, which can be limiting if two features are present at a site, for example traffic signals and a give-way. However, the increase in complexity of coding and transcribing the data could not be warranted, given this relatively rare occurrence. The police also collect information concerning the clear language descriptions of the accident location and the accident event, although they are not included on the STATS 19 form.
The data entry and validation process described in this section relates to the county of Humberside, although most highway authorities in the United Kingdom maintain a similar system. The records are first transcribed onto a computer database by the police data-processing team, and mistakes are iden- tified by a manual check. These records are then validated using a computer program (STATS 21), which checks the consistency of the data. This pro- gram is composed of both intervariable tests (for example, if the Carriageway type variable is entered as a roundabout then the Junction Detail variable must be entered either as a roundabout or a mini- roundabout) and range checks (for example, the car- riageway type code must be between 0 and 9). Any records that do not meet the specified criteria are flagged and subsequently manually corrected.
The data are sent to the highway authority who code the location of the accident from the clear lan- guage description. This is characterised by a five- figure grid reference, which pinpoints the accident to within 10 metres of the initial point of impact. This type of locational positioning can sometimes present problems when intersections are in proxim- ity, since it is difficult to identify the number of accidents occurring at each. To resolve this prob- lem, some authorities have developed link-node sys- tems whereby each accident is coded with a unique intersection number. A second manual check is un- dertaken and the STATS 21 program is rerun to check the variables that have been altered. The data are then sent to the Department of Transport, who compile the national accident statistics (Department of Transport 1993).
The police are also informed of many damage- only accidents, and these are recorded in a separate database managed by the police. In Humberside, they are rarely used for accident investigation and so no grid references are assigned, hence the accu- racy of the locational variables relating to these re- cords could not be checked. In some authorities grid references are assigned to these accidents, so the
Mistakes in accident records: locational variables 263
validation procedure could be undertaken for these data.
It is inevitable that mistakes will be made, be- cause the police have many duties to perform at the scene of an accident. For accidents that the police do not attend, but are notified of by an involved party, mistakes are even more likely because most individuals are not trained to collect this informa- tion. The mistakes in the STATS 19 will not affect the quality of data, provided a system is available that can identify them. The following section ex- plains an improved system that quantifies the num- ber of mistakes remaining for certain location-based items after the current validation procedure has been undertaken.
3. DEVELOPMENT OF A NEW VALIDATION SYSTEM
3.1 Data sources This section describes the process required to
identify mistakes in the accident report form using a GIS. The locational variables investigated consisted Of:
1. Road class; 2. road number; 3. district; 4. speed limit ; 5. Pedestrian crossing facilities; 6. junction control; 7. junction detail; and 8. Carriageway type and markings.
The different items that can be coded for each of these variables are stated on the STATS 19 form (see Appendixes A and B). The correct< type of fea- ture at each location was obtained from a number of sources. Ordnance Survey maps provided infor- mation on road class, road number, and the district boundaries, whilst the location of speed-limit signs, pedestrian-crossing facilities, and junction-control type were obtained from paper maps within the Acci- dent Investigation Section of Humberside County Council. Information on junction detail and carriage- way type were also obtained from Ordnance Survey maps, although a field survey was required to check their accuracy. The information contained in these sources was likely to be more accurate than that from the police, because it is taken directly from maps and a more detailed investigation was under- taken to obtain the data. The coding of several items, such as whether an accident occurred at a T or Y junction is open to some interpretation. In this case, only one individual coded the network, so at least
Road number Road class Speed limit Carriageway type Junction detail Junction control
1 Split zones1 I I
, boukHT\
A Code
Fig. 1. The process to create a new highway feature file.
there would be consistency in the recording of these items.
3.2 Coding the highway features Alternative procedures to provide the coded in-
formation were required because of differences in the nature of the data used, and this is highlighted in Fig. 1. The OSCAR data contain digitised links and nodes, so forjunction detail and junction control all nodes relating to junctions were selected and saved to separate files as points, whilst for road class and number, speed limit, and carriageway type all roads were selected and saved to separate files as links. The location of pedestrian-crossing facilities were manually plotted into the GIS (using the digital map as a guide) and stored as separate points.
For each variable the appropriate file was se- lected and the computer drew a boundary around all the respective links or points. This was necessary because the use of five-figure grid references to lo- cate each accident means that they do not necessar- ily correspond to the road centre lines or junction nodes of the digital map. All accidents falling within this zone could then be considered to be associated with that feature. For the variables of road class, road number, junction detail, and junction control the radius of the zone was 24 m. This was based on:
264 K. AUSTIN
1. the average distance between the centre-line and the edge of the carriageway (3.6 m);
2. the maximum error in the Ordnance Survey maps (0.4 m);
3. the maximum distance from the edge of the junction for which an accident is coded as at the junction (20 m).
The distance of 24 m was also used for thek variables of carriageway type and speed limit to retain consistency in the investigation method. The Department of Transport specifies that any accident located within 50 m of a pedes- trian-crossing facility should have the crossing type coded, even if a pedestrian was not in- volved in the accident. The radius of the zone around each facility was therefore set at 50 m. This included some roads that were not associ- ated with the road where the facility is located, for example, those that were parallel to it. The sphere of influence was therefore nar- rowed to include only those accidents that oc- curred within 24 m of the centre line of that road. These polygon-based coverages were then saved to separate files, one for each variable.
With the existence of such a wide bound- ary, i.e. 24 m, some of the carriageway-type and speed-limit codes could be shown as incorrect even though this was not the case. This would occur when accidents were located on roads that existed within the boundary derived for another road. If a link-node system was adopted, which could automatically snap the accident to the intersection (if the accident was less than 20 m from it) or to the road centre line, a wide zone-boundary would not be needed. Unfortunately, the local authority who partici- pated in this survey had not adopted a link- node system and so a wide zone-boundary was required.
The next stage was to split these polygons at the relevant points. For road class, road num- ber, speed limit, and carriageway type, the zones were divided at the changeover point from one feature to another; although if this was at a junction, the zone was split 24 m from the center of the junction along the joining road. Some of the junctions or pedestrian-crossing facilities were sufficiently close to one another for the zones to merge, so the boundaries were split halfway between the two features. This was undertaken so that any accident located within this zone would be compared with just the code of the nearest junction or pedestrian- crossing facility. The district boundaries were
digitised into the computer as polygons, and no other spatial manipulation was required.
For each variable, the zones were selected and coded with the respective numeric value following the criteria stated on the STATS 19 form, for example, the zone representing an A class road was coded as 3. The areas outside the boundaries were coded as zero. The zone boundaries disaggregated by speed limit for the study area are shown in Fig. 2, and the width of these zones corresponds to the distance of 24 m from the centre lines as explained above.
3.3 Identifying mistakes in the accident variables A flow diagram of the procedure to identify mis-
takes in the accident variables is shown in Fig. 3. The accidents to be validated were selected and saved to a file. They were then located onto a digital map using the grid references stated on the accident record. The point-in-polygon command was used to link the accident and highway feature data so that mistakes in the STATS 19 records could be identi- fied. A graphical representation of how this was achieved is shown in Fig. 4. The accident coverage shows the location of all accidents, each specified with a unique identity number, which is used to relate to the attribute table containing the STATS 19 information. The highway feature boundaries also contain identity numbers so that speed limit from the highway-feature data can be accessed. The point- in-polygon process was undertaken to overlay the points from the accident coverage onto the polygons containing the highway-feature data. Information concerning the polygon that encloses each point was then added to the attribute table of the accident coverage.
An enquiry was undertaken to identify those accidents whose coding taken from the highway data was different from that on the accident record. The example in Fig. 4 shows that for the accident with identity number 1, an identical speed limit was re- corded on the highway feature and STATS 19 records and so would not be identified as incorrect. The accident with identity number 2 had a different STATS 19 speed-limit code to that revealed by the highway-feature data. This could have been because the speed limit was miscoded by the police or be- cause it was mislocated outside the 30 mph zone. The accident with identity number 3 was located outside a road boundary, so the speed limit recorded in the highway feature file was equal to zero, hence the accident was mislocated. The same procedure was adopted to identify discrepancies in the coding of the other variables, although a more complex enquiry was needed to identify mistakes in road class
Mistakes in accident records: locational variables 265
rj 30 mph n 60 mph
m 40 mph n 70 mph * N Fig. 2. A representation of the zone boundaries distributed by speed limit.
and number. The reason was that for accidents that occur at junctions, the class and number of the road where the accident occurred and the road that it joins are recorded as separate fields. Hence, if an accident was located within the boundary of a classi- fied road it would be identified as incorrect only if the highway feature code was different to the STATS 19 class or number in both fields.
For each variable investigated, a locational plot (see Fig. 5) and a table (see Table 1) were produced, which included all accidents identified as incorrect. The table contains the accident details including its reference number, the coding of the feature (from
the STATS 19 record and from the highway feature data), and a description of the location (used to de- termine whether the accident was located correctly). The grid references of those accidents that were incorrectly located were adjusted, whilst an incor- rect value for an investigated variable was altered to that of the highway-feature data.
3.4 Costs incurred to develop this system The costs involved in developing such a system
include those of buying the hardware and software required and maintaining the data. Hardware costs are rapidly falling, so adequate computer facilities
266 K. AUSTIN
highway feature data
Fig. 3. The process of identifying incorrectly coded or located accidents.
could be provided at little over f2000. The cost of GIS software ranges from approximately 22500 to f10,OOO depending on the level of sophistication. Because this system used the more basic GIS tech- niques, one of the less expensive packages was ade- quate. The highway-feature data items have to be located and coded manually, which is a time-con- suming and costly process for a large administrative area. However, most authorities in the United King- dom collect highway inventory data. This requires that teams of engineers travel around the road net- work to code into a computer the location and type of all highway features, including road type, signs, lines, and traffic-control devices. The location of these items could be automatically placed as points over the digital map, which removes the need for additional site visits. The codes for these points were also included, which reduces the amount of manual manipulation to identify and code the items. Unfor- tunately, the highway-inventory data for the area used in this study are stored in a link-node format and were not compatible with the GIS used without an additional piece of software, which could not be made available. However, this type of integration is possible and will mean that this type of application is readily worthwhile. Once the data are contained within the system, the costs of updating the informa-
tion will be minimal. For example, it is unlikely that many more than a dozen new pedestrian crossing facilities would be installed in an average-sized U.K. local authority in a year, and since it would take only a minute or so to enter the data for each facility, these tasks will be performed inexpensively. The cost of adopting such a system in the long run is therefore likely to be affordable to most U.K. local authorities.
4. RESULTS
The mistakes were divided into two categories; those for which the locational text did not corre- spond to the grid reference position (referred to as mislocated) and those for which the locational text corresponded to the grid location but for which there was a difference in the coding of the item between the STATS 19 records and the highway feature data (referred to as miscoded). For each of the variables the number of accidents that were miscoded and mislocated are shown in Table 2.
The mistakes were less than 10% for the vari- ables of district, speed limit, road class, and road number; less than 20% forjunction control, junction detail, and pedestrian crossing facilities; and over 20% for carriageway type. For all variables, the acci- dents that were mislocated were randomly distrib- uted throughout the network, but were usually close to the road where they should have been located. This indicates that the mistakes probably arose dur- ing the selection of the 10-m grid box in which the accident should have been located. Conversely, those accidents that were miscoded were overrepre- sented in several feature types. The following sec- tions contain a more detailed analysis of the results.
4.1 Road class Of those accidents that were miscoded (see
Table 3), 22 (81.5%) were coded as unclassified and located on the B-class road. This total represents 14.1% of the 156 accidents that actually occurred on this B road and indicates that in many instances the police considered that this road was unclassified.
4.2 Road number There was a total of 82 accidents identified as
incorrect, of which 78 were also identified when the road class variable was checked. In these in- stances, both road class and road number were in- correctly recorded. This is so because the existing process accepts as correct accidents that were coded as unclassified and containing no road number or those that were coded as classified and containing
Mistakes in accident records: locational variables 261
MAP PLOT ATtRIBUTE TABLE
Accident coverage
L I I
Highway feature coverage
Merged coverage of accidents and highway feature data
Fig. 4. A graphical representation of the point-in-polygon command.
a road number, even if they were incorrect. The accidents that were miscoded are shown in Table 4, and the same explanations apply to this variable as to road class. For the four additional accidents that were incorrect, road class had been coded correctly, but the number did not relate to the road where the accident took place.
4.3 District In total, 17 accidents were identified as incor-
rect and all were miscoded. One of these was located only 2 m from the boundary, and, because of possi- ble errors in the drawing of the district boundaries, one could not be certain that this was incorrect. Only 1 out of the 45 accidents located less than 100 m from the district boundary was coded incorrectly. This figure may be low because large signs have been erected at the boundaries stating the change from one district to another. Most of the mistakes
occurred at locations more than 300 m from the boundary points where the signs would not be visi- ble. These large signs are rarely erected in the rural areas, which may lead to a greater number of mis- takes in those regions.
4.4 Speed limit @ The GIS validation system identified 131 acci-
dents as being incorrectly recorded, but after com- parison with the locational text, three accidents that occurred on roads subject to a 30-mph speed limit were inside the 40 mph speed-limit boundary and so were correctly coded. Table 5 shows that four of these did not have a standard speed and so an extra enquiry using the current nongeographic system should be included to identify them. This would ac- cept speed limits of 20, 30, 40, 50, 60, and 70 mph only.
The majority of discrepancies (73.5%) were be-
268 K. AUSTIN
7% Accident with STATS 19 and highway data
different for speed limit
Fig. 5. The location of accidents with speed limit identified as incorrect.
Table 1. A sample printout of details from accidents with speed limit identified as incorrect
Accident reference number ST’ATS 19 speed Highway feature speed
limit limit Location
96887 30 40
16988 39590 60
9591 30 159696 30 113591 30 143290 30 47290 40
125187 30 61090 30
40 30
40
40 0
40 0
30
0 0
Fairfax Ave/Bricknell Ave XRDS, Hull A1079 Beverley RdlSculcoates
La/Queens Rd Al 174 About 200 m north of Dunswell
Rbout B1233 Cottingham Rd/Hall Rd/Hotham Rd Thwaite St 50m East JW The Paddock Hull Rd JW Bricknell Ave Cottingham Rd JW Hardy St Cottingham Rd 100 yd west JW Newland
Ave A1079 Beverley Rd JW Pearson Ave, hull A1079 Beverlev Rd JW Grove St. hull
Mistakes in accident records: locational variables 269
Table 2. The number of accidents miscoded and mislocated for Table 4. A comparison in the coding of road number between several variables on the STATS 19 record STATS 19 records and highway feature data
Variable Number Number Percent miscoded
miscoded mislocated or mislocated
Road class 27 51 4.1 Road number 31 51 4.4 Speed limit 83 45 6.8 District 16 0 0.8 Pedestrian facilities 282 46 15.3 Carriageway type 346 48 20.9 Junction detail 115 121 12.5 Junction control 127 121 13.2
tween 30 and 40 mph roads, which was to be ex- pected given that 97.8% of accidents were subject to a speed limit within this range. The data recorded in Table 5 reveal that the proportion of coding mis- takes for accidents subject to 60- and ‘IO-mph speed limits was significantly higher than that for 30- and 40-mph roads (x2 = 85.4 with 1 df, p < .Ol). This indicates the difficulties in coding this variable in the peripheral sections of the urban areas where the speed limit may not necessarily relate to the surrounding land use that usually dictates it.
4.5 Pedestrian crossing facilities A comparison of the coding between the STATS
19 records and the highway feature data is shown in Table 6. The location of refuges were not included in this study because the installation dates were not obtainable. The validation procedure identified 3.2 times as many accidents located within the 50-m boundary but coded as outside than those located outside the 50-m boundary and coded as inside. This bias leads to pedestrian crossing facilities being shown to be safer than they actually are. To correct this, the number of accidents that occurred within 50 m of a pedestrian-crossing facility should increase by 37.9%-this ranges from 11.6% at zebra cross- ings to 350% at subways. The level of coding for subways was lower probably because they are not at grade and so would be difficult to identify.
Table 3. A comparison in the coding of road class between the STATS 19 records and highway feature data
Highway’ feature road class
STATS 19 road class
A B Unclassified
A B
500 - - 134
5 22
Unclassified
-
1223
Highway feature road number
STATS 19 road number 1079 1165 1174 1233 None
1079 422 1165 - 67 z 1 z 1174 1 - 9 1233 w z None -2 5 q 22 1221 Other 1 - - - 2
An additional 23 accidents were located within 50 m “other light-controlled crossing” (i.e., signals) but were coded as near a refuge. At these sites, the refuges were used to stagger the pedestrian cross- ings, although the item should be coded as signals since refuges would be coded only if no other con- trols are present. Thirty accidents that were actually located within 50 m of a signal were recorded as occurring within 50 m of a pelican crossing. This type of control operates in a similar way to a pelican, although pelican crossings would be located away from junctions.
4.6 Junction detail Private drives joining the highway and other
junctions (such as alleyways) were not coded onto the map, so accidents occurring at these types of junction were not checked for mistakes. Of the 121 accidents that were mislocated, 85.5% were coded as 20 m or less from a junction but were located outside this boundary. This may be lower than the true number of mislocated accidents because if the clear language description was not specific to a loca- tion then the accident was considered to be mis- coded. This was especially the case for nonjunction accidents, as only one road name is usually included. Table 7 compares the coding of junction detail be- tween the STATS 19 records and the highway fea- ture data and shows that there were over three times as many accidents located within the junction
Table 5. A comparison in the coding of speed limit between STATS 19 records and highway feature data
Highway feature speed limit
STATS 19 speed limit 30 40 60 70
30 1677 3 40 20 2 z 60 2 1 ; - 70 - 1 3 - Other 4 - - -
210 K. AUSTIN
Table 6. A comparison in the coding of pedestrian crossing facilities between STATS 19 records and highway feature data
Highway feature pedestrian crossing facilities
STATS 19 pedestrian crossing facilities Zebra Pelican Signal Subway Refuge Outside
Zebra 60 us
- - - 10 Pelican - 30 - - 36 Signal - - 42 - 7 Subway - - - J 1 Refuge - Outside 18 G
23 - 17 - 54 22 - 1349
boundary but coded as “not at a junction” than vice versa. As a result of this bias, the number ofjunction accidents is underestimated by 3%.
T-junctions accounted for 63.5% of the mis- coded accidents, which was to be expected because 65.7% of junction accidents occurred at this type of facility. The same comparison can be made for crossroads; this represents 22.4% of junction acci- dents and 19.1% of those that were miscoded.
There were 44 accidents recorded as occurring at a junction but carrying the incorrect junction code; 66.3% of these codes were discrepancies be- tween crossroads and T-junctions. This highlights the difficulty in distinguishing between these types of junction.
4.7 Junction control The accuracy of the coding of uncontrolled
junctions was not investigated because the majority of accidents (66.9%) occurred at the junction of pri- vate drives and the highway or at “otherjunctions.” Table 8 compares the coding of junction control be- tween the STATS 19 records and highway feature data and reveals that all the 71 accidents that were incorrectly coded as inside or outside the function boundary were identified in the junction-detail section.
In addition, 49 accidents coded as “give-way” were within the boundary of traffic signals. Both “give-way” and traffic signals existed at these sites, so all of these accidents must relate to signals as well as to give-ways (except possibly for rear-end shunts on the give-way leg). The manual of the De- partment of Transport (1994) gives no guidance as to which type of junction control should be coded in these circumstances, although, as traffic signals are the most dominant control type, it is probably appropriate to use this value.
4.8 Carriageway type markings A total of 402 accidents were identified as incor-
rect for this variable, although eight of these were correctly coded but located inside the boundary of another carriageway feature. A comparison of the coding between the STATS 19 records and the high- way feature data is shown in Table 9 and revealed that the items most often miscoded were dual two- lane, single three-lane, and single four or more lanes. This was so because of difficulties in coding the carriageway type at junctions with turning lanes and along roads with bus lanes.
Forty-three miscoded accidents were located at junctions where the number of lanes were increased to facilitate turning traffic; these included (a) 14 acci-
Table 7. A comparison in the coding of junction detail between STATS 19 records and highway feature data
Highway feature junction detail
Not within STATS 19 junction detail 20 metres
Not within 20 metres 543 Roundabout 1 Mini-roundabout -
T 13 Y -
Slip road -
Crossroads 3 Multiple -
Private drive -
Other -
Mini- Slip Private Roundabout roundabout T Y road Crossroads Multiple drive Other
- 46 - - 8 - - - 69 - - -
1 : _‘zz 1 _ - - - - 756 1 - 11 - - - - - 2s - 3 --- - - I-_ - - - -
- - 17 - - 274 - - - - - - 3-_ - - -
- - --- - - 109 - - 3-_ - - - 9
Mistakes in accident records: locational variables 271
Table 8. A comparison in the coding ofjunction control between STATS 19 records and highway feature data
Highway feature junction control STATS 19 junction control Give-way Stop Signal Uncontrolled Outside
Give-way 876 - 49 - 13 stop 3-- - Signal 4 - 195 - 4 Uncontrolled - - 142 Outside 52 - -2 - 543
dents coded as single three-lane when the general carriageway type was single two-lane; (b) 13 acci- dents coded as single four-lane when the general carriageway type was single two-lane; (c) 16 acci- dents coded as dual three-or-more-lanes each way when the general carriageway type was dual two- lanes each way.
The manual of the Department of Transport (1994) gives no guidance as to whether this item should be coded to include these lanes or not. How- ever, the general carriageway type would be more useful for analysis purposes.
Of the 346 accidents that were miscoded, 192 occurred along a 2.75km stretch of road. This road included three- and four-lane single carriageway in- corporating a bus lane and a section of two-lane single carriageway road without a bus lane due to road width limitations. Only 29.8% of accidents along this stretch of road were coded correctly and the accuracy for each section type was (a) 58.2% for single four-lane; (b) 82.6% for single two-lane; (c) 1.4% for single three lane with a 5-m lane in one direction and two 2.5-m lanes in the opposite direction (one of which was a bus lane).
These results indicate that bus lanes were gener- ally not considered as separate lanes in carriageway type. They should be included because vehicles use these lanes at all times, even though some classes of traffic are prohibited during certain periods.
A further study to investigate the accuracy of the Carriageway Type variable on roads with bus lanes would be useful to ascertain whether the re- sults of this study are unique. Clarification should also be given as to whether turning lanes at junctions should be added to the standard number of lanes.
CONCLUSIONS
This GIS-based validation system has verified mistakes in the coding and locating of accidents not identified by the current system. For the variables of Road Class, Road Number, Speed Limit, and District the level of mistakes were less than 10%. This figure is low probably because the features are unambiguous and are recognised by people with some local knowledge. The level of mistakes for Pedestrian-Crossing facilities, Junction Detail, and Junction Control were between 10% and 20%. All require the estimation of distance, so it is probably inevitable that a greater number of mistakes will be made. Carriageway Type was the variable most of- ten inaccurately coded, mainly due to uncertainty as to whether bus lanes should be coded as additional lanes. This may be a peculiarity of the sample, and a larger study should be undertaken to identify whether this phenomenon is unique.
If all highway authorities adopt this GIS-based system, the variables discussed in this paper could be removed from the police form altogether as they would be added to the computer database automati- cally once an accident has been placed onto the digital map. This means that over 2.1 million fewer items of data would need to be collected nationally by the police. This estimate is obtained by multi- plying the number of reported accidents occurring in Britain in 1992 (233,025) by the eight variables tested in this system and adding this to the number of junction accidents (143,161) multiplied by 2 (to account for the extra codes of road class and number
Table 9. A comparison in the coding of carriageway type or markings between STATS 19 records and highway feature data
Highway feature carriageway type or markings
STATS 19 carriageway Dual 2 Dual 3 or Single 2 Single 3 Single 4 or type or markings Roundabout One-way lanes more lanes Single track lanes lanes more lanes
Roundabout 71 3
- - - 3 - - One way - - - - 1 - - Dual 2 lanes - - 154 - - 7 2 15 Dual 3 or more lanes - - 21 - - 2 Single track - - - - 3 3
2 2 -
Single 2 lanes 1 4 11 - - 1127 113 52 Single 3 lanes - - 1 - - 26 25 14 Single 4 or more lanes - - - - - 22 32 155 Unknown - - - - - 10 - 1
272 K. AUSTIN
when an accident occurs at a junction). The annual costs of recording changes in the road network are likely to be minimal and the information could be used for other purposes, so this type of system is likely to be worthwhile.
Acknowledgments-The work undertaken in this project would not have been possible without the help of the Director of Techni- cal Services, Humberside County Council, the Chief Constable of Humberside Police, the Ordnance Survey for the use of the OSCAR road centre-line data and ESRI (U.K.) for the use of ARC/INFO GIS. The author would like to thank Peter Shenherd of Humberside County Council and Miles Tight and Howard Kirby of the University of Leeds for their encouragement and helpful suggestions, which have contributed to the work reported here. This work was undertaken under a U.K. research student- ship awarded by the Science and Engineering Research Council.
REFERENCES
Austin, K. P. The collection and use of additional sources of road safety data in highway authorities. Traffic Engi- neering and Control 34: 540-543; 1993.
Campbell, H.; Masser, D. The impact of GIS on local government in Great Britain. Conference of the Associ- ation of Geographic Information, Birmingham, En- gland, May 12-14, 1991.
Chatfield, B. V. et al. Introduction to comprehensive com- puterised safety recordkeeping systems (CCRS). Transportation Research Circular 293. Washington, DC: Transportation Research Board, National Re- search Council, 1985.
Department of Transport. Instructions for the completion of road accident reports: STATS 20. London: HMSO; 1994.
Department of Transport. Road accidents Great Britain 1992. London: HMSO; 1993.
Department of Transport. The road traffic act. London: HMSO; 1988.
Environmental Systems Research Institute. PC ARC/ INFO reference manual. Redlands, CA: ESRI: 19%.
Howard, B. V.; Young, M. F.; Ellis, J. P. Appraisal of the existing traffic accident data collection and recording system. South Australia. Report number CR6. Can- berra, Australia: Department of Transport; 1979.
Ibrahim, K. ; Silcock, D. T. The accuracy of accident data. Traffic Engineering and Control 33:492-4%; 1992.
Shinar, D.; Treat, J. R.; and McDonald, S. T. The validity of police reported accident data. Accident Analysis & Prevention 15:175-191; 1983.
Dep
artm
ent
of T
rans
port
1 Re
cord
Ty
pe
1 N
ew
acci
denl
re
cord
5
Am
ende
d ac
cide
nt
mcc
fd
.2
Polic
e Fo
rce
I%
.3
5 6
7 8
9 10
11
Acci
dent
Re
f N
o m
.4
Seve
rity
01 A
ccid
ent
fi
1 F
ool
2 S
or~
ous
3 S
lagh
I
13
I. 15
.5
Num
ber
of V
ehic
les
WI
.6
Num
ber
of
II 17
18
Casu
alty
Re
cord
s I
I1
.7
.6
I .9
Day
of W
eek
1 S
unda
y 2
Mon
day
3 T
uosd
av
4 W
udno
sdav
S
Th”
r&y
6 F
riday
7
Sat
urda
y
MI
Ml”8
Tim
e
30
3,
32
1.10
Lo
cal
Aut
honl
y I
1 I
I I A
ccid
ent
Rec
ord
A
tten
dan
t C
ircu
mst
ance
s lr
‘u
11
Loca
tion
10 d
igit
refe
renc
e N
o 1.
17
Junc
tion
Con
trol
1 A
ulho
tised
pe
rron
2 A
utom
alk
balk
si
gnal
33
34
35
36
37
30
39 4
0 41
42
lrlll
Eas
ting
43 N
ot-t
hing
12
1 st
Roa
d Cl
ass
0 1
Mc!
tom
ay
2A(M
) 3A
48
5c
6 U
nclo
ssllm
d
7 rlo
ol
8 1 Au
thon
ty
0 IJ
MI
Onl
y 44
45
10
4,
13
1 si
Roa
d Nu
mbe
r I
I II
14
Carr
iage
way
Ty
pe
or
Mar
klng
s 1
Rou
ndab
oul
(on
circ
ular
hi
ghw
ay)
2 O
ne
way
st
reet
3 D
ual
carr
iage
way
.2
la
nes
4 D
ud
coni
agow
oy
.3
or m
ore
Inn0
6
5 S
Ingl
o ca
rria
gew
ay
ringl
o lro
ch
rood
6 S
ingl
e ca
rrio
gow
sy
.2 lon
or
(0~
) oo
ch
dim
&m
) 7
Sin
gle
carin
gaw
ay
.3 lcm
us
(tw
o w
oy
=+
JQdt
y)
6 S
ingl
e ca
rria
gow
ey
.4 o
r mom
la
nes
(Iw
o
way
cap
acity
)
9 U
nkno
wn
.15
Spee
d Li
mit
mph
.16
Junc
tion
Deta
il 0
Not
01
or w
ith,”
20
mov
es
al l
unct
ion
1 R
ound
ob’o
ul
2 M
mI-
roun
dabo
ut
3 ‘T
or
s1a
ggor
od
~U”C
tlO”
4 ‘Y
’iunc
von
6 S
lip
rood
6 C
fon~
ro”d
8
7 M
ulllp
lo
~u”
cIIo
” II
US
l”O
p”““
,”
d”V
ll “1
111
111:
1111
11
0 01
l11!
, l11
11<
,111
111
3 S
op
sign
4
GM
w
ay
sign
or
mar
king
s
5 U
”c0”
volle
d 55
1 .16
2n
d R
oad
Clas
s 1
Mot
orna
y
2A(W
3A
40
5c
6 U
ncln
srili
od
68
57
5s
50
1.19
2n
d R
oad
Num
ber
m 60 6
1 Pe
dest
rian
Cro
ssin
g Fa
cilit
ies
@II
0 N
o ao
ssin
g Ia
dliO
er
with
in
50 m
eve8
1 Z
ebra
2 Z
ebra
cr
ossi
n s
canv
dlod
by
sch
ool
cros
sing
pa
ve
3 Z
oko
cros
sing
co
ntro
lled
by o
lhor
au
thol
isod
pe
rson
4 P
dica
” 5
Oth
er
ltght
co
nvol
led
toss
ing
6 O
lhor
si
tes
conv
ollo
d by
rch
od
cros
sing
pa
trol
7
Oth
er
siw
s co
nvol
led
by o
!her
au
thod
sed
pers
on
6 C
enva
l re
tuge
- n
o oth
er
conv
ols
9 F
oolb
ndgo
or
sub
way
Ligh
l C
ondi
tions
E
DA
YLI
GH
T
1 S
troo
ur li
ghts
7
mol
ros
or m
ore
high
2 S
Voe
l lig
hts
unde
r 7
mev
es
high
3 N
o 61
rool
lig
htin
g
4 D
aylig
ht
rvoe
l lig
htin
g un
know
n
DA
RK
NE
SS
6 S
lnm
l lIg
hta
7 m
olro
~ or
rnc
fv
high
(l/
f)
0 S
VIM
JI ll
gl~t
~ un
d”f
7 “w
vu~
hi
gh
(ill)
7 N
o SIIWI llg
hlln
g
0 S
l”,~
~t
Ilohl
r un
l~l
!) i
~:I,
k,,,,
zs
“,,,,
1,1,
ll,lll
llllr,
III
IkIII
1WII
7r
Sta
ts
19 (
Rev
2I9
1)
.22
Wea
ther
I F
ine fw
ihou
r hi
ph
win
ds)
2 R
ankg
(vdl
O”i h
igh
win
ds)
3 S
now
ing
(with
out
hiih
w
inds
)
4 F
ine
wim
hi
ih
win
ds
5 F
taiii
ng
witi
hi
ih
win
ds
6 sn
awin
g w
iti
hiih
w
inds
7 F
og
(or
mis
t if
haza
rd)
6 O
#wr
9 U
”km
w”
.23
Roa
d Su
rface
C
ondi
tion
fi 1 D
w
2 W
&D
amp
3S”o
w
4 F
msv
lm
5 F
lood
(s
wim
w
awr
over
3cm
s (1
Inc
h)
bP)
.24
Spe
cial
Con
ditio
ns a
t S
ite f
j 0
Non
e 1
AuD
mal
k T
RH
IC
Sig
rul-o
ul
2 A
uDam
atic
T
raffi
c S
igna
l pM
Jal!y
de
tect
ne
3 P
erm
anen
t ro
ad
sign
ing
d&&
a or
obsc
ursd
4 R
oad
wor
ks
peae
n
5 ka
d ~
urla
ca
defe
ctiv
e
.25
$foz
gew
ay
Haza
rds
b
1 D
islo
dged
bw
hicl
a ka
d in
car
riage
way
2
Otlw
r ob
@cl
in
utig
eway
3 In
vd~m
enl
with
pr
evio
us
acci
dent
4 D
og
ill
cam
.9ge
way
5
Oth
er
anim
al
in c
anio
gew
sy
.26
Ove
rtak
ing
Man
oeuv
re
h Pa
ttern
s N
o lo
nger
re
quire
d by
Ihe
~epu
rme”
~ of
Tra
nspo
ct
5s
(19
70
7,
.27
DTp
Sp
ecia
l Pr
ojec
ts
IT]
Fig.
A
l. T
he
STA
TS
19 f
orm
. (c
ontin
ued)
Veh
icle
Rec
ord
Y
P
I 2
2.1
Rec
ordT
ype
m
1 N
ew
vehi
cle
reco
rd
u-J
5 A
men
ded
vehi
cle
mco
rd
!.2
Pol
ice
For
ce
!.3
Acc
iden
t R
e! N
o
!.4
Veh
icle
Ref
No
?S T
ype
of V
ehic
le
01
Peda
l cy
cle
03
Mot
or
Icoo
tW
12
13
14
I I
I I
15 1e
In
02 M
oped
04 M
otor
cy
cle
06 I
nval
M
Trk
yd
05 c
ombm
abon
07 O
mer
lh
me-
whe
elad
ca
r 06
Tsx
l
09 C
ar (
Iour
w
heel
ed)
10 M
inib
us/M
otor
ca
rava
n 11
PSV
12G
ocds
no
love
r 1
l/2
bns
UW
(1.
52
tonn
e8
13 G
oods
ov
er
1 l/2
to
ns
UW
(1
52 t
onne
s)
14 O
rher
m
olor
ve
hlde
15
Oth
er
“on
m&
r ve
hide
1.6
Tow
ing
and
Art
icul
atio
n tl
0 N
o to
wIa
rtkx
latio
” 1
Arli
cula
ted
vehi
cle
2 D
oubl
e/m
ultip
le
trai
ler
3 C
arav
an
4 S
ingl
e tr
aibr
5
omer
to
w
18 19
.7 M
anoe
uvre
s III
01
Rev
ersi
ng
02 P
arke
d 03
Wai
tmg
IO g
o ah
ead
bul
held
up
04 S
oppi
ng
05 S
tart
ing
06U
M”
07 T
urni
ng
leh
06 W
aitin
g to
tur
n le
tl
09 T
urni
ng
right
10
Wai
ting
M t
llrn
right
11 C
hana
ino
lane
to
leh
12 C
hsni
ii la
ne
to t
ight
13
Ove
rtak
incr
m
otiro
ve
hicl
e on
its
offs
ide
_ _
14 O
vert
akin
g sl
alio
naty
ve
hicl
e on
its
ofls
lde
15 O
vert
akin
g on
ne
arsi
de
16 G
oing
ah
ead
lek
hand
be
nd
17 G
oing
ah
ead
tight
ha
nd
bend
18
Goi
ng
ahea
d o&
r
M
?I
1.6
Voh
lcle
Mov
omon
l Co
mpa
ss
Poin
t m
F
rom
T
o 1N
2N
E
3E
4SE
,:”
&kd
md
. “:::k
or
b
m
0 P
nrko
d . (1
1 karb
22
23
.9 V
ehic
le L
ocat
ion
at l
ima
177
of A
ccid
ent
01 L
eavi
ng
tie
mai
n ro
ad
02
Ente
ring
tho
mai
n re
nd
03
On
mai
n ro
ad
04
On
min
or
rood
05
On
serv
ice
road
06
O
n la
y-by
or
har
d sh
ould
er
07
Ente
ring
lay-
by
or h
ard
shou
ldor
06 L
eavi
ng
lay.
by
or h
nrd
shou
ldor
m
0”
a
cycl
owny
10 N
ot
on c
xmin
gow
ny
.lO
Junc
tion
Loca
tlon
01
Veh
icle
al
Firs
t lm
pacl
ij
0 N
ot
al j
unch
on
(or
with
in
20
met
ros/
22
yard
s]
1 Ve
hicl
e ap
proa
chin
g ju
ncllo
n/ve
hicl
e pa
rked
al j
unct
ion
appr
oach
2 Ve
hicl
e in
mid
dle
of j
unct
ion
3 Ve
hicl
e cl
eare
d ju
ncU
on/v
ehtc
lo
park
ed
al
junc
tion
BXII
4 D
id
not
impa
ct
11 Sk
iddi
ng
and
fi O
vert
urni
ng
0 N
o sk
IddI
ng,
jack
knili
ng
or o
vert
urni
ng
1 S
kidd
ed
2 S
kidd
ed
and
over
turn
ed
3 Ja
ckkn
ifed
4 Ja
ckhn
iled
and
over
turn
ed
5 O
vert
urne
d
1.12
HII
Obl
ecf
In
Carr
Iage
way
00
Non
e
01 P
rovb
ul
a&do
nt
02 R
oad
wor
ks
’
03 P
nrko
d vo
hkle
. I
II 04
Pork
ed vo
hkle
.
unlit
05
Brk
&e
(roo
f)
06 B
rklg
o (S
k%)
07 B
olla
r&re
fuge
06 O
pan
door
ol
voh
kle
09 C
ontr
aI
Isla
nd
or r
ound
ilbou
l
10 K
orb
11 C
Mor
ob
ject
28
.13
Veh
icle
Lea
ving
n
.14
Car
riage
way
_
u 0
CM
not le
ave c
arria
gew
ay
1 Le
h ca
niag
eway
ne
arsi
de
2 Le
ft o
xria
gew
ay
near
side
an
d re
boun
ded
3 Le
h ar
riogo
way
sl
raig
hl
ahoa
d at
)un
cbon
4 Lo
ll cz
mln
gew
ay
oIls
Id
onto
ce
ntra
l
5 Le
ft c
anla
gow
ay
oHsi
de
onto
ce
nfra
l m
sorv
afio
n
rese
rvat
ion
and
rebo
unde
d
6 Le
b co
ning
ewey
ol
lsid
e cr
osse
d ce
ntra
l
rose
rvat
bn
7 Lo
ll cn
”i”go
woy
O
lfSM
o
6 Lo
h co
nhgo
way
on
skb
and
mbo
undo
d
Hit
Obj
ect
Off
Car
riage
way
K
YI
00 N
ono
01 R
oad
sign
/Tm
ffic
slgn
al
02 L
amp
post
03 T
&gr
aph
pole
lElo
ctric
ity
pole
04 T
ree
05 B
us
stop
/Bus
sh
elle
r
06 C
entr
al
cras
h ba
nier
07 N
ears
ide
or o
ffsid
e cr
ash
barr
ier
06 S
ubm
erge
d in
wal
er
(com
plet
ely)
09 E
nter
ed
ditc
h
10 O
ther
pe
rman
ent
obje
ct
31
- 15
Veh
icle
Pre
fix/S
uffix
u
Lette
r P
refix
/Suf
fix
kner
or
one
of
tie
follo
win
g
code
s -
0 M
ore
lhsn
tw
enty
ye
ars
old
(al
end
of y
ear)
1 U
nkno
wn/
cher
ishe
d nu
mbe
r/no
t ap
plic
able
2
For
elgn
ldlp
lom
atu
3 M
ilita
ry
4 T
rade
ol
abr
2.
2.’
2.’
2.’
2.:
2.:
2.:
2.:
2.:
2.; 16
FIr
sI P
qlnt
of
Impa
ct
0 M
d no
t hnp
uc
fi
1 F
ront
28
eck
3 of
fsid
e 4
Nea
rsid
e 33
34
35
I7 O
fher
Veh
icle
Hit
II I1
(V
EH
Ref
No)
33
37
.?a
16 P
art(
s) D
amag
ed
000
0 N
on
e 2B
adc
4 N
ears
ide
6 U
nder
side
1 F
mnt
3
offs
ide
5 R
oof
7 al
l lo
ur
s&s
39
19 N
o of
Axl
es
0 N
o bn
ger
requ
ired
by t
he
Dep
ar(m
ent
of T
rans
porl
40
4,
!O M
axim
um
Per
mis
sibl
e G
ross
Wei
ght
m
l.W
ic
tonn
e8
(Goo
ds
vehi
cle
only
)
11 S
ex o
f D
river
fi
1 M
ale
2 F
emal
e
3 N
ot b
acod
43
4.
!2 A
ge o
f D
river
m
(Y
ears
es
timat
ed
il ne
cess
ary)
!3 B
reat
h T
est
tl 0
Not
ap
plic
able
1
Pos
itive
2 N
egat
ive
3 N
ot
requ
este
d
4 F
aile
d to
pro
vide
5 D
river
no
t co
ntac
ted
al
time
!4
Hif
and
Run
0
olhe
r 1
‘Hit
and
run’
2
Non
-slo
p ve
hide
no
t hl
l 47
40
49
50
!5
DT
p S
peci
al P
roje
cts
[Ill] (co
nti
nu
ed)
Cas
ual
ty
Rec
ord
1 Re
cord
Ty
pe
1 N
ew
casu
alty
re
cord
5
Am
ende
d ca
sual
ty
reco
rd
2 Po
lice
Forc
e El
5 6
7 0
D
1011
3 Ac
cide
nt
Rel
No
[m
1.2
13
I‘
4 Ve
hicl
e Re
f N
o m
]
>
15 1
6 17
5 Ca
sual
ly
Ref
No
m]
6 Ca
sual
ty
Clas
s h
1 D
river
or
Rod
or
2 V
ehid
e or
pllh
on
pors
engo
f
3 P
edes
tria
n
.7
Sex
of C
asua
lty
1 M
ale
2 F
emal
e
20
21
#.8
Age
of C
asua
lly
In
(Yea
rs
estim
ated
if
nece
ssar
y)
I.9
Seve
rity
01 C
asua
lly
fj 1
Fot
n!
2 S
OriO
Ul
3 S
liiht
8.10
23 2
4
Pede
stria
n Lo
catio
n m
00
Not
po
dwlri
~n
01 I
n cn
niag
owoy
cr
orrin
g on
pud
ostr
ion
cros
sing
02
In
cani
aww
ay
cros
sing
w
ithfin
zw
zag
linos
_.
_
appr
oach
to th
e cr
ossi
ng
03 I
n ca
ning
way
cr
ossi
ng
with
in
rig-z
ag
lines
exit
the
cros
sing
04
In c
ania
g~oy
cr
ossi
ng
olto
wlw
ro
wth
ln
50 m
otro
d of p
odos
lrian
cror
ring
05 In
ulrr
iogo
woy
C
rosG
llQ o
lsow
hom
06 o
n lo
otw
oy
or v
utgu
07 O
n ro
lugo
or
con
trol
Is
tand
0,
ros
urvo
tion
06 I
n C
Bnv
o ot
cor
rioge
woy
no
t on
ruh
lgo
or
cont
rol
isla
nd
06 I
n w
ingo
woy
no
t uo
sung
10
unk
now
n -.
-
#.l 1
Pede
stria
n M
ovem
ent
0 N
ot
pede
stria
n
1 C
ross
ing
horn
di
vers
ne
arsi
de
2 C
ross
ing
born
di
vers
ne
arsi
de-m
aske
d
by p
ark
cx r
tatio
wy
vehi
cle
3 C
ross
ing
horn
dt
iven
ol
tsid
e
4 C
ross
ing
horn
dt
iver
s of
fsid
e -
mas
ked
by
park
ed
or s
tabn
afy
vehi
cle
5 In
can
iagw
ay
stat
iona
ry.
not
cxos
sing
(sta
ndin
g or
pla
ying
) 6
In c
arria
gew
ay
stat
iona
ry
_ no
t cr
ossi
ng
(sta
ndin
g or
eyi
ng).
m
arke
d by
puk
ed
or s
tatio
nary
vo
hiik
i 7
Wot
hing
al
ong
In c
aniq
owey
la
dng
trat
tic
6 W
alki
ng
alon
g In
cw
iogo
woy
ba
ck
to t
nlh
8 un
know
n
1.12
Pe
dest
rian
Dire
ctio
n C
ompa
ss
poin
t bou
nd
1N
2NE
3E
4 S
E
5s
6S
W
7w
6NW
or
0 .
Ped
eStr
lan
- sl
andm
g S
UII
I.13
Scho
ol
Pupi
l Ca
sual
ly
E
0 N
ot
D s
chw
l pu
pil
1 P
upil
on p
urno
y to
nrom
lC
hcd
2 P
upil
NO
T on
jw
moy
to
iirom
sc
Jwl
8.14
Sea
t B
ell
Usa
ge
ON
olca
rcfv
~
1 S
alw
ybet
tinus
e 2
Saf
ety
belt
fiasd
-
not
in u
se
3Sat
etyb
eltn
ot6w
4 C
hild
sa
fety
te
ltkam
esr
titte
d -
in “
‘0
5 C
hild
s&
y be
klum
ets
titm
d -
rot
in u
si
6 C
hild
sa
fety
be
llmam
esr
not
fitte
d
7 un
know
n
I.15
Car
Pas
seng
er
fl 0
Not
a
caf
pasw
nger
1
Fro
nt
mat
U
T pa
ssen
ger
2 F
low
se
at u
r pa
ssen
ger
1.16
PS
V Pa
ssen
ger
0 N
ot a
PSV
pos
ton~
r 1
Bou
dmg
2 A
llghl
ing
3 .s
talId
ing
pasr
enw
r 4
Sea
ted
pass
enge
r
31
32
xl
34
8.17
D
Tp
Spec
ial
Proj
ects
[m
]
276 K. AUSTIN
APPENDIX B
An Explanation of the Codes for the Variables Used
The definitions of the codes and the methods of coding for the variables investigated are specified by the Depart- ment of Transport (Department of Transport 1994). Some of the items such as “one-way” are self-explanatory and so are not mentioned here.
Road class--This relates to the strategic importance of the road. At junctions, the class of the joining road is also included, although if three or more different roads join at a junction, the two roads with the highest class are entered; whilst if all the roads are of the same class, then the two roads with the lowest road number are included.
Road number-This is the unique identifying code for the road and must be included for all roads that are not unclassified.
District-This refers to the local administrative area, of which there are eight within Humberside and is an addi- tional variable not coded onto the national STATS 19 form.
Speed limit-The coding of this variable relates to the permanent speed limit in miles per hour. Ones that are temporarily in force (due, for example, to road works) are not recorded.
Pedestrian crossing facilities If an accident occurs within 50 metres of a pedestrian
crossing, then the code for that type of facility should be included. Where two facilities are within 50 metres of the accident, the code should relate to the one that is nearest to the accident.
A zebra crossing contains broad alternating black and white lines painted on the road and is marked by a flashing beacon. This allows pedestrians to have prece- dence over vehicles when they step into the car- riageway .
Pelican crossings are positioned solely for pedestrians and are controlled by lights, which change to indicate the pedestrian’s right of way once a button has been pressed.
Other light controlled crossings refer to junctions with traffic control devices that include an individual light to enable pedestrians to cross safely and are stated as “signal” in this survey.
School crossing patrols relate to a person exhibiting a “stop-children” sign, and when they step into the car-
riageway, vehicles are required to stop. An authorised person is a policeman or traffic warden in uniform who has statutory power to control the traffic. Both of these are not included in the survey because they are not permanent features of the highway.
A refuge is an island, usually placed in the centre of the road which permits pedestrians to concentrate on cross- ing one stream of traffic at a time.
Junction detail For this variable, any accident occurring within 20 metres
of a junction should have the code for that type of junction included. If two junctions are within 20 metres of the accident then the code should relate to the junc- tion that is nearest to the accident.
A roundabout is stated as a one-way circulatory system around a central island with entry controlled by “Give Way” markings.
A mini-roundabout operates in the same way as a round- about but the central island is flush or only slightly raised from the carriageway and has a diameter of less than four metres.
The Tand Yjunctions are three arm intersections differen- tiated by the angle of the approaching road, indicated by the appropriate letter.
Slip roads refer to the merging and diverging lanes for grade separated junctions only.
A crossroads is a junction that contains four arms with an uninterrupted alignment.
A multiple junction contains more than four entry arms. A private drive or entrance joining the highway is only
included if one of the vehicles was using it at the time of the accident and not if the accident occurred within 20 metres of one.
The other category refers to an alleyway or layby, and this is only included if one of the vehicles was using it at the time of the accident.
Junction control-A Give-way requires vehicles to yield to the traffic on the other road, although if no vehicles are present than the driver is not compelled to stop. At junctions with a stop sign, vehicles are required to stop regardless of the traffic conditions on the other road. The authorised person category is not included in the study since it is not a permanent highway feature.
Carriageway type or markings-The roundabout cate- gory refers to both the roundabout and mini-round- about items in the junction detail section. Dual car- riageways refer to roads where a central reserve is constructed to segregate opposing flows, whereas sin- gle carriageways do not contain this.