C TRANSPORT RESEARCH LABORATORY TITLE by A T h e i m p r o v e m e nt of c cid e nt d ata q u alit in Malaysia C J Baguley and R S Radin Umnar Overseas Centre Transport Research Laboratory Crowthorne Berkshire United Kingdom 1./ .11, BAGULEY, C J and R S RADIN UMAR (1 994). The improvement of accident data quality in Malaysia. In: Proceedings of First Malaysian Road Conference, Ministry of Works Malaysia. Kuala Lumpur, 20-23 June 1994. THE IMPROVEMEN'T OF ACCIDENT DATA QUALITY IN MALAYSIA Chris J'Baguley Transport Research Laboratory,(TRL). UNITED KINGDOM' institut Kerja Raya Malaysia (IKRAM) MALAYSIA Radin Umar Radin Sohadi Universiti Pertanian Malaysia (UPM) MALAYSIA Chris Baguley graduated from the University of Wales, Aberystwyth with a BSc in Physics and was subsequently awarded an MSc in 'Physics of the Upper Atmosphere' in 1975. Following a brief period in the food. technology industry with Unilever Research, he has, for the past .18 years, been working in the road safety A ~field at TRL, where he has been .involved in a range of research projects related to driver *behaviour and safety engineering measures. He joined the Overseas Centre at TRL in 1989, initially having .UK-based responsibility .for the Centre's Microcomputer Accident Analysis Package, , and also for the final production of a major road safety engineering guide. He has worked in eight different countries and, for the past year, has been based in Malaysia to initiate a road safety 'research programme at IKRAM Ir. Radin Omar Radin Sohadi graduated from University of Sheffield in Civil and' Structure Engineering (Hons) in' 1984. In 1985 he obtained his M.Eng. degree in Transport and, Traffic Engineering (Sheffield) and served as a lecturer in the Department of Civil and Environmental Engineering UPM'. He is a member of the Institution of Engineers Malaysia and Institution of Highway and Transportation Engineers (UK). His involvement in road safety started in 1986 as a member of Research Committee National Road Safety Council .He was responsible in designing the new police accident reporting system and responsible in the computerisation accident data in Malaysia. His research has helped many authorities in diagnhosing 'accidet problems and: instrumenting saf ety, programmes. in this, country.. He. 'is now the Head. of the Infrastructural .Technology Research Programme at the Faculty of Engineering .UPM, the. Coordinator of the Accident .Reseafrch.. Unit, Faculty of Engineering 'and a~n Associ ate Prof ess''r in the Civil .and ,Environhmental Engineering, Faculty oPf ~ Engineering , Univer.siti..Pertanian Malaysia. 1 THE IMPROVEMENT OF ACCIDENT DATA QUALITY IN MALAYSIA Chris J Baguley (Institut Kerja Raya Malaysia (I-KRAM}/Transport Research Laboratory {TRL}) UNITED KINGDOM and Radin Umar Radin Sohadi (University Pertanian Malaysia {UPM}) MALAYSIA ABSTRACT One of the factors affecting the efficiency of a nation 's transport system is the safety of its road network. In order that as high a level of safety as possible is achieved an important prerequisite is that an accurate and comprehensive accident database is maintained. This should be readily accessible and easy-to- use at both the national and local levels. Only with this can safety be monitored, the specific nature and location of safety problems investigated, appropriate remedial action designed and implemented, and the effect of that action evaluated. This paper describes the steps already taken, as well as on-going research, to help improve the database of Malaysia. The long-term aim is to equip all rural and municipal highway authorities with a standard so~ftware package and relevant-database to enable them to investigate their particular safety problems. A new system was developed and pilot tested in two areas and the results from these were encouraging and, indeed, prompted the national introduction of a new report form by the Royal Malaysia Police in 1992. However, there have been problems associated with the rapid expansion to the national level (primarily related to the recording of accident -locations), and new trials by IKRAM are addressing some of these. 1. INTRODUCTION The ietails of road accidents in Malaysia have been recorded in some form by the R.~oyal Malaysia Police, as they have by police authorities all over the world, probably from the time that they were first called upon to deal with a road traffic accident. However, the recording of details in a standard format which facilitated aggregated analysis really only began in 1974 when a mainframe computer system was installed at Police Headquarters, Bukit Aman. This computer is used to store all police statistics, and the function of the road accident database has been primarily to monitor general accident trends across the country. The Police produce an annual publication, Statistical Report Road Accidents Malaysia, which is a good general source of information at the macro level. 2 However, the database was not easily accessible or useable by highway engineers chiefly because no recording of accident location had been included. In 1989 a redesigned accident coding form was introduced which at least required the recording of the nearest kilometre post for accidents occurring on state or federal roads. The database still contained only limited information as the coding sheet was restricted to a single page. Despite its small size this form tended to require considerable police time to complete owing to the need to look up codes in separate coding books. Since 1990 the Universiti Pertanian Malaysia (UPM) sponsored by the National Road Safety Council (MKJR) has been carrying out a research programme aimed at improving the accident data collection and analysis system in Malaysia (Radin & Aminuddin, 1992). It was decided that in order to facilitate wider usage of the accident data, particularly by highway authority engineers, the database needed to be available on a microcomputer system. The microcomputer accident analysis package, MAAP, produced by the Transport Research Laboratory, UK (TRL) was chosen as the most appropriate software available for both management and analysis of the data (see Hills and Elliott, 1986).. In cooperation with the Royal Malaysia Police (PDRM), a new report form was designed which was intended to be easier to complete and provide more comprehensive data for use by engineers. 2. THE PILOT PROJECT The Police accident coding form is referred to as POL27 (Pin 1/91) and an example of the first page of the new version is given in Fig. 1. It can be seen that most of the information required by the Police reporter is contained on the form itself without the need to look up codes in a separate document. The person filling in the form simply needs to encircle the relevant value for each accident parameter or, in a few cases, to fill in a box. The form also contains provision for more comprehensive location data including space for a location sketch as well as a collision diagram. The two districts of Shah Alam and Seremban were selected as trial areas where the Police agreed to c(?operate in using the new POL27 form in addition to their current version of the form. Training in the completion of the new form was given to relevant police officers and its use in these two districts began fully in 1991. The specification files used by MAAP were modified to be compatible with the new POL27 form, and data is entered onto microcomputer at the police station. Assistance by a UPM staff member in the checking and coding of the data was given on a regular basis to the police stations involved in the pilot project, and this assistance is still continuing. 2.1 Location systems For the improvement of safety by engineering measures, one of the most important items of information recorded about an accident is its precise location because analyses need to be chiefly focused on accident histories at individual sites. These are examined in detail to find common patterns which, by some 4. NO. LAPORAN 3. TAHUN4 ~ L 6. BULAN 1.0111(01 -12, 7. TARIKH 2 .(0 l.-3 1 S.MASA 2 KEJADLAN (0 2359) '~(Silt isVbulatkih 19. SENIS PERMU.KAAN I. Bass 2. Blerurop Boata 63 Bertioap Bitumin/Tar 4. Bersurap Konkuit S. Tanah 20. SI[STEM LALULINTAS I. 11:11 Se alal 0 ~i111 Duo Hulh 3. ~ ~ i Tigo 91 1 Lorong 4. j1J1ifl atau Lebi ayh 9. HAR!I. Ahad2. Isnin (D selasa4. Rabu5. Kharnis6. Jumoos7. Sabtu NO. SIR!: P0L27 (Pin. 10 1) SALINAN ASAL --- i s l s e a w l B x k s m PEGA WA1 YANG MENGISI BORANG TITANGANNO/PKT _ _ _ _ _ _ _ _ _ _ NAMA DISIEMAK OLEII1 .T/TANGAN _ NO/I 5KT ___ --- _ NAMA _______________ I (Sila rujuk buku kod baloi) 110. BILANGAN KIENDERAA.N TERLIBAT I p 1 1 !. BILANGAN KENDERAAN ROSAK [- ;L 12. BILkNGAN P1EMANDU MAT!I ~~ .13. BILANGAN PEMANDU CEDERA IJ 14. BILANGAN PENUMPANG MAT! is. BILAGAN PNtJMPAN4G CEDERA [ 91122 16. BILANGAN PEJALAN KAKI MAT]! 101 17. BILAN4GA.N PEJALAN KAKI CEDERA FeT 1I U - S  ______________________________________ 21. BENTIJK JALAN 1. Luuso2. Slelkoh3. Bulatan4. Simpang Empalljebih (3 Simpang T/Y 6. Simpang -1'-r- 7. Simpang Bertingkcal 22. KUALITI PERMIJ1KAAN Rata 3. Berlubang 2. Menadp 4. Bcrombak 23. KEADAA.N JALAN 63 ~ t 2. Cuaom 24. MNIS1 GARIS I. Gadis Kembarn O adis Tengals Bertndo 3. Salon Schala 4. Pernisal Jalan S. .Pusingan U 6. Tengah Tak Bertanda 25. LANGGAR LARI I. Ya n Tidak 26. "NIS KAWALAN I. Polis2. Agensi Luar 3. Berloanpu Isyarat 4. Lintasan Salan Kaki S. Lintnaan JSalon K~ki Rerarlmpoo. Lintosan Keretapi 7. Garis Lintang Runin 8 8. Kotak Kuning ® Tiada Kawajan 27. LEBAR JALAN F0 -] Meter 28. LEBAR BAHU JALAN M02. 02. (Sila isikan keduo-dua beah lebar bohs jalan dalam unit met~er) 29. JENIS BAHU JALAN I. Bertuarp63 Tidak Berturap 30. SEBAB-SEBAB UTAMA KIECACATAN JALAN I. Bahu Salan RendalslTiniggi 2. Manhole Rendoah /Tinggi 3. Batu Loniggar 4. Salan Berdebui S. Salan Berlubang 6. Jalan Licin 7. Kerosakan.Lampu lsyasos 8. Lintasan Kcreaopi Sempit 9. Jambatan Sempit 10. Tiad~a Guard Rail II1. Tiada/Kuarng Lampu JalanQTiada Berkenaant 31. HAD LMU 63 SO KM. 4. 90 K. 2. 70 KM.1 5. I I0 KM.J 3. 80 KM.1 6. Lain-~lain 32. KEADAAN,PERM1UKAAN JALAN 63DKering 4. Berminyak 2. Basnjir 3. Berpasir 3. Basah 6. ScdangDiperbaiki I8.SENIS KEMALANGAN 1. Maia 2. Pwaha 63 Ringan 4. Rosak Sahaja 133. JEN4IS PERLANGGARAN YANG PERTAMA I. Depan Dengan Depan 2. Langgar Belakang 63Langgar Rusuk Tcpal 4. Langgar Sebclah Tepi S. Ilcrgesel 6. Terhimpit7. Langgar Binoaung 8. Langgar Objek Di Jalan 9. Langgar Objelc Di Luar jalan IC. Langgar Pejalan Kaki It. Terbolik 12. Terbobas 13. Cermin Pecalh Soaaja C. SUASANA SEKCITAR (BulaLkian nombof beakcnaan) 34. CUACA 0 Balk 4. Bcrkabus 2. Angin Kuat 5. Hujan atau Lintanig 35. CAH-AYA I. Siang 3. Gelapl LampsI~ 2. SubuW C)Ge1.1 Tnp Scuija Lanspu D. LOKSI(Sil~aiA~uatkanno bo~ienaan ~A ~ NAMA JALAN/SIMPANG 39. JENIS KAWASAN Ls6. JENIS JALAN 37. NO. LALUAN (Isikan) ) TAMIP N I. Perahan I. ]alan Raya Eksprcs E b) - 2. Pejabai DPersekutuan Ly~~~' Jk ~pn)3. Ka-asoa~n BeliBelah 3. Negeri 11138. JENIS TEMPAT , 4. Pembiaai~rsdu.stri 4. Bonarda~n [7 F .Bandaay. 3. Pekan 5. Janbtasiljansa 3. Lain-lain [IJL~i]2. Bondlar Luar Bandr 6. Sckolah -Lain-li POS KILOMETER TIERDEEKAT:Jorvak Dsii F S-E iRE mc A N Jarak Dan 1 ~TAM~ r4 (N iam Tempnsllandor)- No. Scksycn K( a .T nlal3 dr [~~ 76lKm m JARAK KEJADIAN adolah ~~~ meter daui Pos Kilorneter ke arah 5 -9 M B3A H iJA:01 ~~~~~~~~~~~~~~~(hama tcmp.uLlbanda) Jika TIDAKC TERDAPAT POS KM atau NO. SEKSYEN. silo isikan bahagiw nim. Jarak kcjadin adalah -- km dari (noons tcrnpaslbusdax) ke rams (namna icnipaubndar) Fig. 1 Example page from POL27 accident report form 3 .4 appropriate remedial treatment to a site, the engineer may be able to prevent in the -future. Unfortunately, adequate attention to detail in the recording of accident location is often neglected. MAAP users are encouraged to use more than one system of location coding, and to record normally two types for each accident (one to serve as a cross- check on the other). For accidents in cities or towns, ie. Shah Alam and Seremban, two separate location items are recorded for each accident, namely: - i) map grid coordinates (X, Y or eastings and northings of the national grid system); and 3020 3017 30159 30 13 3012 6 if~if 43 278 / .79 38 ) l 3 4382 i 2 -118I 4: . . Fig.2 Node and grid coordinate system in centre of Seremban 5 ii) a Node or Link number. For this system it is necessary to assign unique numbers (Nodes) to all main intersections on a town map .'The Links are sections of road between these junctions and are simply referenced by the two nearest Node numbers on each side of the accident location. The maps used for coding accident location had both the national grid coordinates and the Node-Link system superimposed; an example of part of such a map for Seremban is shown in Fig. 2. For the rural routes where intersections are much more widely spaced, it has been decided that accident location should be recorded to the nearest lO0m from a 'kilometre post; or rather, the post's unique Section Number from the beginning of the road. For accurate recording by the Police this will mean that kilometre posts do need to exist along each highway. Although installation of kilometre posts can be a relatively costly exercise they are invaluable for highway maintenance purposes as well as accident recording, and are also beneficial to motorists in providing distance information of their destinations. or their position in the event of a breakdown. Thus on rural roads the two location items recorded on computer are:- i) map grid coordinates as above, and ii) estimate of nearest lO0m from previous Section number post. Fig. 3 Accident clusters in Shah Alam using grid coordinates 6 2.2 Site prioritising The above location data are used by the computer package to inform the user of sites with the worst accident histories. If accurate digitised or scanned maps of an area are available, the software can use the X-Y coordinate information of the accident database. to produce accident maps. These simply provide a quick visual indication of where accidents are clustered, as shown in Fig. 3. For rural roads, histograms of accident frequencies can be produced along a complete or part of a particular route. The intervals of each bar can be specified by the user and the program can list the worst kilometre lengths in terms of accident numbers (of types the user may specify), or even the worst lO0m sections if required (as shown in the example in Fig. 4). KILOMETRES accidents 0 .. 9 7 xxxxxxx 1 -1.9 13 *xxxxxxxxxxxx 2 -2.9 106 *xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 3 3.9 128 ****xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx XxXxxxxxxxx~xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 4 49 54*xxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 5 .5.9 24 *xxxxxxxxxxxxxxxxxxxxxxx 6 6.9 28 *xxxxxxxxxxxxxxxxxxxxxxxxxxx 7 -7.9 18 **XxxxXXXXXXXXXXXX 8 8.9 24 *xxxxxxxxxxxxxxxxxxxxxxx 9 -9.9 31 **xxxxxxxxxxxxxxxxxxxxxxxxxxxxx 10 -10.9 20 xxxxxxxxxxxxxxxxxxxx 11 -11.9 22 *xxxxxxxxxxxxxxxxxxxxx 12 -12.9 15 *XxXXXXXXXXXXXX 13 -13 2 xx …-- ------------------------------------- Tlotal= 492 …-- -------------------------- ---------- * Fatal x Injury or Damage Listing of worst loom sections Kilometre Post 3.0 51 accidents Kilometre Post 2.1 50 accidents Kilometre Post 2.0 46 accidents Kilometre Post 3.1 41 accidents Kilometre Post 4.8 23 accidents Kilometre Post 11.2 19 accidents Kilometre Post 3.2 17 occldcnts Kilometre Post 8.0 17 accidents Kilometre Post 9.6 16 accidents Kilometre Post 6.4 14 accidents Fig. 4 Histogram of accidents along a Federal road in Seremban 7 In urban areas, MAAP can list the worst nodes or links in an area and can plot histograms along a route specified by node number, similar to the above kilometre analysis. These priority listings are simply by frequency of accidents but they can be restricted to different accident types (eg. injury accidents only, pedestrians only, etc). From the listed accident severity details it is possible to use an accident points weighting system such as that adopted by the Highway Planning Unit of Malaysia (HPU), ie. accidents involving fatality=6, serious injury=3, slight injury=0.8 and damage-only=O.2, to give an initial priority listing: the top five sites in Seremban are given as an example in Table 1. This Table also shows an alternative method of obtaining the priority listing by ranking by the total cost of accidents (ESCAP, 1985) at the locations. In this case the three ways (accident totals, points and costs) all give the same ranking but often they yield different orders of priority for the sites. Table 1 Priority Listing of worst nodes in Seremban Node Junction Name Accidents 1991 & 92 Accident No. Jalan: Fatal Seri- Minor Dam- Total Points Cost ous age Aces. RM 225 Yam Tuan/Sh. Ahmad 2 - 1 1 55 67 31.8 547,500 214 D.B. Tunggal/Sh. Ahmad 1 3 8 48 60 31.0 387,000 215 D.B. TunggalfLee F. Yee - 4 12 27 43 27.0 280,500 313 Zaaba/Dato Ling - 4 9 22 35 23.6 228,000 224 Yam Tuan/Berhala 1 1 3 14 19 20.2 231,000 It should, of course, be noted that after investigation of the accident types at each junction and site visits/studies the above priority listing for actual remedial work is likely to change, dependent upon the estimated cost effectiveness of appropriate treatments and available budget. 2.3 Demonstration study As a demonstration application of the pilot project database, some in-depth studies of blackspots have been carried out (Baguley, 1992; Radin, 1993). Taking the worst site in Seremban as an example (see Table 1), this is an uncontrolled T-junction, Jalan Sheikh Ahmad with the major road F0001 (Jalan Yam Tuan) located within the main town area. Both roads are one-way only and permit two lanes of traffic. The accident data for this junction were extracted from the available data of 1991 and 1992. Using the 'stick diagram' option of MAAP (Fig. 5) and producing a separate collision diagram, as in Fig.6, it was found that the majority of accidents were side swipes or side impacts that occurred chiefly on the main road with vehicles merging from the side road. A total of six side impact accidents in two years resulted in injury (including one fatality) and all involved motorcyclists. The other main type of 8 Univerafti Pertanlan Mlateyele STICK DIAGRAMI ANALYSIS ACCIDENT RECORD FILES: NOW225 CONDITIONS SET: Mode 1 Node 2 11 2 3 16 17 8 j 9 I 101 111 121 131 11.1 151 161 171 181 806 109 109 I09 112 108 108 108 110 110 110 101 ill ill ill ill [ll 103 103 I TRI 107 119 13 102 101 110 116 113 120 123 103 10 l 11 5 129 130 164 109 I I 7I 1 2 1 2 15 1 7 16 1 1 11 1 4 51 51 1 16 16 171 217 1 w06 116 110 112 120 IIs 109 113 l09 [is 109 109 121 II? 101 120 1ll 113 112 I PERN 0 I 1 0 0 D 0 10 D D 0 0 0 j ' 60 IRK IRK IRK JOK 1 IR 1K IRK I K 6K 6K IR K IRK IRK 16KIR K IRK IR K IRK JK Awl II I 111 11 1o1 1 1 111 1 011 1 1 1 T loll 1 1 1 1 I1 1 1 1 1 1 00110S 1 100100S 1 Iss I 1 1 0 0 1 0s 1 1 100 1 1 1 1 1 I 191 201 211 221 231 241 251 261 271 281 291 301 311 321 33 341 351 361 BIN4 103 I04104 104" 104 106 106 l05 105 l05 107 107 101 101 102 102 102 102 1 160 127 10S lOS 112 130 103 109 10 110 130 103 [ll 105 114 103 106 110 112 RI 1 6 16 16 13 12 11 16 161 I 1 51 11j3 1? 1 512 'I4 140A fls 1.20 110 121 11 112 116 113 116 113 120 120 113 116 101 110 115 11 I PE.)l I PEJl II PEal liii 1 1 1 I1 1 PRH I o I 0 0I 0 01 ) 1 0 0 1 0 I 1 0 0 I1 ~1 OK IlK 16K IRK 160 lo IRK IRK IRK I6K IRK [OK IRK IRK IRK 16K IRK IRK INK I 00 1 I l l o s 1s s 1 0 01 1 1 0 0 1 1 001 00 1 1 1 1 0 0 1 1 0ss I [s s I I I [s Is I Iss IS I I 1 0401 I I 6114. BULAII TRH. TARIKH 0R1. MARI lISA. IRASA PE.. PEI.UKAK PRtO. KEPARAWAJI EK * BAIK I'- SEADON ,~.. REAR EN0 '1. SIOE AMI. AMIMAL 01 * OVERTURN SS * OIDE0IJIPE 000. 0UBLUOSE6JlA ORK. 0ARK ........................................StIck Wu~.vr * Accfdent Code 6ub~er:* 1 .021139 2 .022004 3 .023430 4 .0303137 5 .019100 6 .019730 7 -020055 0 .025006 9 .025632 10.025898 11.000253 12.027504 13-027706 14.028153 15.029890 16.030053 17.005640 18.006068 19.001627 20.006423 21.000420 22.009096 03.010276 24.017604 25.0133&2 26.010386 27.010791 28.012139 29.015955 30.016991 31.000659 32-001813 33.004612 34.001113 30.003466 36.005626 17.001209 38r008200 39.008617 40.008741 I 37I 3 1 391 401 411 421 431 4 1 '5I 461 471 481 491 50l 31 1 5 21 53 1 5 '1 ELM 102 l 0 1 03 103 103 103 l 04 1 04 104 l5 105 106 106 1 06 107 1 08 l~s 106 1 1614 129 110 114 110 119 126 101 106 120 102 105 114 127 129 124 10 123 130 HRI0117 13 171 11 51 5 14 12 12 17 16 1 112 I4 1 6 14 1 1111 MISA 112 108 100 115 121 118 117 116 113 110 107 I116 110 117 112 111 ill 114 I PRH D 1 D 0 0 0 j 10 0 0 0 1 I 1 1 0 D 0 0 1 I EX IR 16K IRK IRK IRK JOK IRK IRK 16K 16K IRK IRK IRK I6K IRK ISO IOK IRK I 00T 1 111JO JO II1 111 1 1 1 1 1 1 1 ss l I Iss I1 1 10ss10s 1 ISS s s I S s 1 olsol o 1 0 10 0 1 1 0 8 0 1 1 1 1 1 1 0 0 0 1 1 1 1 1 q1 1 0 61 060 71 0I 5 1 39 1 60 1 6 11 162 1 03 1 6 2251 461 22 67 lI 68 69 70 71 x 4061 21 31 1 31 3 1j 4 1l 3 1 3 1 6 1 08 6I l' 2 1 141 720I ' 122 6 I IP NEAl12 13 13 14 ll 1 1 4 1 I POJ10 il 1 21 7 P E2l1 1 1 1 Pj R5.1 1 1 1101 101 1-1 1 101011 1-1 1 1 1 KlM ee 0 K1 D K 160 1 0 1 0K 1 D 0KO KI D 16K D K K 1 D K1 D KI F I I I OK I K IR IR IR IR IR JOK IR IR IO IR IR IR IO I I I ss1 001 1 00s11 I 0 1 1I Is [ssI ~1 1 1 1 1 1 O LD I 1 1 1 1 o 1- 1 1 1 1 1 1 41.009150 42.0096744 3.0102654 4-0104524 5s011281' 460123724 47012970~ -1~~165304 49.0178715 0.0181135 1.021177152.0233593 530265055 540277125 55-028721 6.0287910 57.033104 58.0331570 9.03422960.0142326 61.035419 6 2037828 6 3037829 6.40372196 -225-000 Fig. 5 Stick diagram for Node 225, Seremban 9 Fig., 6 Collision diagram for Node 225, Seremban serious accidents, comprising one fatal and four injury, that occurred at or near this junction involved pedestrians. In addition rear-end collisions were common with 12 accidents occurring in the two-year period. 2.3.1 On-site study results To supplement the accident data, a one-day site study was made in which vehicle manoeuvre and pedestrian flow counts, traffic-conflicts (or near-misses - see Baguley, 1984; TRRL, 1987), and vehicle approach speeds were monitored. The pedestrian and vehicle counts are given in Fig. 7 and it can be seen that there are a relatively large number of pedestrians crossing the busy Jalan Yam Tuan, reaching a maximum of 890 during the lunchtime period. In comparison, the number of motorcyclists passing, through the junction reached a peak of over 1 100 during the late afternoon period (1600hrs), and motorcycles comprised about 32 percent of all motor traffic. The traffic conflicts recorded over the six-hour observation period are given in Table 2. These tend to confirm the accident history for the junction. The most frequent type of conflict is the merge, where vehicles from Jalan Sheikh Ahmad tend not to give way to vehicles on the major road. Many drivers were also observed to force their way as quickly as possible over to the left-hand side in order to turn left into Lorong Khalsa or an adjoining petrol station. Approach speeds were HourCa McPck LorryBus 121P Hour CrMcPkuoryBus 13- 1f'I 1 33I1 09-1 47329 69 741 60i 16-I 16161 4 10-11 490 370 69 53 53 1- 57 J22I 3 2I 01 12-1l 400 322 51 56 55 Total 410 155 47 8 5 13-1A 529 2971 53 58 6 6-hour all vehs .total = 625 16- 17 3 4 5 17-i 37 2661 40 5 58 ~~ j Jatan L~~m Total 270 1 1 96 3 3 4 Kee Siong ~~~~6-hour all vehs. total = 5703 I..~~~~~~~~~M.. e .* c~~~~~~a ,, ::0 Hour M/c Others 1 300 [58 90 16001 [8f 90 1700Total 173 235 All vehs. = 408 Jalan Khalsa .. ..- J. Hour Car M/c Picku Lorr Bus 09- I 253 I88 I311 2I1 25 5 10-I I 3081 1361 44 131 12 13-1 I4541 41 5 ~2 16 16- 1 41 17-I 418 66 5 31 1 Total 2333 1144 309 151 7 6-hour all vehs. total = 4015 Ialan Sheikh Ah mad Hour M/c Others Hour M~c Others 1300 61 1300 12 22 Hour Car M/c Picku Lorr Bus 1600 281 57 1 6100 141 18 )9-1 17 480 58 124 76 1700 3T 7- 1700 r -- - 10-il 534 254 75 40 52 Toale 75 868 Totla 36 67 ~2- 1 4831 267 L 71 1 21fi54 Al vehs. = 263 Al vehs. 103 3-1 4503I237I67I7I1 31 775 17-i 3 23 7 [52 -1-4 50 Total 2690 1726 396 156 366 6-hour all vehs. total = 5334 PEDESTRIANS CROSSING IN AREA:- Hour A B .c D 09-1 12 4 1 151 10-1 17 193 13 199 12- 1 33j 259 99 62~ 13-141 ] 329 9 2761 16- 1 7¶~j 169 72 iw 17-1 89 76 4 Total 899 1186 533 1120 6-hour all peds. total = 3738 Fig. 7 Pedestrian and vehicle flow counts at Node 225, Seremban found not to be excessive with a mean speed of 30km/h and 85th percentile speed of 37km/h, which may explain why the large majority of conflicts were classified as 'slight'. Table 2. Traffic Conflicts at Node 225, Seremban Number of Slight and ( Serious ) conflicts at junction of J. Yam Tuan/J. Sheikht Ahmad in hour becinninp, Pedestrian conflicts were also relatively common and pedestrians were observed to have obvious difficulties in finding suitable gaps in the traffic stream. There are currently no pedestrian crossing facilities provided. 10 Hour M/c Otheri 1 300 19 31 16(00 l 61 5 17001 , 3 Total 90 116 All vehs. = 206 Conflict 09:00 10:00 12:00 13:00 116:001 17:00 our Manoeuvre ItItal- type_ _ I_ __ _ II__ Merging~ 34 (1) 29 (2) 38(3) 39(1) 37 38 (1) 215 (8) Pedestrian- 4 0 1 6(1) 8(1) 2 5 22(2) Rear-end- 6 2 7 8 4 6 33 Crossing-8' 4 5 3 1 1 2 16 .... 1.1 11 --- ... .. ... 11.1 j ..1 1.... 1...... --------- ......... .. ........... ... .... 1 1 2.3.2 Proposed countermeasures In view of the very high conflict rate involving merging vehicles, and the need to provide an optimum low-cost solution, the recommendation was to install a solid,' low delineator as shown in Fig. 8. The objective of this would be to narrow the merging stream down to a single lane and phase in the merge more gradually, and also to prevent the staggered crossing manoeuvre from Jalan Sheikh Ahmad to Lorong Khalsa or the petrol station: traffic is able to turn left at a major junction slightly further downstream. The raised pavement extension in the mouth of Jalan Sheikh Ahmad would also serve as an additional traffic calming device by providing a chicane and visual narrowing which should help to reduce the higher vehicle speeds. Fig. 8 Proposed countermeasures of channelisation and pedestrian refuge at Node 225, Seremban The refuge shown in Fig. 8 is intended to reduce vehicle-pedestrian conflicts as drivers attention should be focused at this chicane point. It is also assumed that pedestrians would be attracted to this easier place to cross, as it means they would now only have to make gap judgements about one traffic lane at a time. It was recommended that the hatched area be bounded by large road studs which should serve as an additional warning to the 'road narrows' signs to help reduce th~e likelihood of collisions with the refuge. It is hoped that this method will improve the safety at this junction with minimal disruption to the traffic flow (unlike traffic signals). It has been learnt subsequently that the Municipal Authority is likely to adopt much of this suggestion but will be constructing a pedestrian bridge in favour of the refuge. 1 2 3. NATIONAL IMPLEMENTATION Several minor changes were made to the Police coding form following the experience gained in the pilot areas, and the revised version was introduced nationwide in January 1992 following an extensive training programme for traffi c police officers. As the process of forms being sent to Police Headquarters, Bukit Aman, in Kuala Lumpur for entry onto mainframe computer (together with other crime data) is well established, this system was retained with the new POL27 form. On a monthly basis the data is now downloaded to microcomputer and the file converted into MAAP format (see Baguley, 1992), at which stage it is made generally available and can be interrogated easily using MAAP facilities. Existing investigation papers, including separate 'aftermath' sketches for each accident, continue to be necessary for prosecution purposes and, owing to the considerable additional time required to fill in each new P0OL27 form, it was agreed that the Public Works Departnlent (JKR) provide assistance on location coding. The Police still record as much information as possible about the accident site and the local JKR District offices are sent copies regularly of the two relevant sheets of each accident form. They are required to check location details of each accident and complete the coding of route number, closest kilometre (or section) post number, and nearest lO0m from the relevant post. The completed parts of the forms are then sent to the Highway Planning Unit (HPU) of the Ministry of Works, where the location information is further checked and entered onto computer. This will eventually be merged with the corresponding accident records received from the Police. The process is shown diagrammatically in Fig. 9 and, in theory, the database is now complete for analysis by all interested parties. The number, of recorded accidents in Malaysia is increasing sharply and, as all accidents are required to be reported to the Police, they are currently faced with recording the details of more than 120,000 per year (for 1992). Most of these, of course, involve vehicle/property damage only and a compromise was therefore reached when the new forms were introduced such that the very minor cases, classed as 'no further action' (NFA), would no longer be recdrded on computer and thus would not require a POL27 form. For 1992 this has resulted in a total of about 67,000 accidents on computer. Direct comparisons with previous years' data, however, should at least still be valid for those accidents involving injury. 4. THE NEED FOR FURTHER IM4PROVEMENT When the new form was introduced nationally there were a number of initial problems associated chiefly with a lack of clarity about the new responsibilities, and the fact that these extra tasks had to be absorbed into present work schedules without additional manpower being allocated. Early returns of forms containing location data to HPU were rather poor. Only about 81 per cent of the forms which should have been completed for Federal and Expressway roads Distribution of POL27 ooples:- 1L 2nd (osrfl nony Fig. 9 Production of accident database in Malaysia (20,887 records) were in fact received at HPU. Of these, 34 per cent contained either -incorrect or missing data with respect to accident location. However, by holding a series of workshops-in each State of Malaysia for the Police and local JK staff to determine the problems, followed up by training sessions, HPU is trying hard to improve this situation. One common problem was that descriptions of accident locations on the POL27 form were not always adequate for the local JK staff to be able to recognise the site, eg. the location sketch simply contained the road name without any other reference point. This could possibly be due to a lack of awareness of its importance by the particular reporting officer, or that he was not actually required to attend the scene (as in most damage-only accidents where witness statements are acceptable, but are rarely precise about location). Another common reason given was that the nearest kilometre post to the accident is frequently missing or simply not installed (as on many State roads). Indeed, for the State roads (12,399 accidents) the majority of accident records do not yet contain location information. It should also be noted that, owing to the fact that JKR have responsibility for the Federal and State roads only, there is currently no location coding being Or Ini nal 1 3 2nd (nart) oopy 1 4 entered on the database for roads under the jurisdiction of a Municipal authority, ie. most roads within a town or city. 4.1 Accuracy of recorded location data As an example check on the location data which is now on computer, over 2,000 accidents along Federal Route 1 were searched for the recorded values of the nearest lO00m from the relevant kilometre post, and the frequencies of these are shown in Fig. 10. 800- a600-- ~5 500-- U) 200- z Nearest 1 Q0m from Km Post Fig. 10 Accident frequency on Federal Route 1 by 1/10ths km from Posts It can be seen that, rather than a fairly even distribution -as might be expected, there appears to be a strong bias towards accidents being recorded at the kilometre post, and perhaps also a rounding to 0.5km. The higher numbers at 0.1, 0.2 and 0.3km may also indicate the inclusion of a proportion actually occurring at 0.9, 0.8 and 0.7km respectively. The latter arises from a common confusion of location being estimated to the nearest lO0m from the closest post and not then corrected to the lower Section number of the two posts up and downstream of the accident. Surprisingly, the distribution of lO0m accident location coding for damage-only accident types is very similar to the separate, distributions for fatal, serious and slight injury accidents even though, in the latter cases, the Police will most probably have attended the scene. There is currently no coordinate system used to record accident location as in the pilot project, and thus use cannot be made of the geographical mapping facilities available in the latest version of MAAP. 5. NEW TRIALS In terms of accident location the pilot experiment was very successful but this may have been largely due to the fact that it was a controlled experiment with trained full-time staff who supplied all the mapping needs. In an effort to improve the national situation, therefore, IKRAM have initiated some trials in the local JK District area. 1 5 The first of these is to demonstrate the production and use of base maps in as simple a manner as possible by adapting existing appropriate topographical maps in such a way as to aid the process of accident location coding. Another proposed trial is to utilise recent advances in technology in the field of position logging, using hand-held Global Positioning (GIPS) instruments at the scene of each accident. This should'greatly assist and simplify the process of recording grid coordinates which are, themselves, of increasing importance in view of the recent trend towards the use of Geographical Information Systems. Another application of advanced technology will shortly be attempted by UPM in the storage of accident collision sketches on a national scale. by the use of CD ROMs or optical disks. These are discussed in more detail below. 5.1 Detailed maps Topographical 1:25,000 maps for the area have been acquired from the Department of Census and Mapping which were the most up-to-date available and of optimum practical size. With the use of a calibrated measuring wheel attached to a car, all State and Federal roads in the Kajang area have been slowly driven along and the position of kilometre posts and other landmarks in relation to reference points which are already marked on the map (eg. road junctions) have been recorded as accurately as possible. On average, a feature or fairly permanent landmark was noted at about ¼/ km intervals. This information was transferred to the maps and an example of part of such a map (of which there are 7 for the KaJang area) is given in Fig. 11. Road number ~~~SK Lubuk kftub; 19 bend 2 Jd 351O tk0,>1 85 Ctvr P0 49 -, ~i bnd -Section number (kilometrepot 427000 428000 429000 430000 431000 432~00 Fig. 11Example of part of landmark 1:25,000 map for KaJang area Fig. 1 1 1 6 Where not already shown, one kilometre squares were accurately drawn on the maps corresponding to the position of the national grid coordinates. With the additional landmark information and Police descriptions on the POL27 form, it is hoped that the JKR engineer or technician. will be able to pinpoint accidents much more easily on the appropriate map, and will then be able to check the position of the accident in relation to the nearest kilometre post. The JKR officer will also note the X and Y coordinates for each accident location to the nearest l1in. The use of this method will be monitored and, if successful, it is hoped that the technique will be adopted throughout Malaysia. For towns it would also be desirable for a node system to be assigned as in the pilot project. This, however, necessitates the use of even larger scale maps (eg. 1:5,000), and unfortunately, up-to-date versions of such maps are unlikely to be available for some considerable time. 5.2 Hand-held GPS units A growing number of leading electronics companies are now manufacturing portable Global Positioning System (GPS) units which make use of the American satellite system to receive radio signals from which they are able to calculate present location on earth and display coordinates. The use of these instruments by police attending the scene of accidents to perform the task of recording precise location automatically is obviously a very attractive one, as long as they are accurate, easy to use, and the unit cost is not prohibitive. Tests were carried out on five different units (currently costing between RM3,500 and RMl11,000) at a known GPS survey station near Rawang. Unfortunately, there are several factors that affect the accuracy of the reading, for example, satellite clock errors, ionospheric and tropospheric delay, receiver noise, and even purposeful random degradations of position known as Selective Availability imposed by the Department of Defense, USA. Owing to these factors it was found that the instruments used in the field often displayed readings in excess of 50m in error in one or more of the coordinates and occasionally the error could be as much as 200m. However, it is possible to improve on the accuracy by the use of a differential GPS system (claimed to be within 5-l1in). This is a second GPS ground station at a precisely known location which will receive the same satellite signals as the mobile unit and will therefore be able to calculate the instantaneous errors. The errors can ideally be transmitted immediately via VHF radio to the mobile unit which should then be able to automatically apply the same correction at that instant to its computed position. This system will, of course, substantially increase the cost of providing the Police with adequate coverage and it is therefore important that such equipment is thoroughly tested for practicality and reliability before a decision is taken for wide-scale application. IKRAM will thus shortly be purchasing a system for trials within the KaJang police area. It is proposed that for a period of about six months a team will be on standby to be called out to the scene of accidents, and will themselves record location and other details using a second hand-held GPS 1 7 unit and scale map. 5.3 Collision sketch retrieval Space for a collision sketch is provided as part of the POL27 (pin 1/91) form and sketches of this type have proved invaluable for investigators looking for common patterns of accident, or simply to clarify details of a particular accident. Cataloguing these for the whole nation is currently being carried out by the Accident Research Unit, UPM but this has been found to be both time and space consuming. A project has recently been initiated to develop a system of scanning the sketch page of the POL27 form to be stored on either CD ROM or optical disk. Flowdver, greater storage efficiency is required since a single scanned diagram page consumes about 1Mbyte of memory space. Some kind of compression was thus required and an early trial using special software and a cofax compression card offers considerable improvement by reducing a 1Mbyte sketch diagram in rasta format to about 200Kbytes. Assuming that about 5,000 accidents per month are recorded in Malaysia, this would require the use of a 1Gbyte optical disk (a typical size of optical disk) for storage each month. It is envisaged that when an engineer needs to view the collision sketches of a group of accidents this will eventually be very quick and easy to do via computer, by simple input of the relevant accident reference codes and access to the set of optical disks. 6. SUMMARY AND CONCLUSIONS Although accident data has been recorded on computer since 1974, it has not been readily accessible to engineers and did not contain adequate information to indicate exactly where problem areas exist and what could be done in terms of engineering measures to alleviate the problems. In 1991 a major advance was made in the design of a new accident report form which was both simpler to fill in (principally ticking coded boxes) and more comprehensive with regard to information of direct relevance to engineers. The form was linked to the MAAP microcomputer system and was fully tested in two pilot areas in Malaysia. Problem sites were identified in the pilot areas and some in-depth studies carried out with countermeasure recommendations in order to demonstrate the use of the database in improving blackspots. Following the success of the pilot project the system was introduced nationwide from the beginning of 1992. There were a number of initial problems associated with this and, even after the establishment and clarification of the various responsibilities for the data, the database still remains deficient, notably on the aspect of accident location (of vital importance to the highway engineer). Trials of the introduction of large-scale 'landmark' maps and GPS systems are being carried out. It is recommended that such maps are produced as early as 1 8 possible across the whole country, and if the GPS trial proves feasible, funding should be allocated. to supply this labour-saving equipment to the Police. Without an easy-to-use and comprehensive accident database and analysis system in place, engineers in local highway authorities are much less able to play an efficient role in the regular investigation and treatment of safety problems in their respective areas. This is desperately needed on a wide scale in order that they can contribute effectively to the nation's very necessary efforts to improve road safety. 7. ACKNOWLEDGEMENTS The authors would like to acknowledge the assistance and support of the National Road Safety Council, Ministry of Science, Technology & Environment, and the Royal Malaysia Police (particularly in the pilot areas of Shah Alarn and Seremban). The authors are also grateful to the District JKR. of Kajang and to all research assistants at UPM and IKRAM for their support throughout the duration of the project. 8. REFERENCES [1] Baguley C J (1984). The 'British Traffic Conflict Technique. In: International Calibration Study of Traffic Conflict Techniques. NATO ASI Series F: Computer & System Sciences. Springer-Verlag, Heidelberg. [2] Baguley C J (1992). TRI, Support for MAAP Pilot Project in Malaysia, Technical Paper OU/TP/188/92, MKJR/TRL, Crowthorne, United Kingdom. [3] ESCAP (1985). Improving Traffic Accident Recording and Analysis System in Malaysia. United Nations -Economic & Social Commission for Asia and the Pacific Report, Thailand. [4] Hills B L and G J Elliott (1986). A microcomputer accident analysis package and its use in developing countries. In: Proceedings of Indian Roads Congress Road Safety Seminar, Srinigar. [5] Radin Umar R S and Aminuddin A (1992). Sistem Diagnosis Kemalangan Jalan Raya. Laporan Penyeldikan No. 1. Majlis Keselamatan Jalan Raya Malaysia. [6] Radin Umar R S (1993). Analisis Teperinci Kemalangan Jalan Raya: Projek Pilot Seremban, Shah Alam dan Petaling Jaya. Laporan Penyeldikan No. 4. Majlis Keselamatan Jalan Raya Malaysia. [7] TRRL, (1987). The.Traffic Conflict Technique Guidelines. Transport & Road Research Laboratory. Institution of Highways and Transportation, London.