ESIAION OF BUS RAVEL ORIGIN-DESINAION ARIX ON A CORRIR USING SCREENLINE COUNS Sama J. WIDANAPAHIRANAGE a ad Amal S. KUARAGE b a Road Developmet Authority, Battaramulla, Sri Laka b Departmet of Civil Egieerig,Uiversity of oratuwa,oratuwa, Sri Laka ABSRAC he developmet of a travel origi-destiatio (O-D) matrix is fudametal to trasport plaig studies ad a valuable commo iput for all key activities from plaig to maagemet of trasport. he O-D matrix describes the patter of travel movemets betwee a set of origis ad destiatios i a trasport corridor or regio. he developmet of a O-D matrix requires a large database of travel survey iformatio. Bus passeger travel origi-destiatio matrices are particularly difficult to make, specially where there is o regular practice of ticketig, ad whe buses are overcrowded due to deficiecy of seatig supply. Alteratively, household surveys could be used to develop a O-D matrix, but these require cosiderable resources. herefore, the compilatio of travel O-D matrices for trasport plaig studies is rare. his study was coducted i a busy corridor of public trasportatio i the Colombo metropolita regio i Sri Laka. he O-D matrix is formulated from the screelie couts ad has bee validated. he verificatio showed the percetage level of error of most iter-zoal travel estimates is below 5 percet. Also, the matrix is used to fid out the operatioal parameter of the bus routes such as the average trip legth betwee ay two zoes.. INRUCION he origi-destiatio (O-D) matrix or table is valuable key iformatio, which illustrates the patter of trasport activities betwee two travel zoes. he collectio ad aalysis of data for the developmet of a O-D matrix ivolves much time ad moey. However, much less time ad resources are required where there is a regular ticketig system. he lack of such a system i bus travel is oe drawback which could be highlighted as a reaso for the o-availability of O-D travel matrices i most developig coutries Sri Laka is o exceptio. he developmet of a O-D matrix from screelie couts, such as passeger loadig, boardig, ad alightig is less time-cosumig ad costs less moey. I travel markets where there is a costrait of seatig supply, the coutig of passegers becomes difficult because buses are overloaded most of the time. I such a situatio, these screelie couts are useful for formulatig a estimated O-D matrix. his is used to fid out the operatioal parameters of a bus route such as, its trasit capacity, estimatio of passegers travel distace etc.. he Study Area he trasport corridor selected for this purpose rus through oe of the most ecoomically active areas i Sri Laka. he area covers a extet of 48 hectare of lad with a populatio of.66 millio. (Naayakkara, 00). he corridor covers a total distace of 60 kms alog the wester
coast from Colombo to Beruwala as show i Figure. It also crosses six local admiistrative areas that are called Divisioal Secretariat Divisios, referred to i this paper as travel zoes. hus, the study area icludes six travel zoes, viz., Colombo, Dehiwala, oratuwa, Paadura, Kalutara, ad Beruwala, which are all withi the souther ed of the Colombo etropolita Regio. A cetral bus stad is located i each zoe, with a umber of bus halts alog the corridor for easy access to buses plyig this corridor. here are a total of 5 truk bus routes ad eight feeder bus routes coectig these zoes. All these bus routes ply o this corridor makig a total of 3,950 bus trips per day i both directios. ay busiess establishmets, uiversities, idustries, govermet istitutios, residetial areas, ad a airport are located withi this trasport corridor. his meas that there is a larger volume of ecoomic activity takig place withi this trasport corridor, tha i other parts of the Colombo etropolita Regio.. Bus ravel Network he selected trasport corridor is part of the Colombo Galle Road which is a primary truk road that starts from Colombo ad rus southwards through all the selected traffic zoes. here is mixed traffic flow at preset, of which 0 per cet are buses. he average headway of buses is aroud.0 miutes, with a miimum of oe miute closer to Colombo ad aroud three miutes closer to the souther ed of the corridor at Beruwala. ay secodary roads are coected to the Colombo- Galle Road at several locatios. Each such secodary road also has bus services which trasfer large umbers of bus passegers from the adjacet traffic zoes to the zoal odes withi this trasport corridor. Bam balapitiya (! COLOBO H a w e lla DEHIW ALA K o tta w a Padukka aliba Juctio ORAUW A PANADURA (! Badaragam a Horaa W adduw a Leged Railway Network A Class Roads KALUARA B Class Roads atugam a Kilometers 0.5 3 6 9 BERUW ALA Figure. ap of the South-Colombo etropolita rasport Corridor
.3 Literature Review he available literature has paid little attetio to the estimatio of origi-destiatio matrices from screelie data. ost O-D matrices are developed usig roadside iterview-based travel patters from traffic couts (Zuyle ad Williamso, 980; Bell, 99). However, Wog ad og (998) developed O-D matrices from passeger boardig ad alightig couts. odo ad Hess (003) used bus boardig ad alightig couts to ivestigate pedestria safety ad trasit corridor capacity. But these studies have ot cosidered the costrait of passeger overloadig for the estimatio of O-D trips, as dispatched bus ticket iformatio is available at the time of study. I this paper, bus passeger loadig couts were cosidered together with boardig ad alightig couts to avoid the limitatio that could arise by coutig bus passegers based o ticketig due to o issue of bus tickets by private bus operators. he Govermet owed bus compaies issue bus tickets but they represet oly 30 percet of the total iter-zoal bus supply.. OBJECIVES AND EHOLOGY he aim of this paper is to preset (a) the theoretical approach ad form for developig a bus passeger O-D travel matrix from scree lie data ad (b) the validatio ad (c) the applicatio of this theoretical approach to formulate a O-D matrix for the travel corridor selected.. Data Collectio wo sets of data were collected for this research. he first set icluded a sample of bus boardig ad alightig observatios collected alog the corridor for each of the iter-zoal bus routes. he 5 bus routes were idetified by the study by observig several bus haltig places, which were located alog the corridor. hese were located approximately at equal distaces alog the route. By assumig that most of the bus trips are roud trips, ad that the passegers board the bus for their retur trip at about the same place where they get-off, Navick ad Furth (00), stated that the boardig patter for a route i oe directio is equivalet to the alightig patter i the opposite directio over the course of the day. herefore, the place where a passeger alights from a bus is most likely to be the same or ca be cosidered as beig very close to the place where passeger got o to the bus for the retur trip. herefore, it was assumed that the corridor demostrates a regular patter of passeger boardig ad alightig. he boardig ad alightig couts o all these routes were collected o-board the buses by traied eumerators, who made a total of 90 eumerator trips, with at least three roud trips per route. he couts were made for each stop alog the corridor durig 3 hours of a week day, startig at 06.00 hours ad edig at 9.00 hours which covered both peak period as well as off-peak travel patters. I the secod set of data, roadside (screelie) bus loadig data were collected by maual observatio at selected strategic poits alog a route so that the maximum umber of bus routes could be observed at these screelies. he screelie occupacy survey of bus passegers was somewhat difficult as these routes usually have large umbers of stadig passegers, ad also because the sizes of buses vary makig estimatio difficult. o overcome this problem, a estimatio matrix was developed so that the umber of passegers could be estimated based o the size of the bus, ad the level of loadig, both of which were easier to observe agaist precouted passeger occupacy i buses of various sizes as show i able below. 3
able. Passeger Occupacy i Buses Bus Size ea Passeger Load by Level of Loadig 3 4 5 6 7 Empty < / > / Full Full Capacity Overflow +Capacity < 0 Seats 3 8 5 8 3 37 0-9 Seats 4 7 8 7 3 39 58 30-39 Seats 5 4 3 38 56 67 Over40 Seats 6 3 3 46 48 69 0. heoretical Formulatio of O-D atrix I developig a geeral approach to formulate a O-D matrix from screelie data, it is assumed that there are umber of traffic zoes across which a bus operates o a give route or corridor. he iter-zoal travel demad from origi to destiatio ( ( )) at the ed of zoe (-) is illustrated i Figure. I this estimatio, it was assumed that the roadside loadig observatio poit (LOP) was situated ahead of the zoal boudary. Origi O K = B K ZB - B LD B ( ) ( ) LOP ZB - Zoe - Figure. A Bus Route Crossig th Zoes ( ravelig from O to D) LD Where, ( ) = he total bus passeger loadigs observed at a roadside loadig observatio poit located withi (-) th zoe from origi o to destiatio d. ( ) = he iter-zoal bus passeger demad betwee the (-) th zoe ad th zoe from origi o to destiatio d. B = Bus passeger boardig i the (-) th zoe from roadside observatio poit (LOP) to the zoal boudary (ZB - ) from the directio of origi o to destiatio d. A = Bus passeger alightig i the (-) th zoe from roadside loadig observatio poit (LOP) to the zoal boudary (ZB - ) from origi o to destiatio d. K = B K = K A K + B A A Destiatio D = he cumulative bus passeger boardig up to the roadside loadig observatio poit (LOP). 4
K = A K + A = he cumulative bus passeger alightig up to the roadside loadig observatio poit (LOP). he first step is to calculate the differece betwee the cumulative passeger boardig ad the cumulative passeger loadigs. It is determied by the iter-zoal passeger travel (IP) betwee the (-) th zoe ad the th zoe give i equatio () which estimates the average umber of bus passegers who would be travellig across the (-) th zoal boudary (ZB - ) whe a bus moves from origi o to destiatio d. IP = BK K = K = A K + B + B I estimatig the total passeger demad at the (-) th zoal boudary (ZB - ), further adjustmets are required to estimate the umber boardig ad alightig betwee the roadside loadig observatio poit (LOP) ad the boudary. his ca be give by equatio (). ( ) Similarly, equatio (3) is derived to estimate the total iter-zoal passeger demad at the same boudary whe travellig from destiatio zoe d to origi zoe o. I this estimatio, it was assumed that the roadside loadig observatio poit (LOP) was situated ahead of the zoal boudary. A ( B A ) = + * LD BK AK B B A A + + K= K= A ( ) () () ( B A ) = * LD BK AK B B A A + + K= K= ( ) ( ) (3) For this type of arragemet the geeralized form of origi to destiatio travel matrix for the give bus route ca be illustrated as. ( ) ( ). 0 = 0 3... 0... ( )... 0 ( ) (4) 0 (4) 5
Where, ij is the estimated iter-zoal passegers travel betwee i th zoe ad j th zoe. (i=,,.,) ad (j=,,,);the coefficiets, 3,., ( ) are the factors that are derived from the bus passeger boardig ad alightig couts o the bus route as show below, A j ij = Pij * Bm m= (5) Where, P ij is the ratio of all passegers who alighted i zoe j to the ratio of passegers who boarded from zoe i to travel up to zoe j which is expressed as a percetage. I this study, the ij value for was ivestigated based o experiece of the route boardig ad alightig patter. m= B m is the cumulative bus passeger boardig up to destiatio d. A j is bus passeger alightig i the j th zoe..3 Applicatio for a ravel Route his theoretical approach is used to estimate the iter-zoal travel for the selected travel route. ij he first step is to determie approximate values for the parameters. he followig example illustrates how this ca be doe. Estimatio of Iter-zoal Bus Passeger ravel o oratuwa - Colombo bus Route (Example Route Number 0) he oratuwa - Colombo bus route (oe of the 5 routes observed i this study) plies across three traffic zoes ad the roadside passeger loadig observatio poits for this routes were located at the aliba Juctio (LOP-), ad Bambalapitiya (LOP-) as show i Figure. I this example, the observatio poits were situated ahead of the respective zoal boudaries whe buses operate from oratuwa to Colombo. he observed average daily passeger demads at C these two locatios were,34( LD (, J )) ad 3,44( LD (, C BP) ) respectively i the oratuwa to C Colombo directio, ad 7,950( LD (, C J )) ad,304( LD (, BP) ) respectively i the Colombo to oratuwa directio. he summary of passeger boardig, ad alightig o this route is show i able. his was estimated usig a 0 per cet sample from 93 buses operatig o this route. hese buses were foud to make 465 bus trips i both directios durig the day of the survey, which covered approximately 95 per cet of the estimated total iter-zoal passeger retur trips o this bus route. 6
able Passeger Boardig ad Alightig Couts (Bus Route Number 0) Locatio Withi oratuwa Zoe aliba Juctio (LOP-) Withi Dehiwala Zoe Bambalapitiya (LOP-) Withi Colombo Zoe Colombo to oratuwa Average passeger boardigs = ( o) 9 Average passeger alightigs = oratuwa to Colombo Average passeger boardigs, C = Average passeger alightigs, C A = B A ( ) 67 B ( ) 55 ( ) 6 o, C, C B ( J ) = 0 A ( J ) = 6 B ( J ) = 4 A ( J ) = 5, C, C B A ( ) = 0 B ( ) = A ( ) = ( De) = 65 ( BP) = 78 De, C, C B A ( ) = 4 B ( ) = 0 A ( ) = 0 ( Co) = 5 BP, C, C B A ( ) = 7 B ( ) = 36 A ( ) = 64 Co o De BP Co o De BP Co As i able, the average passeger boardigs per bus withi the oratuwa zoe is 55 passegers ad alightig is 6, whe the buses operate from oratuwa to Colombo. At the loadig observatio poit (LOP) which is withi the ext zoe (Dehiwela) the boardig is 4 passegers ad alightig is 5 passegers. Calculatio ravel from Colombo to oratuwa. he loadig observatio poit i the Colombo zoe is located at Bambalapitiya (LOP-), before the Colombo zoal boudary. he usig Equatio, the iter-zoal travel betwee Colombo ad Dehiwela would be, C. IP ( Co) = [ BCo ACo + BBP ABP ] = 44 (6) Similarly, the iter-zoal passeger travel betwee Dehiwela ad oratuwa would be, C. IP ( De) = [ BCo + BDe ACo ADe + BBP + BJ ABP AJ ] = 30 (7) he estimatio of total bus passeger travels at the Colombo zoal boudary was. ( BCo ACo ) Colombo = + LD * BP (8) IPCo = 3,359 passegers per day. he LOP-, is located before the Dehiwela zoal boudary. Similarly, the estimatio of total bus passeger travels at the Dehiwela zoal boudary was. ( BDe ADe ) Dehiwala = + LD * J (9) IPDe =0,70 passegers per day. hereafter we make the assumptio that P (Colombo to Dehiwela) is 95 % as cosidered i Equatio 5. he the coefficiet estimates, Colombo Dehiwala ( A + A ) 0 De J Colombo Dehiwala =.95* = 0.57,,,,, (0) C C C C C BCo + BBP + BDe + BJ + Bo 7
( ij ) (Colombo-Dehiwela) = 3,433 passegers per day / (Dehiwala-oratuwa) = 776 passegers per day ad (Colombo-oratuwa) =9,96 passegers per day. Calculatio ravel from oratuwa to Colombo. I oratuwa to Colombo bus travel, the loadig observatio poit is located further away from the oratuwa zoal boudary, he usig Equatio, the iter-zoal travel betwee oratuwa ad Dehiwela would be,, C, C, C, C, C IP = B A + B A = () ( o) [ o o J J ] 38 Similarly, the iter-zoal passeger travel betwee Dehiwela ad Colombo would be,, C, C, C, C, C, C, C, C, C IP = B + B A A + B + B A A = () ( De) [ o De o De J BP J BP ] 8 Hece, usig Equatio 3, the estimatio of total bus passeger travels at the oratuwa zoal boudary was., C, C, C ( BJ AJ ), C oratuwa = LD, C * J (3) IPo =,639 Passegers per day. Similarly, the estimatio of total bus passeger travels at the Dehiwela zoal boudary was., C Dehiwala, C, C ( B A ), C De De, C IP De = * LD (4) BP = 8,43 Passegers per day. oratuwa Dehiwala Assumig the P (oratuwa to Dehiwala) is 90 % as i Equatio 5. he the coefficiet estimates,, C, C ( ) 0 A De + AJ oratuwa Dehiwala =.9 * = 0.84, C, C, C, C, C (5) Bo + BJ + BDe + BBP + BCo Similarly, Iter-zoal travel (oratuwa-dehiwela) =,4 passegers per day ad (oratuwa-colombo) = 9,497 passegers per day ad (Dehiwela-Colombo) = 8,746 passegers per day. he aalysis was carried out for all of the other bus routes ad estimates were thereafter made usig the summatio of all the ij travels as show i equatio (6) relevat to the travel from origi o to destiatio d to obtai the iter-zoal bus demad ( ij ) for that zoal pair. ij = ij j= (6) he estimated iter-zoal demad ad the observatios are tabulated i able 3. 8
3. VALIDAION OF O-D ARIX Validatio is a importat step to check the accuracy of the theoretical approach of estimatio before wider applicatio. able 3, shows the compariso betwee the estimated travel demad ad the estimated observatios, which are computed from the sample travel data. he observed travel volumes were from data collected from two differet trasport developmet studies doe o this corridor (OCH,000 ; CRSP,997),due to o-availability of the complete sets of data for a sigle study. able 3. Validatio Results for Iter-zoal ravel Estimates. Zoe Name rip Legth Validated Year -004 % error of From Zoe o Zoe (km.) Obs. ij ij Colombo Dehiwela 0.0 59,305 68,95-6.7 Colombo oratuwa 0.0 7,90 3,55-5.48 Colombo Kalutara 43.0 5,34 4,87 6.94 Dehiwela oratuwa 40.0 0,57 8,68 5.5 Dehiwela Paadura 7.0 7,896 8,047 -.9 Dehiwela Kalutara 33.0,783,545 8.55 oratuwa Kalutara 3.0 4,5 3,675.8 Obs. =Observatio ij = Bus demad from i th zoe to j th zoe. able 3, shows that the highest flow occurs for iter- zoal travel of the shorter distaces cetred aroud Colombo. Estimated demad for Colombo - Dehiwala ad Colombo - oratuwa travel is sigificatly more tha the observed demad. he high egative ad positive error terms idicate that a larger sample could have yielded a better accuracy. However with the least accuracy at +/- 5% there is sufficiet cofirmatio that the theoretical approach has a acceptable level of accuracy for use i trasport plaig studies. 5. CONCLUSIONS he paper documets the formulatio ad calibratio of a method of calculatig a travel O-D matrix for a trasport corridor from screelie passeger loadig couts, ad boardig ad alightig couts. his has bee verified through a validatio returig idividual errors of below 5 percet. he deferece of this method from other methods is that this method overcomes limitatios posed by overloadig of buses, ad istaces whe bus ticket iformatio ad bus dispatchig iformatio are ot available for estimatig a O-D matrix. his approach eables estimatig iter-zoal travel legth i a corridor to a high level of accuracy where a umber of buses ply across a umber of differet travel zoes. his method is useful for the bus idustry because the compilatio of such travel iformatio with miimal survey observatios ca save both moey ad time. ACKNOWLGEENS he authors gratefully ackowledge the fiacial assistace provided by the Asia Developmet Bak to coduct this research, ad the facilities provided by the Departmet of Civil Egieerig, Uiversity of oratuwa, Sri Laka. 9
REFERENCES Bell,.G.H.(99) he Estimatio of Origi- Destiatio atrices by Costraied Geeralized Least Squares, rasportatio Research 5, 3-. Colombo etropolita Regioal Structural Pla, (997) rasport Ifrastructure ad Service Requiremets for Colombo etropolita Regio Fial Report, Urba Developmet Authority, Sri Laka, pp.-3. ado, A.V. ad Hess,.(003) Pedestria safety ad rasit Corridors Research Report No.0, Departmet of Urba Desig ad Plaig, Uiversity of Washigto, Seattle, Washigto, pp. 0-7. Navick, D.S. ad Furth, P.G.(00) Usig Locatio Stamped Farebox Data to Estimate Passeger iles, O-D patters, ad Loads, Paper preseted to the 8 th Aual eetig of the rasportatio Research Board, Washigto D.C. Naayakkara, A.G.W.(00) Cesus of Populatio ad Housig, Prelimiary Results, Departmet of Cesus ad Statistics, Colombo, Sri Laka, pp.-3. Outer Circular Highway, (000) Feasibility Study Report Vol.0, Road Developmet Authority, iistry of Highways, Sri Laka, pp. 5-9. Wog, S.C. ad og, C.O.(998) Estimatio of ime Depedet Origi-Destiatio atrices for rasit Networks, rasportatio Research A 3, 35-48. Zuyle, H.J. ad Williamso, L.G.(980) he ost Likely rip atrix from raffic Couts, rasportatio Research B 4, 8-93. 0