Using Accident Records To Prioritize Roadside Obstacle Improvements in New Mexico

Transcription

1 1 ransportation esearh eord 147 Using ident eords o Prioritize oadside Obstale mprovements in New Mexio JMS D. BOGN and JOM W. HLL BSC his paper ontains a desription of a proess for identifying setions of rural New Mexio nterstate, primary, and seondary highways with signifiant fixed-objet aident experiene. Data bases for the analysis are the omputerized aident reord and roadway inventory systems for the 3-fisal-year period from 198 to he rate quality ontrol method is used as the statistial tehnique to identify those setions of roadway that are most in need of examination. ident and inventory information are ombined to alulate ritial aident rates for setions of roadway on eah of the three systems. Calulated ritial rates are then ompared with the atual rates on eah setion, and a listing of setions arranged by ritiality (the differene between the atual and ritial rates) is obtained. his listing is used by New Mexio State Highway Department personnel to prioritize loations for implementing safety improvements" he proedure is appliable to the analysis of other subsets of aidents as well. Data quality is ritial for the proper appliation of the tehnique; fators other than aident experiene, in addition, must be onsidered in the ost-effetive development of aident redution ountermeasures. National statistis <!.> indiate that over 63 perent of fatal traffi aidents involve only a single vehile. ven when pedestrian and pedalyle aidents are exluded, single-vehile aidents still aount for over one-half of all traffi fatalities. he most frequently ited "first harmful event" for single-vehile fatal aidents on a national level is ollision with a fixed objet (48 perent of all single-vehile fatalities), followed by ollisions with pedestrians and pedalylists (29 perent) and overturning aidents (17 perent). ural highways, whih aount for 44 perent of the nationwide total vehile miles of travel (VM), aount for 56 perent of the single-vehile fatal aidents. Despite the emphasis on lear roadsides for nterstate highways, oupant fatalities in single-vehile aidents remain the largest omponent of the fatality toll on these failities. raffi aident data in New Mexio reflet many of the national trends. n 1983, for example, 69 perent of New Mexio's fatal aidents were of the single-vehile type, while for the 3-fisal-year period from 198 to 1982, over 58 perent of the nonpedestrian fatal aidents involved a single vehile <ll. Beause of the rural nature of New Mexio, 7 perent of fatal aidents ourred in rural areas, even though less than 6 perent of the state's VM take plae in these areas. Other New Mexio raffi aident harateristis deviate from national norms. New Mexio's fatal aident rate is onsistently one of the highest in the nation, while the perent of fatal aidents involving multiple vehiles is among the lowest. Fatal overturning aidents, for example, are overrepresented, probably beause of the relatively unluttered roadside terrain in many parts of the state (3). On the other hand, fixed-objet aidents our somewhat less frequently than expeted. he sum of these two aident lasses ould generally be desribed as run-off-the-road aidents. Fixed-objet aidents in 1982 aounted for 14 perent of the state's total aidents and 12 perent of its fatalities. Beause of the relative ease with whih fixed-objet ountermeasures may be implemented, a study was undertaken in New Mexio to establish a manageable set of hazardous loations and to develop improvement priorities. he purpose of this paper is to report on a proedure for ranking in priority order rural fixedobjet aident site improvements in New Mexio. lthough the general aspets of the proedure are not unique (_!), ertain features of its appliation are examined in greater detail. he omputerized aident reord and roadway inventory systems are ombined to identify roadway setions with ritial fixed-objet aident rates. Computerized data are supplemented by the use of hard-opy aident reports; photologs; and individual site visits. lthough rural nterstate segments were initially examined from the reords system, prinipal emphasis for improvements is given to rural federal-aid primary (FP) and federal-aid seondary (FS) failities in the state. Fixed-objet aidents involving guardrails, moreover, are not onsidered beause they were the subjet of a previous study (_~). Subsequent setions of this paper ontain desriptions of fixed-objet aident experiene in the state, disussions on the appliation of the rate quality ontrol proedure for identifying ritial fixed-objet aident loations, and an outline of other onsiderations in the ranking in priority order of improvement sites. While the speifi examples developed in this paper apply to rural, single-vehile aidents in New Mexio, the general priniples are appliable to other signifiant subsets of traffi aidents. FXD-OBJC CCDN CHCSCS Computerized aident reords show that during the 3-fisal-year period from 198 to 1982, approximately 143, aidents were reported in New Mexio. Other aidents, partiularly those involving a single

2 Brogan and Hall 11 vehile, no doubt ourred but are not in the reord system for several reasons <il hese reported aidents were distributed by aident lass as indiated in able 1. he data show that fixed-objet aidents are an important, although not major, portion of New Mexio's aident experiene. BL 1 idents by Class for Fisal Years otal idents ident Class No. Perentage No. Other vehile 87, Fixed objet 2, Overturning 11, Pedestrian/pedalylist 3, Other 2, otal 143, ,492 BL 2 Single-Vehile, Fixed-Objet idents by Loation and Severity Loation Fatal njury PDQ otal ural 18 2,164 4,2 6,472 Urban 71 3,279 9,877 13,227 otal 179 5,443 14,77 19,699 Note: Severity index =(fatal+ injury)/total. Fatal idents Perentage Severity ndex Further analysis found that the 2,14 fixed-objet aidents shown in able 1 inlude 315 aidents involving two or more vehiles. he remaining 19,699 single-vehile, fixed-objet (SVFO) aidents are distributed by loation and severity as shown in able 2. lthough two-thirds of the SVFO aidents our in urban areas, over 6 perent of the fatal rashes and almost 4 perent of the injury aidents are in rural loations. he severity index (the ratio of fatal plus injury aidents to total aidents) is thus onsiderably higher for rural areas undoubtedly beause of higher speeds, different types of fixed objets, and the probable underreporting of some rural, single-vehile, property-damage-only (PDO) aidents. he 6,472 rural SVFO aidents inlude many that ourred on nonfederal-aid roads, suh as other state, ounty, ndian reservation, and u.s. Forest Servie roads. he exlusion of aidents on these other roads as well as fixed-objet aidents involving guardrails (,l provided a sample of 3,432 nonguardrail SVFO aidents on New Mexio nterstate, FP, and FS systems during the 3-year period. t is this set of aidents (approximately 2 perent of New Mexio's total aident experiene) that is examined in greater detail in this paper. Figure 1 shows the proess used to selet the aidents of interest, beginning with a omputer tape of all New Mexio traffi aidents for the 3-year period and ending with a set of 3,432 nonguardrail SVFO aidents on rural federal-aid highways. lthough the average severity index (.38) for these aidents on the nterstate system is idential to those for the FP and FS systems, the systems learly differ in their roadway and roadside design harateristis. Previous researh has doumented substantial variation in severity indies as a funtion of the type of objet struk ll n alternate tehnique for assessing the severity of impat with a partiular objet type is to weight the average NHS osts for fatal, injury, and PDO aidents by the probability of these severity levels. he probabilities an be estimated by the observed relative proportions of impats with a speifi objet type that r esult in a fatal, injury, or PDO aident. Beause of the subjetive aspets of the NHS osts and the lak of homogeneity of objets within a partiular ategory, the results of this analysis are relative rather than absolute. he average ost for all aidents examined in this projet was $8,8. he severity indies and estimated osts of rashes involving various objet types are given in able 3. he absene of trees and utility poles along New Mexio's rural nterstate is refleted by the proportions of rashes involving these fixed objets. Culverts, medians, and traffi signs, on the other hand, omprise a larger perentage of nterstate than FP and FS rashes. Fenes (right-of-way fenes on the nterstate) and embankments (in reality, onsisting of both ut and fill slopes) are struk with similar frequenies on both systems. he ategory "other" inludes barriades and onstrution material and equipmenti these fixed-objet types onstitute a slightly larger portion of the aident experiene on the nterstate system. he average ost values exhibit substantially more variation than the severity indies. o a ertain extent, this ost analysis proedure emphasizes those objets that are more likely to result in a fatality. However, aution must be exerised in using small aident samples where one or two fatalities an dramatially alter the average osts. Other harateristis of rural SVFO aidents are sununar ized in able 4. n general, these aidents our on urves more often than on tangent setions and are more onunon on downgrades and during hours BL 3 Fixed-Objet ident ype by oad System for New Mexio for Fisal Years nterstate FP and FS Fixed Objet Severity Severity ype Perentage ndex Cost($) Perentage ndex Cost($) Culvert , ,7 mbankment , ,3 Bridge , ,1 ree , ,9 Utility pole , ,4 Dith 6.3 8,7.37 8,4 Median or urb ,7.24 3, Guard posts , ,4 Fene , ,9 raffi sign , ,8 Other , ,1 Note: Severity index (S)= (fatal+ injury)/total; FP and FS= federal-aid primary and seondary, respetively.

3 12 ransportation esearh eord 147 Selet only fixed objet aident Fixed Objet idents (214 ),... (14.) Single Vehile Fixed Obiet idents ( 19699) ( l 3. 7). f ]iminate multiple vehile aidents *Number of Observations **Perent of otal i liminate urban aidents ural SVFO e. (6472) (4.5) l....,. liminate "other" roads ural SVFO. on JS FP FS (4424) (3. l) - S.. =.37. on JS FP nterstate tFS (737) - i liminate G aidents (studied 1 previously) ura 1 Non-G SVFO (3432l (2.4 - S.. =. 3 B FP & FS (2695) - Futa.1 (2) lnjury (257) PllO (46) Fatal ( 4 5) nj ury (9 78 ) PDO ( 1672) FGUK l ident seletion proedure. BL 4 Charateristis of ural, Nonguardrail SVFO idents for New Mexio for Fisal Years nterstate FP and FS (%) (%) Weather Clear ainy Horizontal alignment Straight Curve Vertial alignment Level Grade Lighting Day 51 4 Night 49 6 of darkness. he results are onsistent with those reported in the tehnial literature and the findings of previous studies of run-off-the-road aidents in New Mexio Clril DNFCON OF CCL LOCONS Disussion in the previous setion entered on the harateristis of nonguardrail, single-vehile fixed-objet aidents on New Mexio's rural nterstate, primary, and seondary roadway systems. lthough this information is perhaps interesting from an aggregate statistial standpoint, it is of little use in identifying hazardous loations sine speifi objets on moderate volume rural highways are rarely struk more than one and, beause exposure, in terms of volumes and setion lengths, is not speifially taken into aount. he tehnial literature (7) disusses a number of tehniques for the identifiation of hazardous loations. ording to one survey (~), the most ommonly used proedure is based on the number of aidents, followed by tehniques using aident rates and rash severity. he shortomings of relying primarily on the number of aidents are well doumented. lthough rate-based tehniques inorporate exposure, thus eliminating one problem, they indue artifiially high rates when only a few aidents are ombined with low exposure. f reasonably

4 Brogan and Hall 13 omplete and reliable data are available, this defiieny in the identifiation of hazardous loations may be overome through the use of qualityontrol, statistial tehniques, ommonly known as rate quality ontrol methods (4,9). pproximately 15 states make some use of this te"chnique in the proess of identifying hazardous loations (1) Subsequent setions of this paper ontain disussions on the appliation of the rate quality ontrol method to the identifiation of rural, fixed-objet aident loations in need of remedial ation. he tehnique itself is disussed firsti appliation of the tehnique to rural federal-aid highways in New Mexio is then onsidered. H QULY CONOL MHOD he rate quality ontrol method alulates a ritial aident rate (C) for eah setion of roadway. he value of C is a funtion of the systemwide aident rate (), the amount of VM on the setion (m), and a fator (k) based on the desired level of statistial signifiane. he relationship is C= + {k[(/m) 1 / 2 1} +.5/m (1) he term /m is an estimate of the variane of the aident rates, while.5/m is a ontinuity orretion. genies using this tehnique for the general. identifiation of hazardous loations reportedly use a variety of levels of signifiane (a) ranging from.5 to.5, and possibly higher. he hoie of a establishes the value k in the equation, with lower values of a orresponding to higher values of k and resulting in a shorter list of hazardous loations. Higher values of a redue the likelihood that a truly hazardous loation will be overlooked, but the larger list of loations generated in this proess will inlude many sites that are not atually hazardous. Under onditions of finanial and personnel onstraints, it may be appropriate to selet a value of a that will result in a manageable list of loations warranting further study. using a normal distribution table, the hoie of a.5 yields a value for k of 1.645, and quation 1 redues to C= + {1.645( (/m)l/21} +.5 m (2) where C setion rural nonguardrail, fixed-objet ritial aident rate (aidents per million VM on the setion), systemwide rural nonguardrail, fixed-objet aident rate (total SVFO aidents per total travel, not the average of individual setion rates), and m =million VM on the setion. he ritial rate is obviously greater than the systemwide aident rate. t dereases with inreasing travel on the individual study setions. f the amount of travel and the SVFO aident experiene on a setion are known, the atual setion rate an be alulated and ompared with its ritial rate. Within the limitations imposed by the quality of the traffi aident and travel data, setions where the atual rate exeeds the alulated ritial rate are said to be hazardous at the 5 perent level of signifiane. PPLCON OF H QULY CONOL CHNQU ppliation of the foregoing tehnique for establishing roadside obstale improvement priori ties in New Mexio involves identifying setions of rural nterstate, primary, and seondary highways with higher-than-ritial, fixed-objet aident experiene. he initial step involves ombining the omputerized aident reord and roadway inventory data files to determine the for eah of the three roadway systems. he next step involves using these data files to alulate and ompare the atual and ritial rates on eah roadway setion. Setions are then ranked aording to their ritiality, that is, the differene between the atual setion rate and the ritial rate for that setion. he proess is shown in Figure 2 and is desribed in the following paragraph. he ritial, fixed-objet aident rate alulation begins by seleting from the aident file those aidents of interest, in this ase, rural single-vehile, nonguardrail, fixed-objet aidents on the nterstate, primary, and seondary systems. he total travel by system for the 3-year study period is estimated by using the individual setion lengths and annual average daily traffi ounts (Ds) in the inventory file. he average rates are then alulated for eah system. here are a number of tehniques for establishing individual roadway setions. deally, setions should be homogeneous with respet to roadway design and development, traffi volume, and speed. While short setion (<.5 mi) are more likely to be homogeneous, they also have limited aident experiene he New Mexio roadway inventory establishes setions prinipally on the basis of the onstrution ontrats under whih they were built. hus, the setions vary in length, but are reasonably onsistent in design and operational features. he inventory ontains 35, 1,237, and 1,28 setions on the rural nterstate, primary, and seondary systems, respetively. nventory setions average 3 mi in length, but there is onsiderable variation among individual setions. o failitate the analysis and redue the number of individual roadway setions to be onsidered, the traffi volumes of adjaent setions were ompared. f Ds on adjaent setions differed by less than 1 vehiles per day (vpd), the setions were ombined. his proess redued the total number of study setions from 2,57 to 967. Determination of ritial aident rates on these setions is aomplished by first alulating the VM on eah setion (the produt of D, setion length, and the 1,96 days in the 3-year period). ritial aident rate is then alulated using VM and the previously alulated aident rate for the roadway system. hose setions on whih the atual rate exeeds the ritial rate are flagged and ranked aording to the differene between the two rates. POGM SULS Calulated systemwide SVFO aident rates for the nterstate, primary, and seondary systems are.114, O.187, and O. 344 aidents per million VM, respetively. s expeted, the nterstate rate is relatively low (one fixed-objet aident for eah 8. 8 million VM) while the seondary rate is three times higher (one fixed-objet aident for eah 2.9 million VM). Use of these systemwide rates results in the following er itial rate alulations at the 5 perent level of signifiane for individual setions on eah of the three systems: C C C {.555[ (l/m)l/ 2 ]} +.5/m (S) {. 711 [ (l/m) 1/21 } +. 5/m (P) {.965[(1/m)l/21} +.5/m (S) (3) (4) (5)

5 14 ransportation esearh eord 147 Delete ll But (S P S) ural SVFO Non-G idents Delete ll But (S P S) ural Set1ons Calulate Systemwide ates for (S. P S) Calulate tual ident ates on ndividual Set i ans (e)calulate Critial ident ate (C) for (New) Setion (e)calulate VM on ndividual Set i ans (D * len~th * 3 years Delete No Ho FGU 2 Calulation of ritial fixed-objet aident rates for New Mexio for fisal years where S is nterstate system, P is primary system, and S is seondary system. For the purposes of this analysis, the value of m is the number of million VM on a setion during the 3-year study period. he ritial rate relationships given in the preeding equations are plotted by roadway system in Figures 3, 4, and 5. Observed aident rates above the plot are ritial at a level of signifiane of a =.5. he absissa of eah figure is atually the D on a setion exatly 1 mi long. t may also be interpreted as the daily VM on a setion with a length other than 1 mi. For example, to apply Figure 4 to a 3-mi roadway setion with an D of 5,, the figure is entered at 15, on the absissa. lthough the priniples underlying the development of these figures have general appliability, the atual figures are only valid for the 3-year study period on rural New Mexio highways. he appliation of this proedure identified a total of 14 nterstate, 47 primary, and 59 seondary setions that had atual aident rates higher than the ritial rates. n other words, between 1 and 14 perent of the setions on these routes were judged to be hazardous. n example of the output information showing the five most ritial setions on the FS system is shown in able 5. he information listed in able 5 shows the administrative route number on whih the setion is loated, the beginning and ending mileposts of the er i tial setion, the number of SVFO aidents on BL 5 Fixed-Objet idents on ural New Mexio Seondary oads ate per Million VM dministrative Beginning nding No. of oute No. Milepost Milepost idents Daily VM tual Critial o.o

6 Brogan and Hall l. 5 L D N l. O.5. Critial egion Sys temwide verage. l l l l l VG DLY FFC FGU 3 Critial fixed-objet aident rates for rural New Mexio nterstates for fisal years the setion during a 3-year period, the daily VM on the setion, and the atual and er i tial aident rates on the setion. he setions are ranked by ritiality, that is, the differene between the atual and ritial aident rates for the setion. lthough it is not obvious from the abbreviated listing in able 5, a slightly different ranking is obtained if the setions are arranged in dereasing order by the ratio of atual to ritial aident rates. Listings suh as that shown in able 5 an be used by New Mexio State Highway Department engineers for establishing priorities for those setions of various roadway systems in the state that appear to warrant more attention in the amelioration of fixed-objet aident hazards. o be useful, however, the list must be restrited to a manageable number of setions that an be examined in greater detail. he 12 highway setions identified by this proess were judged to be too many to be aommodated within the onstraints of this program. One logial approah for shortening the list is to use a smaller value of a. his has the effet of inreasing the value of k in the ritial rate equation, thus inreasing the value of C and reduing the number of ritial setions. prinipal shortoming of this approah is that 2 perent of the setions, inluding several short setions near the top of the list, had only 2 or 3 aidents during the 3-year period. lthough these setions may truly be ritial, it is also quite possible that the misoding of a single aident's loation by as little as O.l mi ould alter the setion's lassifiation from hazardous to safe. preliminary attempt to restrit the number of study setions involved the use of utoff values, expressed as number of aidents per mile (1.2 for the FP, 1.7 for the FS). lthough this effort eliminated a few short setions with low travel and 2 or 3 aidents, its prinipal effet was to eliminate longer setions. separate analysis of onstrution and projet planning reords also revealed that several setions had reently been reonstruted, and these setions were dropped from further analysis. OH CONSDONS lthough the use of omputerized aident reord and roadway inventory data files provides an exellent guide for identifying roadway setions with high fixed-objet aident experiene, in reality, several additional steps are neessary before aident ountermeasures may be implemented. Beause it is often diffiult to determine with ertainty from omputerized reords whether the same objets are being struk repeatedly, the next logial step in the proess must be a review of hard-opy aident reports for roadway setions of interest. his proess unovered a number of instanes in whih the aident loations were misoded. Beause roadside improvements are typially made at spot loations rather than overextended setions, a more thorough review of identified setions is neessary before ountermeasures may be implemented. s a first step in this proess, photolog reviews of the er i ti al setions were undertaken to identify speifi objets along the roadside that may warrant attention. his effort

7 16 l. l l.o.9.8 L.7 F Critial egion.6 D N.5 D D. l Systemwide verage. l l VG DLY FFC FGU4 Critial fixed-objet aident rates for rural New Mexio primary systelll8 for fisal years L 2. s 2. Critial egion l.5 D N.5. l 2 D 3 Systemwide verage D l VG DLY FFC FGU 5 Critial fixed-objet aident rates for rural New Mexio seondary systems for fisal years

8 Brogan and Hall 17 was quite helpful, although ertain types of roadside fixed-objets (dithes, ulverts) were not readily disernible on some photologs. Finally, field reviews of ritial sites are essential in identifying loations where improvements are to be made and in establishing the proper type of remedial ation. SUMMY ND CONCLUSONS n attempt has been made to desribe New Mexio's rural single-vehile, fixed-objet aident experiene and to develop a proedure, using the rate quality ontrol method, of identifying roadway setions that have unexpetedly high fixed-objet aident rates. lthough this proedure has been applied here to a partiular aident type, it should be realized that the proess is appliable to most aident subsets with suffiient sample size. he proess has, in fat, also been used to examine single-vehile, run-off-the-road aidents on New Mexio's seondary system (~). However, several onerns regarding the use of this proedure should be reognized. First, there may be roadway setions that have aident rates just below ritiali hanging the statistial level of signifiane, then, will affet the number of setions identified as ritial. Seond, it would appear from an examination of the New Mexio data that the loational information in the roadway inventory file is superior to that in the aident reord systemi inorret milepost oding of an aident loation ould thus affet the rate for that setion. hird, it is entirely possible that a ompletely different group of setions would be identified as ritial if other aident subsets were onsidered. omprehensive program of roadside safety in the state should thus reognize the importane of these other aident types. Finally, it should be realized that past aident experiene is not the only reliable indiator of hazard. Past aidents annot be eliminatedi with proper evaluation and development of ountermeasures, however, the number of future aidents may be redued. CKNOWLDGMNS he work reported on in this paper was funded by the raffi Safety Bureau, New Mexio Department of ransportation. ess to the omputerized aident reords and the roadway inventory was provided by.u. nderson of the Division of Government esearh, the University of New Mexio. FNCS 1. Fatal ident eporting System National Center for Statistis and nalysis, NHS, Washington, D.C. 2. New Mexio raffi ident Data raffi Safety Bureau, New Mexio ransportation Department, lbuquerque, pril J.w. Hall and P. Zador. Survey of Single Vehile Fatal ollover Crash Sites in New Mexio. n ransportation esearh eord 819, B, National esearh Counil, Washington, D.., 1981, pp J.. Deaon. dentifiation of Hazardous ural Highway Loations. n ransportation esearh eord 543, B, National esearh Counil, Washington, D.C., 1975, pp J.w. Hall. Guardrail nstallation and mprovement Priorities. l!!, ransportation esearh eord 868, B, National esearh Counil, Washington, D.C., 1982, pp J.W. Hall and J.D. Brogan. ngineering use of ident eords. eport FHW-NMDO-84-1, FHW, U.S. Department of ransportation, May F.M. Counil. ident esearh Manual. eport FHW-D FHW, U.S. Department of ransportation, he 1983 nnual eport on Highway Safety mprovement Programs. Offie of Highway Safety, FHW, U.S. Department of ransportation, valuation of Criteria for Safety mprovements on the Highway. oy Jorgensen & ssoiates, Gaithersburg, Md., J.. Laughlin. Methods for valuating Highway Safety mprovements. NCHP 162. B, National esearh Counil, Washington, D.C., pp. he opinions, findings, and onlusions are those of the authors and do not neessarily reflet the views of the raffi Safety Bureau. Publiation of this paper sponsored by Committee on raffi eords and idents nalysis.