62 TRANSPORTATION RSARCH RCORD 1281 Rral Aident Rate Variations with Traffi Volme J. W. HALL AND 0. J. PNDLTON The natre of the relationship between horly traffi volmes and horly aident rates on rral highways in New Mexio was examined. The data base onsisted of traffi volmes at 44 permanent ont stations and 3 years' aident experiene on 10-mi roadway setions srronding these stations. The highest aident rates orred dring hors with the lowest traffi volmes. Over the range of onditions examined, aident rates dereased with inreasing traffi volmes and with inreasing volme/apaity ratios. However, bease of the moderate traffi volmes on these roadways, it was not possible to determine the effet on the aident rate as horly traffi volmes approah apaity. In rral states sh as New Mexio, frther stdy of this isse shold fos on higher-volme loations fond in rban areas. For over 50 years, analysts have attempted to identify the relationship between traffi volmes and aident rates. Knowledge of this relationship, opled with information on the apaity of the highway setion, wold allow a planner to estimate the safety impliations of projeted traffi growth and potential improvements in highway apaity. With some notable exeptions, previos researh has fond that the rate of traffi aidents on setions of roadway, as measred by the nmber of aidents per million vehile miles (mvm) of travel, inreases with inreasing traffi volme. However, the data spporting this relationship are highly sattered and are sensitive to the design and operational featres of the highway. The impliations of the existene of a definite relationship between traffi aident rates and the ratio of rrent or projeted traffi volme to apaity (vie) are qite signifiant. It might be expeted, for example, that above some toff vale, the rate of aidents wold intensify with an inrease in vie ratios. If this point and the natre of the relationship old be established, then a ompetent analyst old predit the safety impliations of a growth in traffi volme or a hange in highway apaity. This ability wold be extremely benefiial to an ageny attempting to selet an optimal program of highway improvements nder the onstraint of a limited bdget. It old also alert engineers and planners that partilar roadway setions are approahing onditions that might ase roadway safety to beome degraded. As a pratial matter, the relationship between aide.nt rates and vie ratios is inherently diffilt to establish. It is generally agreed that the reporting of traffi aidents is not omplete, althogh there may be ways to address this problem (J). In addition, the qality of aident data reporting, espeially with respet to rash loation, is sspet. Frthermore, J. W. Hall, Civil ngineering Department, University of New Mexio, Albqerqe, N.Mex. 87131. 0. J. Pendleton, Texas Transportation Institte, Texas A&M University, Division 7, CTTI Tower, College Station, Tex. 77843-3135. the reliability of traffi volme data is ertainly sbjet to qestioning. In other words, most state data systems are not able to provide aident or volme data in a omplete, arate format, whih is neessary for this type of analysis. A stdy was ondted to determine if the traffi reords maintained in New Mexio old be sed to identify a tentative relationship between aident rates and vie ratios on rral highways. BACKGROUND The orrene of a traffi aident reflets a shortoming in one or more omponents of the driver-vehile-roadway system. Typially, the orretion of problems assoiated with one of these omponents is sffiient to keep an aident from orring. Ths, althogh many individal fators may ontribte to an aident, improvements to highways an have a signifiant effet in reding rash experiene. It is prdent for a highway ageny to operate a program designed to identify and orret loations or sitations with nsally high aident experiene. Sh a program is hampered, however, by the partially random natre of rash orrene and the qality of available aident data. For this reason, there is onsiderable interest in srrogate measres, sh as vie ratios, that may help in illenlifying problems. One of the earliest stdies of the relationship between traffi volmes and aident rates was reported in 1937 by Yeh (2). The stdy examined the aident rates on setions of twolane roads in New Jersey and related these rates to the average daily traffi (ADT). Yeh (2) onlded the following: It is interesting to note the rather definite relationship between daily traffi volme and the nmber of aidents per million vehile-miles. In other words, as the average daily traffi inreases, aident experiene, on a million vehile-mile basis, likewise inreases to approximately 7,000 vehiles per day, whih is onsidered by many athorities to be the apaity of a two-lane highway. Beyond that point, primarily bease of ongestion and the resltant derease in speed and flexibility of movement dring heavy traffi hors, there is a gradal derease in the aident rate, despite the inrease in traffi. Althogh Yeh (2) does not doment the original data set sed in the analysis, he does present a graph of the relationship between aident rates and ADT. Figre l shows Yeh's data along with his best-fit line, whih shows the relationship between the two variables. The data dispersion in Figre l is a reasonable representation of the amont of satter reported sbseqently by other researhers.
Hall and Pendleton 7 6 GI a. 5 -Ill 4 IC 3 GI.;.. 'ti ( 2 0 2 4..... Average Daily Traffi 6.. 8 10 (Thosands) FIGUR 1 Relationship between aident rates and ADT (2). Lndy (3) analyzed traffi aident experiene on 659 mi of for-, six-, and eight-lane California freeways from 1960 throgh 1962. For eah type of faility, aident rates inreased with inreasing ADT. In addition, Lndy observed that, for any given ADT, the for-lane freeway had the highest aident rate and the eight-lane freeway had the lowest rate. Lndy developed a set of regression eqations of the following form: Aident rate = a + ~ * ADT a,~ 0 Althogh the vales of a are similar for the three types of freeways, the vales of~ derease from for- to six- to eightlane failities. A disssion of Lndy's paper by Byington (3) noted that it may be appropriate to inorporate segment length in an aident predition model. Another disssion by Champagne (3), sing Lndy's data, showed a onsistent inrease in aident rates as the vie ratio inreased. An NCHRP stdy ( 4) examined the aident experiene on roadway setions in three states in relation to traffi volmes as well as an assortment of highway design and operational featres. The stdy fond that the single-vehile aident rate dereased with inreasing ADT, whereas the mltiple-vehile aident rate inreased with inreasing ADT. However, the researhers were nable to define a relationship between total aident rate and ADT bease they fond "in some instanes the rate inreased with inreasing ADT and in other instanes the rate dereased with inreasing ADT." A 1969 state-of-the-art review (5) examined these and other artiles in the tehnial literatre and onlded the following: It has been shown that aident rates tend to inrease sharply with average daily traffi. Whether the levels of servie whih are being inreasingly sed to desribe highway operation are related to the aident rate is nknown. Definition of this relationship might provide a sefl method to predit ftre operational effetiveness of roadway designs. The previosly ited stdies fosed on the relationship of aident rates to ADT. In the mid-1960s, Gwynn (6) examined the horly aident experiene on a 3.8-mi setion of a forlane divided highway in New Jersey. Dring the 5-year stdy period (1959 throgh 1963), the ADT on this setion was approximately 64,000 vehiles per day (vpd), for a total travel of 445 mvm. Diretional horly volmes on the setion ranged from 200 to 3,300 vehiles per hor (vph). Gwynn fond the highest aident rates (41mvm to 61mvm) dring hors in the low-volme ranges. Rates dereased to approximately 21mvm dring those hors in whih the diretional volme was between 1,000 and 1,800 vph. At even higher horly volmes, rates inreased to 41mvm to 51mvm. Sbjet to some satter in the data, a plot of the aident rates as a fntion of horly volme assmed a U shape. Dart (7) stdied 5 years of aident experiene (1962 throgh 1966) on 1,000 mi of rral highway in Loisiana in an attempt to determine whih geometri variables ontribte the most to traffi aidents. As part of this effort, he analyzed the relationship between aident rates and a traffi volme ratio, with the latter defined as the ratio of peak-hor volme to the servie volme at Level of Servie (LOS) B. Callation of the traffi volme ratio reqired that speifi design and operational featres be olleted for eah of the 246 stdy setions. The ratio varied from a low of 0.04 to a high of 2.12. Dart reported that, for the traffi volme ratio range of 0.1 to 1.2, whih presmably aonted for the blk of the stdy setions, the aident rate exhibited a ontinos inrease with inreasing traffi volme ratio. Taylor (8) desribed a hazardos rating formla inorporating both aident and nonaident measres. The model inlded the vie ratio "bease it inorporates the basi volme information, and yet normalizes these data to ompensate, to some extent, for the nmber of lanes, traffi mix, ontrol devies, et." Bease the model was designed to serve as a sreening tool for identifying hazardos loations, Taylor minimized the data reqired for its appliation by defining the vie ratio in a nontraditional manner as ADT divided by 24 times horly apaity. On the basis of this definition and tehnial inpt from a panel of experts, a vie ratio of more than 0.52 was jdged as hazardos. As a point of omparison, the high-volme setion in New Jersey stdied by Gwynn wold by this definition have a vie ratio of approximately 0.44. In 1980, Orne (9,p.20) desribed some preliminary efforts to examine the relationship between traffi aidents in Mihigan and the atal traffi volmes at the time of the aident. This approah, whih had also been sggested by Gwynn (6), is hampered to a onsiderable extent by the availability of reliable traffi volme data at aident sites. More reently, researhers in Greee (JO) examined the relationship between traffi aident rates and vie ratios on an 11.2-mi setion of a for-lane ndivided toll road. Bease the faility was a toll road, with aess available only at the ends of the setion, it provided reliable traffi volme data. Althogh traffi volmes were not speifially indiated, an analysis of the nmber and rate of aidents for the 89-month stdy period sggested that ADT vales were in the range of 17,000 to 18,000 vpd. The analysis onentrated on 9.3 mi of the road, whih experiened 517 aidents, for an average aident rate of approximately 1.21mvm. Using the tehniqes of the 1985 Highway Capaity Manal (HCM). (JJ), he researhers allated the apaity of eah of the 15 stdy setions and determined the vie ratio and level of servie at the time of eah aident. For LOS A and B operating onditions, whih aonted for 81 perent of the travel and 71 perent of the aidents, the aident rate was approximately 63
64 TRANSPORTATION RSARCH RCORD 1281 10 perent below the average rate. At LOS C, the rate was lose to the overall average. However, at vie ratios greater than 0.65, orresponding to LOS D throgh F, the aident rates inreased sharply. For LOS F, a fored-flow ondition, the aident rate was approximately 2.91mvm. The reslts of the Greek stdy are enoraging, bt they mst be viewed irmspetly. At the most basi level, how well the proedres of the 1985 HCM apply to the driving environment in Greee is not known. Also, the aident stdy ondted over the 7-year period was sbjet to hanges in the environment, even thogh there were no hanges to the roadway itself. In addition, the aident sample size was relatively small in omparison with several of the other stdies ited previosly. Perhaps the most extensive evalation of this sbjet was a 1967 to 1975 stdy (12) of eight setions of for-lane interrban road in Israel. The stdy, whih was limited to fatal and injry aidents on weekdays, ontained thorogh analyses both of single- and mltiple-vehile rashes. The stdy inlded two primary proedres: (a) a time-seqene analysis for eah roadway setion and (b) a ross-setion analysis on an annal basis. Single-vehile aident rates were extraordinarily high for flow rates below 250 vph. The mltiplevehile aident rates were more diverse, with half the sites showing a sbstantial inrease in rates for flow rates greater than abot 900 vph, and the remaining sites exhibiting little hange with inreases in horly traffi volmes. When the two rash types were ombined, the reslts were dominated by the data for mltiple-vehile rashes. More speifially, those stdy setions that enompassed a broad range of traffi volmes had a U-shaped relationship when aident rates were plotted as a fntion of horly volme; the minimm rate orred near 500 vph. The remaining for sites, three of whih did not have horly volmes in exess of 1,000 vph, did not show an inrease in aident rates as horly volmes inreased. The tehnial literatre offered some sefl gidane for the development of the methodology to be sed in this stdy. It appears valid to assme that there is a relationship (albeit nknown at this time) between traffi aident rates and traffi volme. Bease a variety of driver, vehile, and roadway fators ontribte to rash orrene, it wold of orse be nrealisti to expet traffi volme variations to explain a large share of the differenes in aident rates. The data satter observed in these stdies is a partial refletion of this problem. The literatre sggests two approahes for ahieving a sitable sample for stdying this isse. One proedre is to selet a omparatively short setion of road and to stdy its volme and aident harateristis for an extended period of time. By sing a single, reasonably homogeneos site it is possible to thoroghly monitor the parameters of interest. On the other hand, the site hosen may not be representative of the remainder of the road system. Frthermore, aident stdies that ontine for an extended period are not able to aont for system-wide hanges in the driver, vehile, or environment variables. An alternate approah is to se a larger nmber of stdy setions. A major advantage of this tati is that it enompasses the harateristis of a broader setion of the roadway system and offers the potential for ondting the stdy over a shorter period. However, it is diffilt to ensre that the setions are omparable with respet to those design and operational harateristis that an affet rash orrene. All things onsidered, it appears that the se of mltiple setions of onstant length is the sperior stdy proedre. STUDY DSIGN The essential reqirement for a stdy of the relationship of aident rates to vie ratios is reliable information on traffi aidents, horly traffi volmes, and fators that inflene highway apaity. New Mexio maintains a ompterized aident reord system that inldes approximately 50,000 aidents reported annally throghot the state. The system allows aidents to be seleted by rote and loation and lassified aording to harateristis of interest to this stdy. Traffi volme data pose a somewhat more serios problem. In New Mexio, rral traffi volmes are olleted at 50 permanent ont stations, and short-term onts are ondted throghot the state on a rotating basis. A reent ritiqe by the New Mexio State Highway and Transportation Department identified valid onerns abot the reliability of data from the short-term onts. Althogh the root problems that affeted the data qality have been orreted, it will take 1 or 2 years to develop a sitable traffi volme data base for the rral state highway system. Therefore, this stdy relied on traffi volme data from the permanent ont stations; these data are reportedly reliable. Another prinipal data need is the information neessary to allate apaity. Roadway setions at the stdy sites inlde two-lane roads, a ople of for-lane divided highways, and for-lane freeways. With a small sarifie in aray, all an be assmed to have 12-ft lanes and lear roadsides. When neessary for apaity allation, a reasonable jdgment an be made regarding driver familiarity with the roadway. A major fator affeting apaity is the ombination of terrain and traffi omposition, espeially of heavy trks; the.se parameters are available for some of the sites and old be measred at the others. For two-lane sites, the extent of the road sbjet to passing restritions is neessary for the allation of apaity; nfortnately, this information is not readily available in existing reord systems. ASSMBLING TH DATA BAS In aordane with Byington's reommendation (J), the stdy plan involved the identifiation of onstant-length segments arond eah of the permanent ont stations. The length hosen for this prpose was 10 mi. A review of maps and the roadway inventory file fond that six of the permanent ont loations were on setions of road with signifiant nmbers of aess and egress points, to the extent that the volme onted at the permanent ont station may not reflet the volme on the entire 10-mi setion; therefore, these sites were deleted from the analysis. At three additional sites, the permanent ont station was on a setion bonded by a pair of major intersetions or interhanges less than 10 mi apart; in these ases, the setions were shortened to 5 mi. In the remaining 41 ases, it was possible to identify a 10-mi setion
Hall and Pendleton 65 arond the permanent ont station that was reasonably homogeneos and free of signifiant aess points. The reslting 44 stdy sites, with a length of approximately 425 mi, had a ombined annal travel of 960 mvm (on the basis of the FY 1988 travel volmes). The average daily traffi at individal sites ranged from 500 to 26,200 vpd. The next step in reating the data base involved a searh of the ompterized aident reords to identify all reported traffi aidents (fatal, injry, and property damage only) on these 44 setions for the 3-year period from 1985 to 1987. The searh identified a total of 2,006 aidents, for an overall aident rate of 0. 70/mvm. The aidents were sbseqently ategorized by hor of the day and ollision type (single- verss mltiple-vehile rashes). DATA ANALYSIS In an effort to establish a tie to the reslts reported in the tehnial literatre, the initial phases of the analysis examined the volme and aident experiene for the entire data base. Figre 2 shows the horly variation in total annal vehile miles of travel for the 44 stdy setions. Peak travel ors in the hors of 4:00 to 5:00 p.m. and 5:00 to 6:00 p.m., both of whih aont for approximately 7.4 perent of the ADT. Bease of rral travel harateristis, there is no disernible morning peak bt rather a gradal inrease in horly travel from abot 7:00 a.m. throgh the late afternoon. At the other extreme, the hors beginning at 2:00, 3:00, and 4:00 a.m. eah aont for abot 0.9 perent of the daily travel. Figre 3 shows the horly distribtion of aidents for the 3-year stdy period. The most striking featre of this graph is that it exhibits sbstantially less horly variation than was evident for the traffi volmes. The maximm nmber of aidents (127) orred in the hor beginning at 6:00 p.m. (6.3 perent of the daily total), followed by 5:00 p.m. and 7:00 a.m., eah with abot 5.6 perent. By ontrast, the lowest nmber of aidents (54) orred between 4:00 and 5:00 a.m., aonting for 2. 7 perent of the total. When horly aident rates are allated, therefore, the peak rates do not or dring those hors with the greatest nmber of aidents bt rather dring those hors with the lowest volmes of traffi. As shown in Figre 4, the peak aident rate (3.2/mvm) orred between 2:00 and 3:00 a.m.; all the hors from 11:00 p.m. to 60 'i 40 ( 20 FIGUR 2 ly travel variations. 140 "' ~105,, (,. 70? Ill) I I/) Ill) 35 '11 0...._ FIGUR 3 Traffi aidents, 1985 throgh 1987. 3.2.---------------------------------~. 2.4 QI. ti -; 1.6 - QI :2 0.8 ( 0 2 4 6 8 10 12 14 16 18 20 2'2 24 FIGUR 4 Traffi aident rates, 1985 throgh 1987. 5:00 a.m. had aident rates that were more than twie the overall daily average of 0.7/mvm. The finding that aident rates are high dring the hors of darkness is not nexpeted. However, the relatively onstant daytime aident rate was not antiipated. For the relatively high-volme hors between 6:00 a.m. and 7:00 p.m., only the hors beginning at 7:00 a.m. (0.78/mvm) and 6:00 p.m. (0.74/mvm) had aident rates greater than the average for the entire day. Figre 5 shows the same information inlded in Figre 4 bt in a manner that highlights the pattern of aident rates verss average horly volmes. Over the range of volmes stdied, the data demonstrate that rral aident rates derease with inreasing horly volmes. Althogh this same pattern emerges in sbseqent analyses, a ase-and-effet relationship was not assmed, espeially between the low traffi volmes and their aompanying high aident rates. However, mentioning two potential explanations for this pattern is appropriate. Single-vehile aidents have been shown (13) to be a serios problem in New Mexio, aonting for approximately 45 perent of the state's highway fatalities. ly aident rates were therefore allated separately for single- and mltiple-vehile aidents. For the entire day, the rates of single- and mltiple-vehile aidents were 0.44/mvm and
66 TRANSPORTATION RSARCH RCORD 1281 3.2 ------------------------------. 2.4 Q. ti -:a 1.6 - ~ 0.8. o.o. _ 0 100 200 300 400 Average ly Volme FIGUR 5 Traffi aident rates, 1985 throgh 1987. 500 0.26/mvm, respetively. The plot of horly aident rates shown in Figre 6 sggests two findings. First, the major omponent of the peaking of aident rates in the late evening and early morning hors is bease of single-vehile aidents; this ertainly makes sense, bease the lower traffi volmes dring these hors lessen the opportnity for mltiple-vehile onflits. Seond, the rates of single- and mltiple-vehile aidents dring daylight hors are similar; with the exeptions of the hors beginning at 1 :00 and 2:00 p.m., the singlevehile aident rate is always greater than the mltiplevehile aident rate. Nmeros harateristis differ between daytime and nighttime operation, espeially on rral highways. 0Il' the one hand, enforement presene is mh lower on these roads at night, ths reding the likelihood that a reportable aident will be reported. On the other hand, there is ample evidene that the driving environment hanges dring the hors of darkness. Visibility is obviosly worse, and drivers dring the low-volme hors are more likely to be tired or impaired. The role of single-vehile aidents and the omplexities of nighttime operation vis-a-vis horly volmes in ontribting to the sbstantial aident rate variations at low flow rates is beyond the sope of this projet, whih is more onerned with the relationship at high volmes. Despite the pattern shown in Figre 5, it may be that, over the volme range of interest to the analyst, aident rates on rral New Mexio roads are independent of horly volmes. A flaw is inherent in the reslts shown in Figres 2-6. Speifially, the 44 stdy sites inlde a mix of two-lane and for-lane roadways, with obviosly different volmes and apaities. This mixtre of the two types of sites tends to mask the effets both at low and high volme levels. To resolve this shortoming, the data base was divided into two sets: (a) a grop of 24 low-volme setions with ADT vales less than 4,500 vpd and (b) a grop of 20 high-volme setions with ADT vales greater than 4,500 vpd. With two exeptions, this groping divided the data base into two-lane and forlane failities. Graphs analogos to Figres 4 and 5 were prepared for eah grop. The 24 low-volme sites had an average ADT of approximately 2,200. The total nmber of aidents was 567, for an average aident rate of 0.97/mvm. The hor beginning at 6:00 p.m. had the highest nmber of aidents (55); on the basis of a mean volme of 130 vph, the aident rate dring this hor was 1.61/mvm. With the exeption of this hor, aident rates for the hors from 6:00 a.m. to 7:00 p.m. were all below the daily average. As shown in Figre 7, peak aident rates orred dring the early morning hors. The pattern in this figre is similar to that shown in Figre 2, althogh the rates are sbstantially higher. In Figre 8, the aident rates are plotted as a fntion of average horly volmes, whih varied from 13 at 3:00 a.m. to 172 at 4:00 p.m. As a oarse approximation, the two-diretional apaity of these setions old be estimated at 1,800 vph; the absissa in Figre 8 therefore orresponds to a range of vie ratios from 0 to approximately 0.10. The trend of dereasing aident rate with inreasing average horly volme is evident, with the only signifiant exeption orring at 6:00 p.m. The 20 high-volme sites, with an aggregate length of 185 mi, had an average ADT of nearly 11,300. On the basis of the total of 1,439 aidents dring the period 1985 throgh 1987, the average aident rate was 0.63. In ontrast to the two-lane roads, the peak nmber of aidents (87) orred at 7:00 a.m., when the aident rate was 0.75. The horly variation in aident rates shown in Figre 9 displays the familiar pattern of below-average rates dring the daytime and above-average rates at night. However, peak aident rates are only half the vales for the low-volme setions. 3 ::: ::: Q. ti -IV 2 - i:::j ; 0 "' Single Yeh " Mltiple Yeh FIGUR 6 Single- and mltiple-vehile aident rates. 6.0.-----------------------------------. ::: ::: 4.8 Q. 3.6 ti IV 2.4 i:::j ; 1.2 FIGUR 7 ADT 45 00 vpd Aident rates on low-volme setions.
Ha ll and Pendleton 67 6.0 :& 4.8 ftl II 1:1. 3.6.; IC 2.4 II "' ADT 4600 vpd :; 1.2 0.0 0 30 60 90 120 150 180 :::&.; 3.0 ~ 2.0 ftl IC 1.0 ADT 4500 vpd II. "' :; 0.0 0 180 360 540 720 900 Average ly Volme FIGUR 8 Aident rates on low-volme setions. FIGUR 10 Average ly Volme Aident rates on high-volme setions. :::& 3.0 ~ 2.0.; IQ IC 1.0 II "' :; ADT 4600 vpd 0.0 FIGUR 9 Aident rates on high-volme setions. The relationship between aident rates and average horly volme is shown in Figre 10. In this figre, the maximm peak horly volmes of 821 vph (at both 4:00 and 5:00 p.m.) orrespond to a vie ratio vale in the range of 0.15 to 0.20. Over the range of moderate traffi volmes shown in this figre, there is no indiation that aident rates inrease as the vie ratio inreases. On the basis of the onditions stdied and the relationships displayed in Figres 2-10, it is possible to draw the following intermediate onlsions: 1. Aident rates tend to be below the daily average dring the daylight hors and are sbstantially greater dring the low-volme hors of darkness. 2. Over the range of average horly volmes stdied on rral New Mexio highways, aident rates derease with inreasing volmes. However, when the very low-volme hors are ignored, it may be that aident rates are independent of flow rates. 3. The major differene between daytime and nighttime aident rates appears to be attribtable to the nsally high single-vehile aident rates dring hors of low traffi volme. In addition, the single-vehile aident rate is, in most ases, higher than the mltiple-vehile aident rate. 4. Althogh New Mexio's permanent ont stations are loated at representative loations on rral two-lane roads, they do not embrae a fll range of horly volmes and vie ratios. The highest average horly volme at any of the twolane setions is 365 vph, orresponding to a vie ratio of approximately 0.20. Bease of the limited latitde of vie ratios, it is not possible to se the New Mexio data base to develop a meaningfl relationship for these failities over the fll range of vie ratios. 5. Aident rates on New Mexio freeways and other forlane highways are abot 65 perent of those on two-lane roadways. However, bease of their signifiantly higher volmes and marginally higher vie ratios, the mltilane highways provide a better opportnity for establishing the relationship between aident rates and vie ratios. These findings are not partilarly srprising in hindsight. However, a better nderstanding of this sbjet reqires setions of road with higher vie ratios. The next task, therefore, involved a more detailed stdy of the highest volme forlane setions in an effort to extend the relationship shown in Figre 10. HIGH-VOLUM FOUR-LAN ROADWAYS The trade-offs between the long-term examination of highvolme setions and the shorter-term evalation of a more extensive portion of the roadway system were desribed previosly. Althogh overall preferene was given to the examination of a larger nmber of setions, it was neessary to fos on the highest volme sites if information was to be developed regarding the pper ranges of the vie ratio. It was reognized at the onset that an individal 10-mi setion, even thogh examined for a 3-year period, wold exhibit signifiantly more variation in volme and aident parameters than the larger grop of high-volme setions shown in Figres 9 and 10. Nevertheless, this orse of ation was the only one feasible within the onstraints of the existing data base. As a reslt, the three mltilane, permanent ont stations with the highest ADT vales were examined in greater detail. Two of these setions are on the Interstate highway system, whereas the third is on a federal-aid primary system. The harateristis of these setions are presented in Table 1.
68 TRANSPORTATION RSARCH RCORD 1281 TABL 1 SUMMARY OF THR HIGH-VOLUM SITS Station 99 Rote s 84 Milepost 168.61 1988 Volmes ADT 26214 Ave Min Hr 54 Ave Max Hr 2354 %HC (1987) 4 Highest ly Volmes 50th Highest 2817 30th Highest 2869 10th Highest 3157 Highest 5064 Capaity (Approx.) Design Speed 60 PHF 0.9 Driver Commter Terrain Rolling Capaity/dir 2864 Capaity 5700 V/C Ratios Ave Min Hr 0.01 Ave Max Hr 0.41 50th Highest 0.49 30th Highest 0.50 10th Highest 0.55 Highest 0.89 270 520 I-40 I-10 176.07 155.25 19226 19185 219 204 1389 1345 25 22 1605 1613 1640 1642 1740 1740 1978 1951 70 70 0.9 0.9 Weekend Weekend Rolling Level 1851 2808 3700 5600 0.06 0.04 0.38 0.24 0.43 0.29 0.44 0.29 0.47 0.31 0.53 0.35 As shown in Table 1, the ADT on these high-volme setions ranged from 19,200 to 26,200 vpd. On the two Interstate setions, more than 20 perent of the traffi onsisted of heavy ommerial (HC) traffi, whereas on the divided for-lane rral arterial, only 4 perent of the traffi was HC. This partilar harateristi, along with the natre of the terrain, has a signifiant effet on highway apaity. In general, d;-ivers on the Interstate setions are relatively nfamiliar with the roadway, whereas those near Station 99 tend to be ommters, whih has an effet on highway apaity. Data are also shown for seleted highest horly volmes. The highest horly volme is the two-way traffi dring one of the 8,760 lok hors (e.g., 4:00 to 5:00 p.m.) that was the highest for the entire year. A similar definition applies to the 10th, 30th, and 50th highest hors. As indiated by the data, the setion on US-84 has some rather extreme peaking harateristis, with the vie ratio at the highest hor being 0.89; by ontrast, the omparable figres on the Interstate setions are 0.35 and 0.53. Some level of stability is reahed, however, when the 30th highest horly volme, a vale often onsidered in highway design, is examined. At this point, the vie ratios range from approximately 0.30 on I-10 to 0.50 on US-84. As shown in Table 1, vie ratios hange only slightly between the 30th and the 50th highest hors. The data reveal that even if a stdy is restrited to the highest horly volmes at the highest ADT sites, vie ratios at New Mexio's rral permanent ont stations rarely exeed 0.5. From this stdy's perspetive, it is diffilt to evalate the extreme points on the vie rve. If, for example, the highest hor of the entire year at Station 99 (5:00 to 6:00 p.m. on the last Monday in Agst) had no aidents dring 1985 throgh 1987, the reslting aident rate wold be O.OOlmvm. On the other hand, with one aident dring this partilar hor, the aident rate wold be 6.581mvm; intermediate rates are not possible. In fat, none of the 147 aidents of Station 99 dring 1985 throgh 1987 orred between 5:00 and 6:00 p.m. on the last Monday in Agst. Looking at a broader slie of time, all of the hors from noon to 7:00 p.m. on the last Monday in Agst were among the year's 30th-highestvolme hors. Dring the 3-year stdy period, there were no aidents dring any of these hors on the last Monday in Agst. Contining one step frther, there was only one aident dring these 7 hr on any Monday in Agst, yielding an aident rate of 0.45 verss an overall rate of 0.51 at this site. Finally, dring all hors of the 13 Agst Mondays in 1985 throgh 1987, there were two aidents on this setion, for a rate of 0.44. The onlsion from this miroanalysis of the highest volme setion is that evalation at this level an prode sporadi and inonsistent reslts. As a ompromise between the earlier analysis that groped 20 for-lane failities and the preeding analysis that examined speifi hors with the highest volme, an intermediate approah was taken. The three high-volme sites haraterized in Table 1 were examined individally and olletively. Figres 11-13 show the relationship between aident rates and traffi volmes at the three sites. Althogh the best-fit lines in these figres are generally similar to those seen in the earlier graphs, the data exhibit a higher degree of satter bease, with smaller amonts of travel at a speifi site, a differene of one or two aidents an ase a sbstantial
Hall and Pendleton 69 : 4.5 ICI. or fti 3.0 ~ 1.5 na1a- ot 9.02 2a1n t.t 001 Mowt1 2.8 -----------------. :: :: 2.1 ICI. or -; 1.4 ~ 0.7 - - - -r- 0,0... 0 450 900 1350 1800 Average ly Volme FIGUR 11 Aident rates on US-84. FIGUR 14 Three high-volme ont stations. : :: 2.4 ICI. or -; 1.6 ~ 0.8 0.0.. FIGUR 12 Aident rates on 1-10. 2.0 : : 1.5 ICI. or ft! 1.0 ~ 0.5.. 0.0 FIGUR 13 Aident rates on 1-40. hange in the aident rate. These wide variations are moderated to some extent when the aident and travel data from the three sites are ombined. As shown in Figre 14, data from the three sites allow the average aident rate to be evalated for average horly volmes as high as 1,700 vph. As a rogh approximation, this average volme level orre- sponds to a vie ratio of 0.34. The inevitable onlsion is that, over the range of volmes and vie ratios fond at the highestvolme rral for-lane roadways, aident rates derease sharply as average horly volmes inrease to abot 900 vph (vie = 0.18). Between horly volmes of 900 and 1,700 vph, aident rates level off at approximately 70 perent of the overall average aident rate. It is not possible to determine from these data if a frther inrease in traffi volmes wold be aompanied by an inrease in aident rates. CONCLUSIONS AND RCOMMNDATIONS An attempt was made to determine if a relationship exists between horly aident rates and the ratio of traffi volme to apaity. Knowledge of any sh relationship wold help engineers and planners assess the safety impliations both of projeted traffi growth on existing highways and of highway improvements designed to inrease apaity. The tool old therefore be a valable spplement to rrent analysis tehniqes. This objetive was not ahieved bease peak traffi volmes on the rral highways examined are rarely as high as 50 perent of apaity. Althogh the variation in aident rates at low flow rates was domented, onlsions were not reahed for the potentially more ritial ase of high vie ratios. For sh highways, it is probably impossible to resolve this isse. The highest aident rates were shown to or at low volmes. This finding was expeted bease these onditions or at night, when aident rates are known to be higher. Bt it also raises a related isse onerning the type of aident that shold be inlded in the aident rate. It was assmed that vehile onflits, and in trn aidents, wold inrease as ongestion inreased beyond some threshold vale. However, mltiple-vehile rather than single-vehile aidents wold inrease. Figre 6 showed that the peaking of aident rates dring the low-volme hors of darkness was ased primarily by the orrene of single-vehile aidents, althogh few rashes of this type an be attribted to ongestion. Frther stl!dy of this sbjet might ahieve better reslts if restrited to mltiple-vehile rashes.
70 TRANSPORTATION RSARCH RCORD 1281 The New Mexio State Highway and Transportation Department has implemented a set of proedres that over time will improve the qality of traffi volme data olleted at short-term ont loations. The availability of a more extensive data base, possibly inlding sites with higher vie ratios, might provide an opportnity for frther stdy of this topii;;. The other major variable onsidered in this stdy was traffi aident experiene. The tehniqes sed highlight the importane of high-qality aident data, espeially with respet to rash loation. Signifiant progress was made on several major omponents of the problem. General trends were identified, and data problems that need to be addressed before frther progress an be made were highlighted. Two preliminary stdies, desribed in the projet report (14), examined the feasibility of applying the tehniqes disssed herein to rban intersetions and freeway setions. Althogh there are potential data problems with both of these highway elements, they both provide the opportnity to evalate operation at high vie ratios. A proposed stdy will attempt to evalate the relationship between aident rates and vie ratios at high-volme intersetions in the Albqerqe area. ACKNOWLDGMNTS Aess to the aident data sed in this stdy was provided by the Division of Government Researh at the University of New Mexio. Traffi volme data were provided by the New Mexio State Highway and Transportation Planning Brea. The ooperation of these agenies is appreiated. RFRNCS 1.. Haer and A. S. Hakkert. xtent and Some Impliations of Inomplete Aident Reporting. In Transportation Researh Reord 1185, TRB, National Researh Conil, Washington, D.C., 1988, pp. 1-10. 2. A. H. Yeh. Improvements to Rede Traffi Aidents. Pro., ASC, 1937, pp. 1775-1785. 3. R. A. Lndy. ffet of Traffi Volmes and Nmber of Lanes on Freeway Aident Rates. In Highway Researh Reord 99, HRB, National Researh Conil, Washington, D.C., 1965, pp. 138-147. 4. J. K. Kihlberg and K. J. Tharp. NCHRP Report 47: Aident Rates as Related to Design lements of Rral Highways. HRB, National Researh Conil, Washington, D.C., 1968. 5. G. T. Slatterly and D.. Cleveland. Traffi Volme. In Traffi Control 111d Roadway lements-their Relationship to Highway Safely (Rev.), hnpter 2. Al 111tive Safety Fondation, Washington, D.C., 1969. 6. D. W. Gwynn. Relationship of Aident Rates and Aident Involvements with ly Volmes. Traffi Qarterly, Jly 1967, pp. 407-418. 7. 0. K. Dart and L. Mann. Relationship of Rral Highway Geometry to Aident Rates in Loisiana. In Highway Researh Reord 312, HRB, National Researh Conil, Washington, D.C., 1970, pp. 1-16. 8. J. I. Taylor and H. T. Thompson. Identifiation of Hazardos Loations. Final Report. Report FHWA-RD-77-83. FHWA. U.S. Department of Transportation, D e. 1977. 9. D. Orne. Traffi, Aident and Highway Data-State Perspetive. In f1111jli!me11ti11g Hig/11w1y Safety /111pro11m e11ts, ASC, Washington, D.C., 1980. 10. J.M. Frantzeskakis and D. I. Iordanis. Volme-to-Capaity Ratio and Traffi Aidents on Interrban For-Lane Highways in Greee. In Transportation Researh Reord 1112, TRB, National Researh Conil, Washington, D.C., 1987, pp. 29-38. 11. Speial Report 209: Highway Capaity Manal. TRB, National Researh Conil, Washington, D.C., 1985. 12. A. Ceder and M. Livneh. Relationship Between Road Aidents and ly Traffi Flow-I and II. Aident Analysis and Prevention, Vol. 14, No. 1, 1982, pp. 19-44. 13. J. D. Brogan and J. W. Hall. Using Aident Reords to Prioritize Roadside Obstale Improvements in New Mexio. In Transportation Researh Reord 1047, TRB, National Researh Conil, Washington, D.C., 1985, pp. 10-17. 14. J. W. Hall and 0. J. Pendleton. Relationship Between VIC Ra1ios and Aident Rates. Report FHWA-HPR-NM-88-02. FHWA, U.S. Department of Transportation, Jne 1989. Pbliation of this paper sponsored by Commiee on Traffi R eords and Aident Analysis.