Accidents on Rural Interstate and Parkway Roads and Their Relation to Pavement Friction

Transcription

1 Transportation Kentcky Transportation Center Research Report University of Kentcky Year 1973 Accidents on Rral Interstate and Parkay Roads and Their Relation to Pavement Friction Rolands L. Rizenbergs James L. Brchett Cass T. Napier Kentcky Department of Highays Kentcky Department of Highays Kentcky Department of Highays This paper is posted at UKnoledge. researchreports/1113

2 ELIJAH M. HOGGE SECRETARY COMMONWEALTH OF KENTUCKY DEPARTMENT OF TRANSPORTATION FRANKFORT, KENTUCKY 461 BUREAU OF HIGHWAYS JAMES E. GRAY COMMISSIONER October 31, 1973 WENDELL H. FORD GOVERNOR H.2.24 MEMORANDUM TO: J. R. Harbison State Highay Engineer Chairman, Research Committee SUBJECT: Research Report No. 377; Accidents on Rral Interstate and Parkay Roads and Their Relation to Pavement Friction, KYHPR-64-24; HPR-1(9), Part II. Report No. 339, dated September 1972, bore the same title as shon above. The previos report as based on 1969 traffic volmes, 197 accident data, and 1971 skid-resistance measrements. The report no sbmitted is based on the same 1971 skid data bt 197, 71, and72 accident data, and 1971 traffic volmes. It seemed necessary to re-analyze the first trends in terms of more contemporary data. Accident data extends throgh the year before and the year after the skid measrements ere made. The re-analysis confirms the trends previosly obtained. The Interstate and Parkay systems provided test sites on a grand scale. Accident reporting and retrieval procedres provided by the Department of Pblic Safety and AADTs provided by the Division of Planning -- combined ith or skid tests -- made the analysis possible. The same srvey or inventory of skid resistances as sed in the 1974 Interstate Cost Stdy. The report, as did Report 339, addresses straightforardly varios contexts in IM (transmitted by FHWA, Agst 1, 1973) and PPM (May 3, 1972) insomch as it establishes a critical skid resistance for 7-mph highays (SN 7 = 27) and an inventory of the Interstate and Parkay systems (for 1971). The fact that pavement sections fell belo the critical vale is nmistakable. The skid test inventory is inclded as an appendix to the report. The analysis does not indicate ho mch margin of safety above the critical vale shold be provided nor does it, in fact, establish an SN 7 of 27 as a minimm standard. Hoever, recognition of the critical vale, together ith other eighting considerations, ill srely gide the establishment of goals or minimm standards of skid resistance for 7-mph highays. Performance eqations for bitminos and portland cement concrete P,_avements developed in Report No. 331 (also see STP 53, ASTM, April 1973) are given belo: Bitminos Concrete: or, SN 7 = In (Cmlative Traffic x 1 5 ) SN 7 = log (Cpmlative Traffic) E s = 6.6 R =.638 Portland Cement Concrete: AQDRESS RETURN TO: DIVISION OF RESEARCH, 533 SOUTH LIMESTONE, LEXINGTON, KY. 458

3

4 or, SN ln (Cmlative Traffic x 1 5 ) SN log (Cmlative Traffic) E s = 4.3 R =.734 To obtain a 99.9% assrance that these pavement srfaces old not become critically slick, the respective mean vales old have to be not less than E 8. For instance, if e assme that E 8 is constant for each type of pavement, E s is the mean vale of SN 7 to be soght for a given nmber of vehicle passes. Hoever, sing the performance eqations, e obtain: Bitminos Concrete: x 6.6 = log (Cmlative Traffic) Cmlative Traffic = 2.88 x 1 5 Portland Cement Concrete: x 4.3 = log (CmlatiVe Traffic) Cmlative Traffic = 1.4 x 1 5 If there ere 1, velticles per day in the lane considered, the 99.9% confidence limit old be assred for 29 days for bitminos concrete and 1 days for portland cement concrete. To obtain the same degree of confidence throgh 1 or 2 million vehicle passes ith some other type of srface -- assming the same scatter or variability -- its mean vale at 1 and 2 million vehicle passes old have to be E s that is, the mean vales old have to be 47 and 4, respectively. Hoever, the same degree of confidence cold be obtained ith a srfacing material having a mean vale of 3 if E s :;:: I and E s = 3. The performance eqations cited are based on lo-season measrements; for the greater part of the year, the mean vales old be higher. The data ere not stratified or sorted ith respect to sorces and types of sands and limestone coarse aggregates. Extensive project ltistories are needed to establish performance eqations and to flly qalify sand-asphalts (Special Provision No. 59-C) and Open-Graded Plant-Mix Seals as presently being tested on US 31 W north of Elizabethton. Lighteight aggregates and slags shold be evalated in the same ay. Interim decisions addressing portions of IM seem navoidable. In order to provide similar gidance for other primary rotes (specifically those ith posted speed limits of 6 mph), e are sbmitting (as an attachment hereto) a copy of a graph, from a pending report, shoing critical vales of skid resistance measred at 4 mph. A vale of SN 4 = 44 is apparent. Tills is eqivalent to abot 32 (bitminos concrete) and abot 3 (portland cement concrete) at 7 mph. A higher demand for traction is indicated on roads ith ncontrolled access althogh the posted speed is loer. The performance eqation for bitminos concretes on primary rotes other than Interstate and Parkays (also from Report No. 331) differs somehat from the one previosly cited. There as not sfficient data to develop a comparison eqation for portland cement concretes. The bitminos pavement eqation is given belo: or, ln (Cmlative Traffic x 1 5 ) SN 4o log (Cmlative Traffic) E s = 5.9 R =.636 Letting SN 4 eqal to E s, as before,

5

6 x 5.9 = log (Cmlative Traffic) Cmlative Traffic = 1.68 x 1 5 By comparison, the performance of bitminos concrete is a little more favorable on Interstate and Parkay roads than on Primary rotes. Hoever, neither asphalt nor concrete pavements are shon by these analyses to qalify for long-term service nder the conditions defined. If a confidence level of abot 95% had been assmed, e old have obtained the folloing: Bitminos Concrete (Interstate): x 6.6 = log (Cmlative Traffi c ) Cmlative Traffic = 1.39 x 1 6 Portland Cement Concrete (Interstate): x 4.3 = log (Cmlative Traffic) Cmlative Traffic = 3.8 x 1 5 Bitminos Concrete (Primary): x 5.9 = log (Cmlative Traffic) Cmlative Traffic = 3.42 x 1 5 A possible recorse or interim measre, not mentioned in IM , old be to limit et eather speeds to sch a level sch that the stopping distance old not exceed that of a vehicle traveling the posted speed (presmably the dry eather design speed). A relatively high percentage of pavements old meet this condition if 7 mph speeds ere redced to 45 mph and 6 mph speeds ere redced to 4 mph dring et eather. We are nable to cite statistics to verify the effectiveness of this action if it ere to be implemented. Intitively, at least, I believe that sand-asphalts no described by Special Provision No. 59-C ill provide a minimm variability and a minimm loss of skid resistance ith time and traffic. The speed gradient (loss of skid resistance ith increasing speed) may prove to be higher than for Open-Graded Plant-Mixed Seals. Hoever, I expect Open-Graded Plant-Mixed Seals to polish more nder traffic (lose skid resistance) bt maintain a lo speed gradient. Sand-asphalts, as no described in Special Provision No. 22-C, may qalify in time for 6-mph speeds. Additional performance histories are rgently needed. Respectflly sbmitted, SJ# ;/ )./ Jas. H. Havens Director of Research JHH:d Attachments ee 1 s: Research Committee

7

8 8,-----,----,-----,----,----,-----,----,-----, 7 6 >- <(_, O::o 1- :z:: 5 z > Clz -o <)_ <> - <t = 4 o <(. ll::o: 3 I >- ::; ;l: 2.: 6. < 3, e 3POO :::: 12,1..,... 6 C:6. OJ 1 o,: o-----;25----,o:----;,5----4:----4f5----5o SKID NUMBER (4 MPH)

9

10 TECHNICAL REPORT STANDARD TITLE PAGE 1. Report No. 2. Government Accession No. 3. Recipients Catalog No. 4. Title and Sbtitle Accidents on Rral Interstate and Parkay Roads and Their Relation to Pavement Friction Report Dote October 1973 Performing Orgonilolion Code 7. Athorls) R. L. Rizenbergs, J. L. Brchett. and C. T. Napier B. Performing Organilation Report No I Performing Organization Name and Address Division of Research Kentcky Brea of Highays 533 Soth Limestone Lexington, Kentcky 458 Sponsoring Agency Nome ond Address 1. ll Work Unit No. Conlroct or Grant No. KYHPR Type of Report and Period Covered 14. Interim Sponsoring Agency Code 15 Spplementary Notes Prepared in cooperation ith the US Department of Transportation, Federal Highay Administration Stdy Title: Pavement Slipperiness Stdies 16. Abstract Friction measrements ere made ith a skid trailer at 7 mph (31 m/s) on 77 miles (124 km) of rral, for-lane, controlled-access rotes on the interstate and parkay systems in Kentcky. Each constrction project as treated as a test section. Accident experience, friction measrements, and traffic volmes ere obtained for each. Varios relationships beteen et-eather accidents and skid resistance ere analyzed. The expression of accident occrrence hich correlated best ith skid and slip resistance as et-eather accidents per 1 million vehicle miles. Accidents increased greatly as Skid Nmbers (7 mph or 31 m/s) decreased from 27 ± I. Analysis of Peak Slip Nmbers and accident occrrences indicated similar trends Key Words 18, Distribtion Statement Skid Resistance Peak Friction Slip Resistance Incipient Friction Accident Rates Friction Needs Pavement Friction Secrity Clossif. (of this report) 2. Secrity Classif. {of this page) 2J. No. of Pages 22. Price Unclassified L_ Form DOT F 17.7!s-69l Unclassified

11

12 Research Report 377 ACCIDENTS ON RURAL INTERSTATE AND PARKWAY ROADS AND THEIR RELATION TO PAVEMENT FRICTION INTERIM REPORT KYHPR-64-24, HPR-1(9), Part II by Rolands L. Rizenbergs Research Engineer Chief James L. Brchett Research Engineer Principal Cass T. Napier Former Research Engineer Division of Research Brea of Highays DEPARTMENT OF TRANSPORTATION Commonealth of Kentcky in cooperation ith the U. S. DEPARTMENT OF TRANSPORTATION Federal Highay Administration The contents of this report reflect the vies of the athors ho are responsible for the facts and the accracy of the data presented herein. The contents do not necessarily reflect the official vies or policies of the Brea of Highays or the Federal HighaY Administration. This report does not constitte a standard, specification, or reglation. October!973

13

14 INTRODUCTION To assre safe highay travel in et eather, pavements mst be designed to have sfficient and endring skid resistance to enable drivers to perform ormal driving tasks ithot risk of skidding and( or) loss of vehicle control. In emergencies, a driver may be compelled to brake hard and, ith conventional braking systems, may experience skidding regardless of ho skid resistant the pavement may be, Anti-locking brake systems minimize the risks of skidding and permit the driver to retain directional control of the vehicle. Any vehicle ill skid, ith potential loss of control, hen the demand for braking force exceeds the tractive force. As friction (traction) increases, greater deceleration is available, and a drivers chances of avoiding collision or skidding off the road are increased. Ideally, et pavements shold provide as mch traction as dry pavements. In a practical and realistic sense, hoever, the qestion remains as to hat minimm level of friction a pavement shold provide to safegard the pblic from nde hazards associated ith et-eather driving. Little satisfaction derives from merely maintaining a friction level at or near a critical vale. The critical vale, hoever, may serve as a criterion for posting et-eather speed restrictions and for design of srface corses providing a de margin of safety. Investigations elsehere to establish minimm friction reqirements fall into to categories: (I) stdies of driver behavior and, therefore, frictional demands attending driving tasks and (2) analysis of accident data and accident experience as related to paver.. ent friction. Stdies in the first category represent a logical approach bt involve extensive monitoring of representative driver poplations nder realistic roaday conditions and sitations. Interpretations as to hat constittes normal as opposed to emergency reactions or sitations present a problem. Friction factors ths derived cannot easily be related to skid resistance measred ith conventional testers (sch as trailers) operated nder prescribed procedres and conditions of test. Accident rates have been recognized as being higher on et than on dry srfaces; many statistics are available to spport this intitive conclsion. Frthermore, research has shon that accident rates tend to increase as et skid resistance diminies. This relationship is no considered to be intitive and a priori. Hoever, the interaction of many contribting factors sch as roaday geometries, traffic characteristics, driver behavior, etc., together ith ncertainties concerning reliability an. availability of accident data, type of friction measrements, and type of analysis have heretofore obscred relationships beteen accidents and pavement friction. The primary objective of this stdy as to discern a relationship beteen accident experience and pavement friction for rral, for lane, controlled access roads on the interstate and parkay systems in Kentcky. A preliminary report (1) on this phase of the stdy as issed previosly. These highays ere prposely chosen for this initial analysis becase many of the sally confonding variables cold be assmed to have minimal inflence. A similar stdy of other rral rotes is nearing completion. Sbseqent evalations of sch a relationship in conjnction ith economical and technical considerations ill srely gide the establishment of minimm levels of friction. To define a relationship beteen accidents and skid resistance, the effect of all other parameters mst be knon or held constant insofar as possible. By limiting the stdy to rral, for-lane, interstate and parkay facilities, some of the parameters, sch as road geometries, access control, and speed, may, be assmed to remain reasonably constant. Traffic characteristics (volme and density) and pavement srface conditions (et or dry and skid resistance hen et) are, respectively, the regenerative and casative factors. Annal average daily traffic volmes ere obtained for Accident data ere those reported dring the calendar years 197, 1971, and Pavement friction measrements ere made beteen Jne and October 1971 on 77 miles (124 km) of the interstate and parkay systems. Both locked-heel and peak slip resistances- ere measred. Peak slip resistance reaches a maximm hen the rolling velocity of the heel is 8 to 9 percent of the vehicle velocity. This difference in velocity may be attribted to shear strain in the tread rbber and(or) slip beteen the tread rbber and the pavement srface. Hoever, it seems most likely that maximm tractive resistance is achieved hen the shear strain in the tread rbber is greatest and actal slip is least. This peak resistance is often referred to as incipient friction and exceeds the resistance measred by the locked-heel method. In normal driving, the vehicle operates in a pre-slip and cornering mode. Therefore, both locked-heel skid resistance and peak slip resistance at varios speeds.. or some other type of measrement -- may be needed to flly characterize pavements. The measrement(s) hich best correlate(s) ith et-!ather accidents remains to be established.

15 PAST STUDIES In Great Britain, Giles (2} and Sabey (3} noted that the percentage of et-road accidents involving skidding correlated linearly ith skid-resistance (Figre I) as measred ith the British Portable Tester. Minimm friction levels, based in part on this correlation, ere recommended. Those recommendations of minimm skid resistance ere: (l) 55 for tangent, level roads and (2) 65 for crves, intersections, grades and rondabots. McCllogh and Hankins ( 4} stied skid resistance and accidents to set gidelines for srface improvements on Texas highays. They investigated 517 road sections and measred skid resistance ith a trailer at speeds of 2 and 5 mph (9 and 22 m/s). Accidents ere expressed in terms of both fatal and injry accidents per 1 jllion vehicle miles as ell as total accidents (inclding property damage accidents) per 1 million vehicle miles. McCllogh and Hankins also considered measres of accidents in other terms bt indicated three reasons for choosing the final exptessions: 1. Virtally no differences ere observed in the preliminary investigation sing the different measres of accidents -- i.e. total accidents, et-road accidents, or skidding accidents. 2. Classification of accident data so as to obtain the other measres of accidents as time consming, and the nmber of skidding accidents as nreliable becase it as difficlt to determine a skidding accident from available accident statistics. 3. Fatal and injry accidents ere selected to avoid incomplete reporting of accidents since these ere virtally alays reported. From graphs of accidents per 1 million vehicle miles verss coefficient of friction (sch as Figres 2 and 3), McCllogh and Hankins conclded that, even thogh there as a ide scatter of points, the data indicated accidents ere, in general terms, inversely proportional to the coefficient of friction. Similar trends {196 5 Smmer months (April to September) Winter months , (Janary to March, October to December) Correlation coefficient=.9 2 Regression line: y =-1.3 x+112 Percentage of dry road accidents involving skidding oo I I I I I I I I Skid-resistance -mean vale measred each month for eight sites Figre I. Relationship Fond in Great Britain beteen Montly Freqency of Wet-Road Accidents Involving Skidding and the Mean Skidding Resistance in the Month Concerned (from Skidding Accidents, Institte for Road Safety Research, SWOV, The Netherlands, 197). 2

16 >- z. 1 9 > BOO. 7 6 z I z > , _:_.:,. :_:;\:,;i,,::.:ji.: : :? -i::_.:_,.2.,, _ :1 cc id e 1 ts 9 1 COEFFICIENT OF FRICTION AT 2 M.P_H Figre 2. Comparison of Accidents and Coefficient of Friction at 2 mph (9 m/s) (from Reference 4 ) m >- z 4 > > z 4-4 >- 4 I Of I 8 9 LO COEFFICIENT OF FRICTION AT 5 M.P.H Figre 3. Comparison of Accidents and Coefficient of Friction at 5 mph (22 m/s) (from Reference 4). 3

17 and observations ere determined by comparing both fatal and injry accident rates ith coefficient of friction. Based on these findings, McCllogh and Hankins recommended minimm friction levels of.4 at a speed of 2 mph (9 m/s) and.3 at 5 mph (22 m/s). Minimm friction reqirements (Table I), as measred ith the skid trailer (locked-heel) for different mean traffic speeds, ere proposed by Kmmer and Meyer (5). In that stdy, driver behavior and skidding accidents ere correlated ith pavement friction. Hoever, recommended minimm Skid Nmbers (SN) ere based on driver behavior only and ere intended to satisfy tractional demands for normal vehicle manevers encompassing all driving, cornering, and braking manevers performed by a majority of drivers nder normal traffic conditions. Kmmer and Meyer also recommended that friction measrements be made at the mean traffic speed, ths eliminating ncertainties of sing predetermined skid resistance-speed gradients to extrapolate skid resistance vales to other speeds. TABLE I RECOMMENDED MINIMUM INTERIM SKID NUMBERS (From Reference 5) MEAN TRAFFIC SPEED (mph) (m/s) so SKID NUMBER SN b SN c a Skid Nmber measred in accordance ith ASTM E-274 Method of Test. b sn.. Skid Nmber measred at mean traffic speed. c sn4.. Skid Nmber, measred at 4 mph (18 m/s), inclding alloances for Skid Nmber redction ith speed sing a mean gradient of G =.5 SN per mph (1.12 SN per m/s). Mahone and Rnkle (6) stdied 521 road sections and 2,727 accidents on 313 miles (54 km) of interstate highays in Virginia to determine relationships beteen the percentage of et-eather accidents and Predicted Stopping Distance Nmber (PSDN). Friction measrements ere made ith a skid trailer (locked-heel) at 4 mph (18 m/s) and ere converted to PSDN by correlation. Accidents ere expressed as a ratio of et-srface accidents to total accidents (inclding property damage accidents) (Figre 4 ). To reasons ere cited in jstification of the choice of this ratio as a measre of accidents: I. It as ndesirable to be limited to the se of skidding accident statistics, since inadeqate friction cold promote accidents not reported as involving skidding. 2. It is sometimes impossible to determine from accident reports if skidding as a major contribting factor. Mahone and Rnkle observed that, in most cases, the percent of et-eather accidents decreased as the PSDN increased and that, if traffic volme as a factor, apparently the percent of et-eather accidents as loer ith increased volmes. They conclded that on interstate roadwj.ys ith a mean traffic speed of 65 to 7 mph (29 to 31 m/s) the minimm PSDN shold be 42 for the throgh lane and 48 for the passing lane. Moore and Hmphreys (7) stdied 75 hlgh-accident sites, each one-half mile (.8 km) long, in Tennessee. The sites involved 45 accidents in Skid-test (trailer) measrements ere made in 1969, 197, and 1971, and the analysis as based on 4-mph (18-m/s) test data. Percentages of et-pavement accidents ere related to coefficient of friction. They conclded that accident reports do not adeqately indicate hether skid resistance is a significant factor in a particlar accident. Hoever, pavements ith coefficients of friction of.41 and less had almost tice as many et-pavement accidents as srfaces ith higher skid resistances. It as recommended that the minimm acceptable level of skid resistance shold be a coefficient of friction of.4, as measred by the locked-heel skid tester at 4 mph (18 m/s). In instances here speeds and( or) highay facilities ere comparable, minimm levels of friction recommended by different investigators have been in general agreement. For a mean traffic speed of 7 mph (31 m/s), Kmmer and Meyer (5) recommended a Skid Nmber of 46 at 4 mph (18 m/s). Mahone and Rnkle (6) provided a means of converting measred PSDN to SN at 4 mph (18 m/s). Using an assmed speed gradient of.5 SN per mph (1.12 SN per m/s), they recommended 4-mph (18 m/s) SN test vales of 4 for traffic lanes and 47 for passing lanes on facilities 4

18 5 (/) 1- TRAFFIC LANE z LLI 6 PASSING LANE Cl () 4 () <( LLI () cr ::::l (/) I 1- LLI 3 : z LLI () cr LLI ll.. I \ cr\ / I\ / / PREDICTED STOPPING DISTANCE NUMBER v v Figre 4. Comparison of Percent Wet-Srface Accidents ith Predicted Stopping Distance Nmber for Traffic and Passing Lanes (from Reference 6). operating at speeds of 65 or 7 mph (29 or 31 m/s). Moore and Hmphreys (7) sggested a minimm coefficient of friction of.4 (SN = 4) hen measred at 4 mph (18 m/s) for highays ith posted speeds of 5 mph (22 m/s) and higher. McCllogh and Hankins ( 4) recommended a minimm coefficient of friction of.3 (SN = 3) at test speeds of 5 mph (22 m/s), and this is approximately eqivalent to a Skid Nmber of 35 at 4 mph (I 8 m/s) (speed gradient of.5 SN per mph (1.12 SN per m/s)). Unfortnately, McCllogh and Hankins did not measre mean traffic speeds. If mean speeds ere 5 mph (22 m/s), their Skid Nmber old agree closely ith the reconnendation by Kmmer and Meyer, i.e., a Skid Nmber of 37. Driver behavior stdies by the Franklin Institte ere sponsored by the National Cooperative Highay Research Program, Highay Research Board, nder NCHRP Project 1-12(1), Pavement Friction Coefficients in Driving Tasks. A final report on the stdy has not been pblished. Findings shold enhance the nderstanding of minimm frictional reqirements for normal driving tasks nder varios roaday geometries, speeds, etc. 5

19 DATA ACQUISITION AND COLLATION Traffic Volmes Since traffic volmes vary ith time, any measrement of volme not obtained at the time and location of ejch accident old not precisely represent the volme associated ith the accident. In stdies sch as this, hich cover a system throghot a state, that type of volme measrement is highly impractical The measrement of traffic volme hich is generally available biennially is an annal average daily traffic (AADT). The AADT data for 1971 ere sed in these analyses. Friction Measrements Friction measrements ere obtained sing a Srface Dynamics Pavement Friction Tester (Model 965A) developed by the General ;vtotors Proving Grond and manfactred by K. J. La Engineers, Inc., Detroit, Michigan. This skid trailer complies ith ASTM E 274 (8). The measrements represent friction developed beteen a standard tesl tire (ASTM E 249) (9) and a etted pavement. The locked-heel measrements are expressed as Skid Nmbers (SN); incipient or peak friction is expressed as Peak Slip Nmber (PSN). A description of the skid trailer and procedres applicable to the method of test ere presented in a previos report ( JO). Mesremcnts ere obtained dring the smmer of on all rral, for-lane, interstate and parkay rotes in Kentcky having a posted speed limit of 7 mph (31 m/s). Tes\S e re made in!he left heel path only and at one-mile (.4-m) intervals in each Jane; no less than five tests per lane ere made on each constrction project. The basic test speed as 7 mph (31 m/s). Additional tests ere condcted on selected pavements at 4 mph (H rn/s). Comparison beteen the Skid Nmhers obtained at the to speeds are presented in Figre 5. 8 I 7.. :::!!;._,. 6!;t a: 5 ID :::!!; :::> 4 z :>:: (f) 3 CLASS f r.type BITUMIINOUS SN4 =.92 SN E s 4.1 R =.934 W J( V PCC SN 4 = 1.17 SN7 + 9 Es = 2.a R =.954 / Jf :J./ /e..... / / SKID NUMBER AT 7 MPH SKID NUMBER AT 7 MPH Figre 5. Correlation of Trailer Tests Condcted at 4 and 7 mph (18 and 31 m/s) on Bitminos and Portland Cement Concrete Pavements on Interstate and Parkay Rotes. 6

20 Accident Information Accident data ere obtained from State Police Records, compterized, and maintained by the Department of Pblic Safety. All accidents reported dring the calendar years 197, 1971, and 1972 ere analyzed. Information available from the compter files for each accident is detailed in APPENDIX A. A smmary of accidents on rral, for-lane, interstate and parkay rotes is presented in Table 2. Accidents totaled of hich 3 14 occrred d<ing et-srface conditions. From these accident records, many expressions of accident occrrence may be calclated. Rates of et-srface accidents, dry-srface accidents, fatal and injry accidents, and total accidents (inclding property damage accidents) are commonly calclated. Expressions sed in other investiiations have inclded (1) ratio of et- to dry-srface accidents, {2) ratio of et-srface to total accidents, (3) ratio of et-srface, skidding accidents to total accidents, ( 4) et-srface accient:.. per 1 million vehicle miles, (5) total accidents per 1 million vehicle miles, and (6) fatal and injry accidents per 1 million vehicle miles. Test Sections A test section is defined as 11a section of pavement of niform age and niform composition hich has been sbjected to essentially niform ear along its length (8). Almost all constrction projects fit this definition. Inasmch as the direction of travel for a vehicle involved in an accident as not given in the accident reports, sections inclded both directions of travel. There ere 11 test sections. These are smmarized in APPENDIX B along ith!971 AADTs and other relevant data. On rral, for-lane roadays, most traffic travels in the oter lanes (approximately 8-85 percent), and a large percentage of manevers begin or terminate there. The oter lane, left heel-path Skid Nmbers ere averaged to characterize the skid resistance of the test sections. Distribtions of these vales, SN and PSN, for the 11 test sections are exhibited in Figres 6 and 7, respectively. The relationship beteen SN and PSN is shon in Figre 8. Minimm, average, and maximm vales for each test section are presented in APPENDIX B. Milepoints recorded in accident reports ere sed to describe the location of accidents to the nearest tenth of a mile (.16 km). The nmber of et-srface accidents, dry-srface accidents, total accidents, and the ratios of et- to dry-srface accidents and et-srface to total accidents for each test section are presented in APPENDIX B. Rates of et-srface accidents and total accidents, in terms of 1 million vehicle miles (total vehicle miles traveled nder all pavement conditions), ere calclated for each test section. These rates ere based pon the lengths of sections and the 1971 ADT1 s. These vales are also presented in APPENDIX B and pertain to accidents for a 3-year period. APPENDIX C is similar, bt there the accident data spans 6-month periods (Jne throgh November). 7

21 TABLE 2 SUMMARY OF YEARLY ACCIDENT OCCURRENCE (11 TEST SECTIONS) NUMBER OF ACTIDENTS AND FATALITIES TOTAL TOTAL DRY.SURFACE WET-SURFACE TOTAL DRY-SURFACE WET-SURFACE TOTAL DRY-SURFACE WET-SURFACE TOTAL DRY-SURFACE WET-SURFACE ROUTE ACCIDENTS ACCIDENTS ACCIDENTS FATALITIES ACCIDENTS ACCIDENTS ACCIDENTS FATALITIES ACCIDENTS ACCIDENTS ACCIDENTS FATALITIES ACCIDENTS ACCIDENTS ACCIDENTS FATALITIES ; m m B Kentcky ; m m Trnpike ;; 742 m Jad:oon,, ;; Prcha Parkay Ponnyrile 15 12J Parkay Wcrn eo ;; Kentcky Parkay Blegrass ;o,, % 12! Parkay Mont; in,,.12,_ IJO,, Parkay Tlol # I: II 3H H Cl77 i_1ho )_1 B 57.l6g} Lll4 II

22 12 1 z Qa I- I- l!. a: :::>4 z ADT c::::j - 3 C2Z?Z2l 3) SKID NUMBER ( 7 MPH( 31 MIS)) Figre 6. Skid Nmber Distribtion for ll Test Sections of Rral Interstate and Parkay Rotes. 1 r.,a z ti,_. l! - :54 :::> z l:i; 1% 4 45 I%: I? 18 I: 5 AOT = - 3 = I: f:; f:; I: f:; f:; f:; ro: 1-!- f:; f:; f:; I PEAK SLIP NUMBER (7MPH(31M/Sl) 75 8 Figre 7. Peak Slip Nmber Distribtion for II Test Sections on Rral Interstate and Parkay Rotes. 9

23 Figre 8. Relationship beteen Skid Nmber and Peak Slip Nmber at 7 mph (31 m/s) on Interstate and Parkay Rotes. 9 2 PSN= 2.95(SN)-.3(SN) 8 R 2 = C/) - ::!E f() :::r:: a.. ::!E 1- :: (l) ::::ie :::> z a.. _J C/) <l: a E5 = L----L---..J.._-----L---...I SKID NUMBER (7MPH(31 MIS)) 1

24 SKID NUMBERS AND ACCIDENTS Analysis of Test Sections by Cross Classification To aid in determining the relationship beteen different combinations of traffic volme, Skid Nmbers, and accidents, data for test sections ere arrayed as shon in Table 3. Elements of the array are average et-srface accident rates for all test sections ithin Skid Nmber and traffic volme categories. Similar arrays ere prepared for other expressions of accident occrrence, inclding: (I) total accident rate, (2) ratio of et srface to total accidents, and (3) ratio of et to dry-srface accidents. Thes are presented in APPENDIX D. Analysis of the arrays led to the conclsion that the data needed to be stratified ith respect to AADT to better define the relationship beteen accidents and pavement friction. Plots of et-srface accident rate verss Skid Nmber for each test section (Figre 9a d) frther illstrate the need for sorting and groping the data. Reslting relationships beteen accidents and pavement friction involving et-srface accident rate are presented in Figre I Oa d and similarly for other accident expressions in Figres lla-d, 12a-d and 13a-d. Stratification of data by AADT s shoed improved relationship beteen accidents and Skid Nmbers for some accident expressions, bt not necessarily for other expressions. Althogh a trend as indicated for test sections ith AADTs less than 4, vehicles per day, there as not sfficient data to jstify separate analysis of sections ith the lo AADTs. Considerable variability in data remained after elimination of test sections having AADTs less than 3, and 3,5. Hoever, a trend of decreaseing etmsrface accident rates ith increasing Skid Nmbers as mmistakable. Analysis of test section data contined, seeking that expression of accident occrrence relating best ith pavement friction. This as accomplished by taking elements in the arrays as predicted vales. Actal accident occrrences for each test section ere then compared to this 11predicted vale to obtain deviations. This enabled comptation of a coefficient of correlation for each accident expression. The correlation coefficients ranked the expressions in the folloing order: 1. et-srface accidents per 1 million vehicle miles, 2. total accidents per 1 million vehicle miles, 3. ratio of et-srface to total accidents, and 4. ratio of et- to dry-srface accidents. The degree of correlation as not sfficiently encoraging to enable a decisive selection of the best expression. Analysis to determine the relationship beteen accident occrrence and pavement friction as therefore contined sing all for expressions. 11

25 TABLE 3 WET-SURFACE ACCIDENT RATES* NO OF TEST SECTIONS MPH AADT (VEHICLES PER DAY) (3 1 m/s) SKID NUMBER IS IS.I IS *Accidents per I million vehicle miles

26 SYMBOL AOT SYMBOL AOT i 6 z > 5, 2. i 4.t El II ABOVE.a !! [) IOOOD - 11 II ABOVE 3 SKID NUMBER (7MPH (31M/S)) SKID NUMBER (7MPH (31M/S)} Figre 9a. _Test Section Averages: Wet-Srface Accident Rate for 197 Verss Skid Nmber ith AADT Stratification. Figre 9b. Test Section Averages: Wet-Srface Accident Rate for 1971 Verss Skid Nmber ith AADT Stratification. SYMBOL SYIIIBOL AOT II ABOVE.t B ABOVE SKID NUMBER (7MPH (31M IS)) SKID NUMBER (7MPH (31M IS)) Figre 9c. Test Section Averages: Wet-Srface Accident Rate for 1972 Verss Skid Nmber ith AADT Stratification. Fi:re 9d. Test Section Averages: Wet-Srface Accident Rate for 197 throgh 1972 Verss Skid Nmber ith AADT Stratification. l3

27 c 25 W < - 2 > z c <j < Q - oz > 8 < 3 25 < > z 2 z Q <, z A T e:,.- o- 3ooo Ill \ /1\ / \ \ \ \ \ \ SKID NUMBER (7MPH (31MIS)) SKID NUMBER (7MPH (31M/S)) Figre loa. Comparison of Wet-Srface Accident Rate to Skid Nmber ithot Volme Stratification. Figre lob. Comparison of Wet-Srface Accident Rate to Skid Nmber ith Volme Stratification of AADT of 3, < r > 2 z z 4 c < ;. z c o 5 ACT Ll, - Q z 8 < o ;. 5 \ ADT \ 6- o-4ooo \ \ \ \ /!\ \ / \\ \ l j l. _l. _L J. L _j SKID NUMBER (7MPH (31M/S)) SKID NUMBER (7MPH (31MIS)I 45 5 Figre JOe. Comparison of Wet-Srface Accident Rate to Skid Nmnber ith Volme Stratific-ation at AADT of 3,5. Figre!Od. Comparison of Wet-Srface Accident Rate to Skid Nmnber ith Volme Stratification at AADT of 4,. 14

28 ;;;, > c z 3 c, z - 8. c c z o c ;;; 14, 12 I > z c j, c z 8 8 c c z 6 e Q / 35 ) I 4 I I I 158! 45 5 SKID NUMBER (7MPH (31MIS)) SKID NUMBER (7QMPH (31MIS)) Figre lla. Comparison of Total Accident Rate to Skid Nmber ithot Volmne Stratification. Figre llb. Comparison of Total Accident Rate to Skid Nmber ith Volme Stratification at AADT of 3,. c, z 8 m c c z c ADT!::, - Q / / / /./ ;;; c > z, z > c c z 6 c c 4 A DT 6.- o-qooo II-4I-24 c./ / / / SKID NUMBER (7MPH (31 MIS)} SKID NUMBER (1MPH (31M/Sll Figre llc. Comparison of Total Accident Rate to Skid Nmnber ith Volme Stratification at AADT of 3,5. Figre lid. Comparison of Total Accident Rate to Skid Nmnber ith Volmne Stratification at AADT of 4,. 15

29 .3 z f OJO SKID NUMBER (7QMPH(31MIS)) \ \ il A D T \.1 6 -Q-3 II -3QOI-2AOOO S K 1 NUMBER (7 MPH(31MIS)) \ \ \ \ 45 5 Figre l2a. Comparison of Wet-Srface Accidents/Total Accidents Ratio to Skid Nmber ithot Volme Stratification. Figre 12b. Comparison of Wet-Srface Accidents/Total Accidents Ratio to Skid Nmber ith Volme Stratification at AADT of 3,. Q < AO T --35 II \ \ ), e < _1 5 A T \ \ D,- Q \ / _/ \ \ \ o L. L _L j L _j _L j SKID NUMBER (7 MPH(31MfS)) SKID NUMBER (7 MPH(31M/S)) Figre 12c. Comparison of Wet-Srface Accidents/Total Accidents Ratio to Skid Nmber ith Volme Stratification at AADT of 3,5. Figre 12d. Comparison of Wet-Srface Accidents/Total Accidents Ratio to Skid Nmber ith Volme Stratification at AADT of 4,. 16

30 !- c e o_6 1-. _4 Q o:r: _2 v 1.. l;r!.8 > ::.2 ADT t::. - o-3ooo o L _L J_ so SKID NUMBER (7MPH(31M/S)) o L J _L J L J SKID NUMBER (7MPH(31M/S)) 5 Figre 13a. Comparison of Wet-Srface Accidents/Dry-Srface Accidents Ratio to Skid Nmber ithot Volme Stratification. Figre 13b. Comparison of Wet-Srface Accidents/Dry-Srface Accidents Ratio to Skid Nmber ith Volme Stratification at AADT of 3, ADT 1.. t:.-o- 4ooo v m > c Q ,- Q-35 Ill ol----l L _L J _J L SKID NUMBER (7MPH (31MS)) o L--- L : SKID NUMBER (7MPH(31MIS)) Figre l3c. Comparison of Wet-Srface Accidents/Dry-Srface Accidents Ratio to Skid Nmber ith Volme Stratification at AADT of 3,5. Figre 13d. Comparison of Wet-Srface Accidents/Dry-Srface Accidents Ratio to Skid Nmnber ith Volme Stratification at AADT of 4,. 17

31 Analysis of Test Sections by Averaging Techniqes Three averaging methods ere sed to redce variability and, thereby, to more clearly demonstrate general relationships already apparent in the data sets ith and ithot volme stratifications. A discssion of these methods and the reslting trends follo. Cmlative Averages. To techniqes ere sed to calclate cmlative averages. The first involved calclating the average of each expression for accident occrrence for all test sections having a Skid Nmber less than or eqal to a given vale. The second procedre involved calclating average accident occrrence -- for -, each method of expression -- of all test sections having a Skid Nmber greater than a given vale. These average vales are plotted in Figres 14a-c, 15a-c, 16a-c, and 17a-c. In the first procedre of calclating averages, accident rates for lo Skid Nmbers had the greater inflence pon the average vale obtained. Extreme vales of the expressions for accident occrrence for a test section at high Skid Nmbers ere attenated throgh division by the large nmber of test sections ith loer Skid Nmbers. Ths, the second procedre, hich yielded opposite eightings, as necessary to verify the trends and to insre that large deviations at high Skid Nmbers ere not being masked by the averaging process. The reslting trends ere reasonably similar for et-srface accident rate and total accident rate. For other expressions, the trend lines ere not as similar; and the data points ere qite scattered. Wet-srface accident rate decreased as the Skid Nmbers increased to approximately 28; frther increase in Skid Nmbers reslted in only slight redction in accidents. Since the trends ere similar, and becase of the ranking of accident expressions discssed previosly, sbseqent analyses ere restricted to et-srface accidents per 1 million vehicle miles as the method of expressing accident occrrences. Average Wet-Srface Accident Rates Groped by Skid Nmber. In the second method, test sections ere groped by Skid Nmber. The average et-srface accident rate as calclated for each grop. These averages are plotted in Figre 18a-c. Again the trend indicated a rapidly decreasing accident rate ith increasing Skid Nmbers p to abot 27. The variability as greater than that obtained by the first method becase several grops inclded only one or to test sections, each having eqal eighting as grops containing a larger nmber of test sections. Still, the trends by the to methods ere very similar. Moving Averages. The third method involved calclation of an average et-srface accident rate and an average Skid Nmber for progressively ordered sets of five test sections. The first average as of the five!8 test sections ith the loest Skid Nmbers. The test section ith the loest Skid Nmber as then dropped, and a test section ith the next highest Skid Nmber added. This as repeated ntil all test sections had been averaged in a grop of flve. In cases here more than one test section had the next highest Skid Nmber, one of these as randomly added each time. Test sections ere dropped in the same seqence as they ere added. Reslting averages are plotted in Figre 19a-c. The trend as similar to those dev.eloped by the previos to methods. Hoever, this method indicated a more distinct change in the slopes of tl1f:_ to branches of the crve. At a Skid Nmber belo 27, the et-srface accident rate increased by to to three accidents per 1 million vehicle mifes per Skid Nmber, hereas above 27 the et-srface accident rate decreased nominal1y. The foregoing analysis here involved accident data for the entire 3 years hile skid resistance as measred in the smmer and fall of Pav,ements, of corse, exhibit loer friction in the smmer and fall, bt the measred vales may not be assmed to necessarily represent the loest friction dring the year for a particlar pavement nor for the road system as a hole. The rapid change in the slope of the crves in Figre 19a-c, for instance, may occr at some higher or loer Skid Nmber depending on hen measrements are made. If accident data ere sbdivided into sbsets for to periods of the year and the measred vales reflected the mean Skid Nmber for each period, the rapid change in the slope old then be expected to occr at the same Skid Nmber provided driver behavior remained the same throghot the year. Since skid resistance vales ere not available for the inter-spring period, Figre 2 a-c as prepared to sho the relationship beteen et-srface accident rate and pavement friction for smmer-fall period. In contrast to Figre 19a-c, a slightly higher Skid Nmber at hich the accident rate increased rapidly is evident. The data, hoever, are more scattered presmably de to the redced data base (less than half of all et-srface accidents). A shift toards a higher Skid Nmber old be anticipated becase the accident rates associated ith the 3-year period ere related to friction vales hich ere loer than the mean friction fr a fll year. De to greater scatter of data in Figre 2a-c, trends established in Figre l9a-c mst be accepted as the better indication of the relationship beteen accidents and pavement friction even thogh the accident data in Figre 2a-c more closely correspond to the measred skid resistance. Wet-srface accident rates ere calclated for I million miles of total travel nder all pavement conditions rather than et-srface travel mileage. The

32 tre accident rate for et-srface conditions old be several times higher since pav-..ments ere et only 13 percent of the time. Also, as shon in Table 4, precipitation dring the Jne-to-November periods ere sbstantially less than for the inter-spring periods. Yet, the et-srface accident rates for 197 and 1971 ere higher dring the Jne-to-November periods. Precipitation occrrences in 1972, especially dring inter-spring, as sbstantially more than for the to preceding years. If precipitation for the to 6-month periods (dring 3 years) ere the same, the et-srface accident rate of 19.5 (Jne to November) old be 25.2 compared to 19.3 for inter-spring periods. Therefore, loer skid resistance of pavements dring smmer and fal1 obviosly contribted to an increase in et-srface accidents. c < c, 8 < c < 2 e < LESS THAN OR EQUAL TO THAN Ill - GREATER SKID NUMBER (7 MPH(31M/SI) Figre 14b. Cmlative Averages of Wet-Srface Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3,. c - LESS THAN EQUAL c - > < < 15 c - GREATER THAN - c c, 15 e :;; < - LESS THAN OR EQUAL TO II -GREATER THAN - ---t <O 15 5 SKID NUMBER (7 MPH ( 3IM/S)) SKID NUMBER (7 MPH (31M/S)) Figre 14a. Cmlative Averages of Wet-Srface Accident Rate Verss Skid Nmber ithot Volme Stratification. Figre 14c. Cmlative Averages of Wet-Srface Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3,5. 19

33 O- LESS THAN OR EQUAL TO Ill-GREATER THAN Figre ISa. Cmlative Averages of Total Accident Rate Verss Skid Nmber ithot Volme Stratification.,, L _i J L ! SKID NUMBER (7 MPH(31M!S)) Figre ISb. Cmlative Averages of Total Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3,. - ;;; - 1 > g 9 r : z > 8. : :::: 7 z e 8 O-LESS THAN OR EQUAL TO &-GREATER THAN o o 111-6QL J J j JL j j SKID NUMBER (7 MPH (31 MIS I) 111 -,j 1 - > z 9 e = z > z r - LESS THAN OR EQUAL TO Ill-GREATER THAN L l l o o jc c_-- - oo Figre!Sc. Cmlative Averages of Total Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3,5. SKID NUMBER (7MPH(3IM/S)) 2

34 LESS a-greater THAN THAN OR EQUAL TO Figre 16a. Cmlative Averages of Wet-Srface Accidents/Total Accidents Ratio Verss Skid Nmbers ithot V olne Stratification j: L _j j L.L cc ]: J SKID NUMBER (7 MPH(31M/Sll.4 -LESS THAN OR EQUAL a-greater THAN TO...J ;;: oo.2 - Vll. c.15 Figre 16b. Cmlative Averages of Wet-Srface Accidents/Total Accidents Ratio Verss Skid Nmber ith Volme Stratification at AADT of 3,. O.IO L l SKID NUMBER(7MPH(31MIS)) _4 -LESS THAN OR EQUAL TO a-greater THAN Figre 16c. Cmlative Averages of Wet-Srface Accidents/Total Accidents Ratio Verss Skid Nmher ith Volme Stratification at AADT of 3,5...2 <( SKID NUMBER(7MPH{31M/5))

35 -LESS THAN OR EQUAL TO Ill -GREATER THAN.,.,., v o<.coc-oo----_q-oc,o Figre 17a. Cmlative Averages of Wet -Srface Accidents/Dry-Srface Accidents Ratio Verss Skid Nmber ithot Volme Stratification..2., 2 - LESS THAN EQUAL SKID NUMBER (7 MPH!31 MIS)) Ill - GREATER THAN O< Figre 17b. Cmlative Averages of Wet-Srface Accidents/Dry-Srface Accidents Ratio Verss Skid Nmber ith Volme Stratification at AADT of 3,. ; s OA. 2 -LESS THAN OR EQUAL TO., 2 < Ill - GREATER THAN SKID NUMBER ( 7MPH( 31 MIS)) z.7 lij.6 I Q.5 t! Figre 17c. Cmlative Averages of Wet-Srface Accidents/Dry-Srface Accidents Ratio Verss Skid Nmber ith Volme Stratification at AADT of 3,5.. 2 < SKID NUMBER (7 MPH(31 M/S)) 22

36 4 35 Figre 18a. Average Wet-Srface Accident Rate of 11 Test Sections. Groped by Skid Nmber Verss Skid Nmber ithot Volme Stratification j _j _j _j l. l,j SKID NUMBER ( 7 MPH(31 M/Sll Figre 18b. Average Wet-Srface Accident Rate of 94 Test Sections Groped by Skid Nrnnber.. Verss Skid Nrnnber itb Volme Stratification at AADT of 3, SKID NUMBER ( 7 MPH ( 31 M/S)) 35 3 e Figre 18c. Average Wet-Srface Accident Rate of 82 Test Sections Groped by Skid Nmber Verss Skid Nmber ith Volme Stratification at AADT of 3, C SKID NUMBER (7 MPH(31 MIS)) 23

37 35 cl oc I > z z z oc !) 15 Figre 19a. Five-Point Moving Averages: Wet-Srface Accident Rate Verss Skid Nmber ithot Volme Stratification SKID NUMBER (7 MPH(31M/S )) 35 Figre l9b. Five-Point Moving Averages: Wet-Srface Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3,. ;; 3 oc z > 25 Cj z 2 oc \BO!t /, dj c CJDo SKID NUMBER (7 MPH (31 MIS}} 35 oc 3 z 25 Q 2 oc 15 1 g Bo oso %/, cp (]) c9 Figre l9c. Five-Point Moving Averages: Wet-Srface Accident Rate Verss Skid Nmber ith Volme Stratification at AADT of 3, SKID NUMBER ( 7 MPH(31 MIS}} 24

38 4 Figre 2a. Five-Point Moving Averages: Wet-Srface Accident Rate for Smmer-Fall Periods Verss Skid Nmber ithot Volme Stratification., & CZ) o 3!; z, c <6fb 6 /,O oi; 8 c,., o O o 4 c >, 3 z, o 8 as o 6 8 o oo o b o@ c o o o cp Figre 2b. j, 3 4 SKID NUMBER (7MPH(31M/S)) Five-Point Moving Averages: Wet-Srface Accident Rate for Smmer-Fall Periods Verss Skid Nmber ith Volme Stratification at AADT of 3,. L _L L _L J -- Figre 2c. 3 4 SKID NUMBER (7MPH(31 MIS)) Five-Point Moving Averages: Wet-Srface Accident Rate for Smmer-Fall Periods Verss Skid Nmber ith Volme Stratification at AADT of 3,5. 4 ll 3 o o z c D, cs > 6 D /,. - f: o o OO 3 4 SKID NUMBER (7MPH(31 M/S}l 25

39 PEAK SLIP NUMBERS AND ACCIDENTS As discssed previosly, there is a need to analyse different measrements of pavement friction to determine hich correlates the best ith accident experience. The peak friction force as measred rotinely dring all tests; ths these data ere available for analysis. Test section averages ere arrayed as shon before. Peak Slip Nmbers (PSN) ere sbstitted for Skid Nmbers. The arrays again indicated the desirability for sorting the data by AADT. Wet-srface accident rates again appeared to be the best expression for accident occrrence. Test sections ere groped by Peak Slip Nmber, and average et-srface accident rates ere calclated for each Peak Slip Nmber as shon in Figre 21 a-c. The plots indicate more scatter than as obtained ith Skid Nmbers (Figre!Sa-c) and may be de to each data point representing feer test sections. The greatest change of slope occrred at a Peak Slip Nmber of abot. 57. Similar reslts ere obtained tilizing the five-point moving average, as shon in Figre 22a-c. Stratification of data by AADT at 3, and 3,5 vehicles per day minimized scatter, bt the change in slope remained at the same Peak Slip Nmber. The point of greatest change in slope of crves in Figre 22a-c as at Peak Slip Nmber of 57 and in Figre 19a-c at Skid Nmber of27. According to Figre 8, a PSN of 57 is eqivalent to SN of approximately 27. Scatter of the data in Figre 19a-c in comparison to Figre 22a-c also appear to be similar and, therefore, sggests that both measrements of friction relate eqally ell to accident occrrence. This as somehat nexpected becase of the inherent measrement and chart analysis errors associated ith peak slip resistance (PSN) determination. Peak slip resistance occrs for a very brief period of time dring heel lock-p, and the measrement represents a mch shorter length of pavement than the locked-heel test (SN). For that reason, poor agreement beteen SN and PSN in Figre 8 as credited largely to inaccracies in PSN. If sch a conclsion is valid, some of the scatter of data in Figre 22a-c may be de to errors in PSN. In that event, Peak Slip Nmber may correlate best ith accident experience. SUMMARY AND CONCLUSIONS On rral, for-lane, interstate and parkay (expressay) facilities, et-srface accidents per 1 million vehicle miles correlated best ith skid resistance. Even sing the best statistical expression of accidents, 26 scatter and sprios variability in data seem inevitable. Stratification of the data by AADTs minimized scatter. Averaging methods as a means of developing trends and minimizing scatter beteen variables ere sed in the stdy. Of the averaging methods investigated, the 11moving average11 yielded more definitive reslts. Definite trends ere established in regard to the relationship beteen et-srface accident rates and Skid Nmbers (Figre!9a-c). Wet-srface accident rate decreased rapidly as the Skid Nmber (7 mph) increased to 27; frther increases in Skid Nmber beyond this point reslted in only a slight redction in accident rate. The analysis here involved accident data throghot the year (3-year period) hile skid resistances ere measred in the smmer and fall (1971) hen pavements normally exhibit loer friction vales. As expected, analysis of the accident data for 6-month periods (Jne to November) (Figre 2a-c) indicated a slightly higher friction. The data, hoever, ere more scattered de to the redced data base. Wet-srface accident rates for 2 of the 3 years considered in this stdy ere higher dring the smmer-fall periods (Table 4) even thogh the roads ere et a lesser proportion of time. When adjsted to eqal time of precipitation dring December to May and Jne to November, et-srface accident rates for the smmer-fall periods ere higher for all 3 years. Loer skid resistance of p::vements dring smmer and fall obviosly contribted to an increase in et-srface accidents. Definite trends ere also evident beteen et-srface accident rates and Peak Slip Nmbers. The greatest change in slope of the trend lines (Figre 22a-c) occrred at a PSN 7 of abot 57. Scatter of data as no orse than that for Skid Nmbers. This as somehat nexpected becase of the inherent measrement and chart analysis errors associated ith peak slip resistance determinations. A Peak Slip Nmber of 57 is eqivalent to a Skid Nmber of approximately 27 (Figre 8); and, as shon in Figre l9a-c, a SN 7 of 27 also corresponds to the greatest change in slope of trend lines. This sggests that either the correlation beteen the to friction measrements overshadoed any sbtle differences that may exist beteen accidents and either measrement of friction or that both skid and slip resistance are eqally valid indexes of pavement friction. It shold be reemphasized that the findings cited here pertain to rral, for-lane, limited access, expressay type highays ith posted speeds of 7 mph (3 1 m/s) and that no consideration as given in the analysis to geometries of roadays nor to points of traffic conflicts. High accident or repeat accident locations certainly arrant frther stdy to determine