7......igloaoz. . .3. - nut-"rut?“ .. 7313.3; .331... 33-”...331vmmN‘NMWN 193.33 .m. «33.3.... wfi .wfiflfifi? . Ln... 3 3 .3 ‘a. . .- 3. . 133v...“ 33..- g . . 3:33.. 33.3.“..333 “1333.4“ 3 .13333733... 33 .55 3.. 37 [3.333013 3! . 333. 3. .,...(.33.1. 3.3 3: . 3363. ”Fill 4.133....33.‘ flaw: 33...}...333 .333. cunanufiqchuvymer 33.73 33/. flrfaz 3.3. MCHMH Mm 33~I5333363v3h$r| "final. .3 L.?~3.33 3 3...... u.- ...-...,.x.mu. 3 . u 3 b?! I :L I V V ...... 3 3.3333. 3% 3.3 .. . V 3 , . .. .3... .. riflkmfinfi- . .. H . . . . -. . . _.... . . . .. .. _ 3.4 h . ”Wu... ”iv 3 . u. .. .. .. ,... .3 . _. .. .. 3 3 .3 :3 3 .. 3. .3. .. . . .3 w....- A... ... . .. .3 . 3 ... . . 31.7th- .9 a! ;..o 3. :3 . , :. . . ., H . .. . . .. . . . V 3 33. ukmnhtéé . a: ...m.Ncr3 . . .5. _. a u. . . . V .. . 3. .. z . 3 . .. . . . .33 $fh3fi3h03 afiuummfinm . .. . . . V -. . . 33:33... .3 .3. .. .. .3 f. : . . ., .1 . .3 .. , . : . . . . .3 . . . .. 3 .. . . .. . .. 33.3.3.3? 4&531633 I33 (I! up... I ”haw... I E 3. .3 . . 3. L. .. . 3 . . .. .3. 3 . .3. .3 . I.” ovfiihuuu 3 . .H . mam... .3L.a~..m..V..n.~ .:. - -_ 1.3... . . - 3m” _d .3. .3... . .3 2m fiixnmnmfi “33...... 3 . ’31....“ :5 Lune. w v .51“: 3 i fv ._ -. "i r 333“” “ 3 33'. 333. 3 a!!!) 3. .283; 3- v 3 312.....3 LI fihuolh‘ r. 3...). 3. 0.. 19.34% 3.133%:- 33 Paul... .33. . .IN3333 . 451 ‘i £303. 3 H. $3M... J. NW3. . 3 . .PL. 33.31.33... 33333333.“.3333333 III} 3. 33¢ a a. '. .vlcl 3 3 «133.15.- .I. dbuflm. 3. } '3 find . .C .33'3‘1331... n 31.3? 3).?“ .3. . 3 tau". 3.3... 3 3r. 3% .233“ .- . .3Vm3m3n333fi3mfi393bwu 3 3 .3... fl; .n..1. 3: I2 ~ W 1.3.3., Lava; 3.»...333 . . h. .. .3 . a I u!’ ‘9‘ o 3.3 3- vb 3 utt- 3133 3333 3 .23.: .43 . . . FL . 3t 3 in“. . L P .}.l\3 .333: . €fi$$3finfifimfifi 3». Vufifi....¥.....mzt.. .V 3 3 . .. .32.... 333...; :31; 3.3 Iqfiggmwfiflagagiw. 33.3... .333.3..3.... . . .. . V. _ .. .- ,., : H33..n.u333.33133333333. 3. w 3 an 3 . 3. .fifi. sad-3.9.3? .1. 33.33.33. - .. . 3 . \D xii: . hide? .23.. 3 .3 $313.33... $3., .333333W3. W333? 1933.4 i 13. J6 swur )1 Hum . v I I U” Iv. iv .. . Luvd3fl,v.uc.3.$v.w:...mmf .. 3 . .mmnum.) lgvf-thflbfihpd 3.1.3.... 3.3.3 3.L é’.‘ “9‘ 2 S .IL'F'IN . M hflfigt . t3} .Iln . ['33. 1.1! i .3... . 3.3.... . 3 if» . 3313.33 . .333bfji333lnr 333.33... . 33.3 , 3H . at! 3 333 3.3. 3.3.3.133 31.3 33 .3 3.63... .5 It'll ’- -‘ .O ‘2 «I I’!~ ’ ‘I ‘ .fi'iu'c-Vt‘i VI 3 n4 . nlv 1": )0 £1.3an- 31-110.! 3 I3 I. (Vvtanit! -‘l “-|l\‘.u.6 ‘ Y’f. ‘ III“: .-ls..a.‘.VHIV.vI-V 33 . 3. . . fixuum333mmuflmm Maw...» .3...3..33.n33..3.3..3.. 3 36303§3.§%! .3 3. $3313 3.... 3.... 3 3 3.33 333 :3 .010". (I)... "c.l¢i..<.l. 333 -. ...qulif¢¢3tlrgil32.3 .3. 3 . RI..- .333. . 3.3. 3....-. 3 .. : 33.3.13): 33.3.3.3 333a... . 31.33.. . :4! J3... 33.3.. 3-... I 3.. 31... 333. Hok3ah3.u~fl§l33.w33~v.3n.3lhfiuw .33. 133.333....w4 V. . . 333.33.wn.mw.3.3 mun-“.mbflstar .nb. . , 3?... . . '3 . '3.“ .9V 3..) .35 . .Jv‘ 3'31”.“ 37:31. .7 r! t fluid-I333: 3H.alun.¢3:ndsh.i. : . .3 . 31 13333.3. . 3 . .0633: -h 3.. 33333333533.“ Nd.“ 341A: ,3hvm3wun..um3.uwuun . witllb g . n 1 1 .IILV3.trIVtI . z 5"! 3» Y ti3l3.330h30( I .1. I 3 3Mflh..flfl~nn.un.3n 3 3.. 3...... . +33 3 . 3. .3 53.9“3 33.3.3343"... . .333. 3 . uenuflvnwflsr}..u-anv393m3...3.333.333...2.TJ 3.... 3 33......3 33.33;... V . 3 3 H44 3.39.3353“? $335123... 3.33., a. .3.-- - 3 3 .. . $ . {33.3.33 . 33333.33. . . I . . .3 3 Jumommq. - . . 3 . . 3. - 1"? 33-! 33. a. 3. 1'3 3 3 3.0.“: .. y . . “‘113‘1‘v ‘33.“... fit... 3 f 3w] . - [1.2.3. “IQiII V . . 1.3.353! us ‘3; .1117. . 337334.4353va-3f3bn3 3333333333333; :5. 33:13 .33... 33.3.3.3}..- . . 33 3. D. . 3 . . . ”3...... ...33..n.33....3«..3.3.n3.3..... n.3,". . 33.3. 33331333.. . 3 33333.. 31....3333331835033. . V. . .33...3...m3h,33h 33 . 3- 3.3.3.3- . . 3.3 33 33.337. 3.3.-. .333 Ln333a333 3.3 . .. 33 3.9.23: 373.33.13.33... :3 Ilvtl¢lv§~|33dilpu\3\pl .3 .3 .3 i “I‘zsl‘.|§3lrpl33 3 333338.53‘: . .. . . :3.-. 1.3.3.3113]-.. . v .I‘ DI! c "tV-(z l' lleI'3 $11.33! ‘I. . I 3.939% .WHQWuuuuaPVfifiUMHJF-m 3. .373. 333.33%}! . ll . ”Wm. 33 a. 3: .. 3me. 3.3.3.35? Him-.3 -3 .. 332-... : 3.3333. 33%., 3.3M»... 33. 3 n.33..3.m.333-..w333.3 u I . II. 3. I I IN 3‘ In-» I 3 3 .3 . .9 3.933333 “huh-.3. .33 3332.... .. .33 o I .3....l.5..3..:39 3.1.3.3333. :w .333... 4.5.5. . . . 33333333."; 3333!. Fnhr 331353333 Lvhflfack... 3 . . V ., and»! 33303344333333..3ru33.fl..3.3.i3 3.33 3 33.3 .353. .....-..33 . 3.1....L3t 3. 3...?3auufl... 3.33.13.13.33 3333.33.94. . . . . I. .32. {3.4.3 «33533....33333333335 33:... .3333.-.33%!3..3.3.32.3.3.l .3 .3 lL.nr.5.L.333..3t3: 3.. .. ..3...3....33...3.n33.3.: 33.33.933.33: inunbw ...3.3 .. .. {Ishtar-3333.331. .33.). 3 .. ...43.L..hrxhu_3lr.l..§f.3 3.33 . 3l.Yy|TIO"Av.3 . .I3 .I u 3 . . 3. ole 3 303%.. ‘33,..33l.‘ oV|\ Q . . 4 <3 .m‘ I 3 [.33 .3! 3|! 3". 3 . . . . 3 3 I! «Kn-n3 HE. ‘33.».535 Hula-.33. ‘3; ..3 (L333 r343 3331'3'333I03'Ivr...53333;\v. . 3!. I3 . 3. 3 ..&T- It .n33 . urhlc .36 0: no“! lu3'l '3!..JI..33II I :303333 3333‘ ol3ci33-tofifl1..o.l.¢tall3o331-3933333303l3o\.. ‘ 3 .- I.9.V\Ffl.l-|¢ILI 3L: 3 3|: 3! ‘01.!!9 . ‘lxouvfrlvi v0. ..lo.\|a||!3!09£v>.t '33:: 1-3.3.343 .333.i.3:3 4.333 33 u 3 .13)‘. 3 3 - 373‘ \OK 3 r :1. 03.333! ilvlun. sl. (If. nublnv‘ Lair-fl»... 33'0..~H.tnv....u‘".:u.33 .3. . .3 ”3301.. 3' 7 3.73933“ . . . ... 3.33.. .....HHIIA.M|LIIII.3 3|..qu3’..kfl331.'v.n.ttnul'l..l¢ IOWA-+13 3133.1..3'33 3\na\a.I\Jl'|llllbv3|II3lhn.i...3.)l09.n£.lbllv‘v‘r3 3.3.33.3... 11.39.33.3'I3rc .33."..-6!" 3.331. .330 u 3 3 3 3 L. . 3 33333.3 I.V¢I3l.3'.3w.3.l . 3 ~I| I 3 VH3 . I 1133’ 3 3 3 3 I 3 3| 3 llrn 3,3. I n r r 3 3| 3 l! [3.3" 3.33.333 3 3 '3 3| 3 39333 In: THESIS IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIOIIIIIIIIIIIIIIIIIIIII 3 1293 0157 LIBRARY Mlchigan State Unlverslty This is to certify that the dissertation entitled AN INVESTIGATION OF THE EFFECTIVENESS OF BOULEVARD ROADWAYS presented by Salvatore Castronovo has been accepted towards fulfillment of the requirements for Ph . D . degree in Civil Engineering QWM Major professor Date 5/6/97 MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 PLAOE II RETURN BOX to move this chockout {tom your "cord. TO AVOID FINES Mum on or him dd. duo. DATE DUE DATE DUE DATE DUE ________.__J—- V___———— AN INVESTIGATION OF THE EFFECTIVENESS OF BOULEVARD ROADWAYS By Salvatore Castronovo A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Civil and Environmental Engineering 1997 ABSTRACT AN INVESTIGATION OF THE EFFECTIVENESS OF BOULEVARD ROADWAYS By Salvatore Castronovo The State of Michigan has been constructing directional cross-overs in the medians of boulevard roadways since the 19605. These directional cross-overs were constructed to alleviate the congestion and interlocking that were occurring at bidirectional cross-over locations in boulevard roadway sections. The value of these cross-overs has not yet been determined, even though (apparent) improved operations have been observed after the implementation of this design. The objective of this research is the investigation of the safety benefits of divided highway and directional cross-over median designs. The independent variables that were developed for this research were: roadway cross-section type (i.e., boulevard vs. center lane left-turn); median width; lane width; cross- over type (directional vs. bidirectional) and spacing; speed limits; population density; traffic volumes, percent commercial; intersection density, signal density, and shoulder/curb. The dependent variables that were evaluated are the accident rates and accident types. A data set of the pertinent state highway facilities was generated utilizing the Michigan Department of Transportation’s photo log records and from field reviews. The study segments consisted of 366.3 miles of five-lane roadway, 55.8 miles of seven-lane roadway, and 512 miles of boulevard roadway. Accident data from 1970 to 1990 were retrieved to provide the data necessary to evaluate the dependent variables. An investigation of the mean accident rates of highways with boulevards, compared to the mean accident rates of highways with continuous center lefi-turn lanes (CCLTL), revealed that there is a significant difference, with boulevard highways having a lower mean accident rate for those Michigan highways investigated. The investigation into the difference of the mean accident rate for boulevard highways with medians of different widths provided an indication that roadways with medians greater than 30 feet but less than 60 feet had the lowest mean accident rate of the three median categories investigated. The differences were significant for most accident types when compared to roadways with medians less than or equal to 30 feet and greater or equal to 60 feet. The investigation of the accident rates for boulevard highways with directional and bidirectional signalized cross-overs, while not conclusive, provided an indication that boulevard highways with directional cross-overs had a lower mean accident rate than boulevard highways with bidirectional cross-overs for signalized roadways. ACKNOWLEDGMENTS My sincere appreciation goes to the following individuals for the valuable assistance and encouragement they have provided throughout this research project: Dr. Thomas Maleck, my committee chair, who served as a role model and without whose guidance it would not have been possible to complete this research. My committee members Dr. Chia Shing Lai, Dr. Roy Erickson, Dr. Richard Lyles, Dr. Mark Snyder, and Dr. William Taylor for their perseverance and assistance. The Michigan Department of Transportation, whose support of the research project made possible the collection of data that served as the basis of the investigation. The numerous individuals at the Michigan Department of Transportation who supported the project including Ms. Mary Hudak Baylis, Mr. Brett Kach, and Mr. Arthur Sloboski. The Student’s at Michigan State University who supported the research, including Mr. David DeBerardino, Mr. Paul Dorothy, Mr. Michael Scheuer, and Ms. Sara Nolf. My family who persevered while I pursued this work. My brother Vincent, sister Carmela, and Aunt Josephine Insalaco Belli who provided guidance during my early developmental years. iv I dedicate this dissertation to my wife Joann who provided me support and encouragement, and to my parents Pietro Castronovo and Gaetana Insalaco Castronovo, who emigrated from Sicily to the United States of America and worked many long hours depriving themselves so that their children would receive an education and lead better lives. TABLE OF CONTENTS CHAPTER 1 ........................................................... 1 INTRODUCTION AND RESEARCH BACKGROUND ......................... l 1.1 HISTORICAL PERSPECTIVE ................................. 1 1.2 THE PROBLEM ........................................... 19 1.3 OBJECTIVE .............................................. 19 1.4 RESEARCH APPROACH ................................... 20 1.5 ORDER OF PRESENTATION ................................ 20 CHAPTER 2 .......................................................... 22 LITERATURE REVIEW ................................................ 22 2.1 MEDIAN WIDTH .......................................... 22 2.2 CROSS-OVER TYPES ...................................... 30 2.3 CROSS-OVER SPACING ................................... 33 2.4 DRIVEWAYS ............................................. 33 2.5 TURN-RELATED ISSUES ................................... 37 2.6 TRAFFIC VOLUME ........................................ 40 2.7 RAISED MEDIAN S ........................................ 40 2.8 CCLTLS .................................................. 41 2.9 ROADWAY GEOMETRY ................................... 43 2.10 LITERATURE SEARCH SUMMARY ......................... 46 2.11 RESEARCH OBJECTIVES .................................. 47 CHAPTER 3 .......................................................... 48 RESEARCH METHODOLOGY .......................................... 48 3.1 THE VARIABLES OF INTEREST ............................ 48 3.2 THE RESEARCH DATA .................................... 49 3.2.1 DATA INDEX ....................................... 49 3.2.2 ROADWAY FEATURES .............................. 51 3.2.3 TRAFFIC CONTROL DATA ........................... 52 3.2.4 THE ACCIDENT DATA .............................. 52 3.5 SITES OF INVESTIGATION ................................. 55 vi 3.6 EXECUTION OF THE EXPERIMENT ............................... 55 3.7 DATA INSPECTION ....................................... 59 3.8 UNIT OF ANALYSIS ....................................... 59 3.9 STATISTICAL ANALYSIS .................................. 59 3.10 CHAPTER SUMMARY ..................................... 60 CHAPTER 4 .......................................................... 62 DATA ANALYSIS ..................................................... 62 4.1 INTRODUCTION .......................................... 62 4.2 DATA INSPECTION ....................................... 62 SUMMARY OF DATA INSPECTION ............................... 89 4.3 DISCRIMINATE ANALYSIS ................................ 90 4.3.1 BOULEVARD ROADWAY ALL ACCIDENTS ............ 91 4.3.2 BOULEVARD ROADWAY REAR-END ACCIDENTS ..... 93 4.3.3 BOULEVARD ROADWAY DRIVE-RELATED ACCIDENTS ................................................... 95 4.3.4 CCLTL ROADWAY ALL ACCIDENTS .................. 98 4.3.5 CCLTL ROADWAY REAR-END ACCIDENTS .......... 100 4.3.6 CCLTL ROADWAY DRIVE-RELATED ACCIDENTS ..... 103 4.3.7 DISCRIMINATE ANALYSIS SUMMARY .............. 105 4.4 ACCIDENT ANALYSIS ................................... 105 4.4.1 BOULEVARDS AND CONTINUOUS CENTER LEFT TURN LANES ACCIDENT ANALYSIS ....................... 106 4.4.1.1 TESTING OF FIRST HYPOTHESIS .............. 106 4.4.1.2 SELECTION OF RANGE OF INVESTIGATION . . . . 106 4.4.1.3 INVESTIGATION OF ALL SEGMENTS .......... 108 4.4.1.4 INVESTIGATION OF UNSIGNALIZED SEGMENTS 109 4.4.1.5 INVESTIGATION OF SIGNALIZED SEGMENTS ........................... 1 10 4.4.1.6 INVESTIGATION OF BOULEVARD ACCIDENT RATES: UNSIGNALIZED VS. SIGNALIZED SEGMENTS . . 111 4.4.1.7 INVESTIGATION OF CCLTL ACCIDENT RATES UNSIGNALIZED VS. SIGNALIZED SEGMENTS . . 112 4.4.1.8 SUMMARY .................................. 113 4.4.2 BOULEVARD ACCIDENT ANALYSIS FOR VARYING MEDIAN WIDTHS .......................................... 113 4.4.2.1.TESTING OF SECOND HYPOTHESES ........... 113 vii 4.4.3 4.4.4 4.4.5 4.4.2.2 SELECTION OF RANGE OF INVESTIGATION ....................... 1 15 4.4.2.3 INVESTIGATION OF ALL, UNSIGNALIZED AND SIGNALIZED SEGMENTS ..................... 115 4.4.2.4 SUMMARY .................................. 117 BOULEVARD ACCIDENT ANALYSIS FOR VARYING POSTED SPEEDS .......................................... 118 4.4.3.1 TESTING OF SECOND HYPOTHESES ........... 118 4.4.3.2 SELECTION OF RANGE OF INVESTIGATION ....................... 1 18 4.4.3.3 INVESTIGATION OF ALL UNSIGNALIZED AND SIGNALIZED SEGMENTS ..................... 121 4.4.3.4 SUMMARY .................................. 121 BOULEVARD ACCIDENT ANALYSIS FOR VARYING AVERAGE DAILY TRAFFIC ......................... 122 4.4.4.1 TESTING OF THE SECOND HYPOTHESES ......................... 122 4.4.4.2 SELECTION OF RANGE OF INVESTIGATION ....................... 122 4.4.4.3 INVESTIGATION OF ALL SEGMENTS,UNSIGNALIZED SEGMENTS, AND SIGNALIZED SEGMENTS . . . . 125 4.4.4.4 SUMMARY .................................. 126 BOULEVARD ACCIDENT ANALYSIS FOR DIRECTIONS VS. BIDIRECTIONAL CROSS-OVERS ..................... 126 4.4.5.1 TESTING OF THE THIRD HYPOTHESIS ......... 126 4.4.5.2 INVESTIGATION PARAMETERS ............... 126 4.4.5.3 INVESTIGATION OF ALL SEGMENTS .......... 126 4.4.5.4 INVESTIGATION OF UNSIGNALIZED SEGMENTS 127 4.4.5.5 INVESTIGATION OF SIGNALIZED SEGMENTS ........................... 127 4.4.5.6 SUMMARY .................................. 127 4.5 CHAPTER SUMMARY .................................... 129 CHAPTER 5 ......................................................... 132 CONCLUSION ....................................................... 1 32 5.1 SUMMARY .............................................. 132 5.2 BOULEVARD VS. CCLTL ................................. 133 5.3 UNSIGNALIZED VS. SIGNALIZED ......................... 133 5.4 BOULEVARD MEDIAN WIDTHS ........................... 134 5.5 BOULEVARD MEDIAN CROSS-OVERS ..................... 134 viii APPENDIX A DATA COLLECTION ........................................... 135 APPENDIX B DISCRIMINATE ANALYSIS ..................................... 155 APPENDIX C DATA ANALYSIS RESULTS ..................................... 162 APPENDIX D SIGNIFICANCE TEST RESULTS .................................. 189 REFERENCES ....................................................... 200 ix Table 4.4.1.1 Table 4.4.2.1 Table 4.4.3.1 Table 4.4.4.1 Table 4.4.5.1 Table A.l Table D.l Table D2 Table D3 Table D4 Table D5 LIST OF TABLES Boulevard and Center Left-Tum Lane Accident Rates .......... 107 Boulevard Accident Rates by Median Width .................. 114 Boulevard Accident Rates by Posted Speed Limit .............. 119 Boulevard Accident Rates by Average Daily Traffic ............ 123 t-test Results for Directional vs. Bidirectional Cross-Overs at 0.95 Level ........................................... 128 Investigation Sites ....................................... 135 Boulevard/CCLTL t-test .................................. 189 Boulevard/CCLTL t-test for Unsignalized Roadways ........... 190 Boulevard/CCLTL t-test for Signalized Roadways ............. 191 Boulevard t-test for Unsignalized/Signalized Roadways ......... 192 CCLTL t-test for Unsignalized/Signalized Roadways ........... 193 Figure 1.1.1 Figure 1.1.2 Figure 1.1.3 Figure 1.1.4 Figure 1.1.5 Figure 1.1.6 Figure 1.1.7 Figure 1.1.8 Figure 1.1.9 Figure 1.1.10 Figure 1.1.11 Figure 1.1.12 Figure 1.1.13 Figure 2.1.1 Figure 2.1.2 Figure 3.6.1 Figure 3.6.2 Figure 3.6.3 Figure 3.6.4 Figure 4.2.1 Figure 4.2.2 Figure 4.2.3 Figure 4.2.4 Figure 4.2.5 Figure 4.2.6 Figure 4.2.7 Figure 4.2.8 Figure 4.2.9 LIST OF FIGURES Survey of State of the Practice Letter .......................... 4 State of the Practice Survey Results ........................... 5 Rural Boulevard Section With Bidirectional Cross-Overs and Residential Drives ........................... 7 Urban Boulevard Section With Directional Cross-Overs With Shoulders ................................ 8 Urban Boulevard Section With Bidirectional Cross-Overs ........ 10 Urban Boulevard Section With Back-to-Back Directional Cross-Overs ................................... 11 Urban F ive-Lane Center Left-Tum Lane Cross-section ........... 12 Urban Seven-Lane Center Left-Turn Lane Cross-section ......... 13 Michigan Department of Transportation Bi- Directional Cross-Over Standard ............................ 14 Michigan Department of Transportation Directional Cross-Over Standard ............................ 15 Michigan Department of Transportation Proposed Directional Cross-Over Standard .................... 16 Michigan Department of Transportation Proposed Directional Cross-Over Standard .................... 17 Michigan Department of Transportation Proposed Directional Cross-Over Standard .................... 18 Guide for Determination of Street Widths and Median Treatment . . 25 Relationship of Median Widths to Accident Rates ............... 29 Photolog Equipment ...................................... 57 Loading Optical Disk Into Photolog Machine .................. 57 Operating Photolog Equipment ............................. 58 Photolog Image of Road Section ............................ 58 Accident Percentage for Each Type of Accident, All Segment Lengths ..................................... 64 Accident Severity Percentage for All Accidents ................ 66 Traffic Volume Percentage for All Section Lengths ............. 67 Percent Commercial vs. Percent of Miles, All Section Lengths ..... 69 Population Percentage for All Section Lengths ................. 71 Speed Percentage for All Sections ........................... 73 Mile Percentage for Intersection Density, All Section Lengths ..... 74 Mile Percentage of Signal Density, All Section Lengths .......... 76 Mile Percentage of Driveway Density, All Segment Lengths ...... 78 xi Figure 4.2.10 Figure 4.2.11 Figure 4.2.12 Figure 4.2.13 Figure 4.2.14 Figure 4.2.15 Figure 4.3.1 Figure 4.3.2 Figure 4.3.3 Figure 4.3.4 Figure 4.3.5 Figure 4.3.6 Figure 4.4.2.2 Figure 4.4.3.2 Figure 4.4.4.2 Figure A.1 Figure A.2 Figure A.3 Figure A.4 Figure A.5 Figure A.6 Figure A.7 Figure A.8 Figure A.9 Figure A.10 Figure A.11 Figure A.12 Figure A.13 Figure B] Figure C.1 Figure C.2 Figure C.3 Mile Percentage of Commercial Driveway Density, All Section Lengths ...................................... 80 Mile Percentage of Lane Width, All Section Lengths ............ 82 Mile Percentage of Curb/No Curb, All Section Lengths .......... 83 Mile Percentage of Median Width, All Boulevard Section Lengths .............................. 85 Total Percentage of Cross-Over Type, All Boulevard Section Lengths .............................. 86 Mile Percentage of Section Length, All Section Lengths .......... 88 Discriminate Analysis Results for All Boulevard Roadway Accidents 92 Discriminate Analysis Results for Rear-End Boulevard Roadway Accidents .............................................. 94 Discriminate Analysis Results for Drive-Related Boulevard Roadway Accidents .............................................. 97 Discriminate Analysis Results for All Continuous Center Left-Tum Roadway Accidents ...................................... 99 Discriminate Analysis Results for Rear-End Continuous Center Left-Tum Roadway .............................................. 101 Discriminate Analysis Results for Drive-Related Continuous Center Left- Tum Roadway Accidents ................................. 104 One-Way Anova Results for Varying Median Widths for Boulevard Roadways .................................... 1 16 One-Way Anova Results for Varying Posted Speeds for Boulevard Roadways .................................... 120 One-Way Anova Results for Varying Average Daily Traffic For Boulevard Roadways .................................... 124 MDOT District Map ..................................... 139 Photo Log Instructions ................................... 140 Photo Log Codes ........................................ 141 Data Collection Sheet “A” ................................ 145 Data Collection Sheet “B” ................................ 146 Data Collection Sheet “C” ................................ 147 Data Collection Sheet “D” ................................ 148 Data Collection Sheet “E” ................................ 149 Data Collection Sheet “F” ................................ 150 Work Rules ............................................ 15 1 Field Data Collection Instructions .......................... 152 Field Data Collection Codes ............................... 153 Data Collection Sheet “G” ................................ 154 Discriminate Analysis Example ............................ 155 Accident Rates for Boulevard and Center Left-Turn Lane Accidents 162 Accident Rates for Unsignalized Boulevards and Center Left-Tum Lanes ........................................ 163 Accident Rates for Signalized Boulevards and Center Left-Tum Lanes ................................................. 1 64 xii Figure C.4 Figure C.5 Figure C.6 Figure C.7 Figure C.8 Figure C.9 Figure C.10 Figure C.11 Figure C.12 Figure C.13 Figure C. 14 Figure C.15 Figure C.16 Figure C.17 Figure C.18 Figure C.19 Figure C.20 Figure C.21 Figure C.22 Figure C.23 Figure C.24 Figure C.25 Figure C.26 Figure C.27 Figure D.1 Boulevard Accident Rate by Median Width for All Accidents . . . . 165 Boulevard Accident Rate by Median Width for Rear-End Accidents 166 Boulevard Accident Rate by Median Width for Angle-Straight Accidents ............................................. 167 Boulevard Accident Rate by Median Width for Angle-Turn Accidents ............................................. 168 Boulevard Accident Rate by Median Width for Head-On Lefi Accidents .......................................... 169 Boulevard Accident Rate by Median Width for Driveway Accidents ............................................. 170 Boulevard Accident Rate by Median Width for Other Types of Accidents ........................................... 171 Boulevard Accident Rate by Speed Limit for All Accidents ...... 172 Boulevard Accident Rate by Speed Limit for Rear-End Accidents . 173 Boulevard Accident Rate by Speed Limit for Angle Straight Accidents ............................................. 174 Boulevard Accident Rate by Speed Limit for Angle Turn Accidents ............................................. 175 Boulevard Accident Rate by Speed Limit for Head-On Left Accidents ............................................. 176 Boulevard Accident Rate by Speed Limit for Driveway Accidents ............................................. 177 Boulevard Accident Rate by Speed Limit for Other Types of Accidents ............................................. 178 Boulevard Accident Rate by Average Daily Traffic for All Accidents ............................................. 179 Boulevard Accident Rate by Average Daily Traffic for Rear-End Accidents ............................................. 180 Boulevard Accident Rate by Average Daily Traffic for Angle Straight Accidents ....................................... 181 Boulevard Accident Rate by Average Daily Trafiic for Angle Turn Accidents ......................................... 182 Boulevard Accident Rate by Average Daily Traffic for Head-On Lefi Accidents .......................................... 183 Boulevard Accident Rate by Average Daily Traffic for Driveway Accidents ............................................. 184 Boulevard Accident Rate by Average Daily Traffic for Other Types of Accidents ...................................... 185 Boulevard Accident Rate for Cross-Overs .................... 186 Boulevard Accident Rate for Directional Cross-Overs .......... 187 Boulevard Accident Rates for Bidirectional Cross-Overs ........ 188 Tukey-b Test Example ................................... 194 xiii CHAPTER 1 DREE B 1.1 HISTORICAL PERSPECTIVE Divided highways provide a physical separation between opposite directions of vehicular travel. This separation of travel may create problems in accommodating certain traffic movements; such as, left-hand turns, U-turns, and access to property abutting the opposite direction of travel. The design of median openings (cross-overs) was first discussed by the American Association of State Highway Officials (AASI-IO) in 1954. A primary concern with this type of highway is the median width needed to accommodate U-turning vehicles. The criterion established in 1954, to allow a passenger vehicle to turn from the outer lane of one roadway to the outer lane of the other roadway or from the inner lane of one roadway to the outer shoulder of the other roadway, was a median width of 20 feet. A 30-foot median width was established for a single unit vehicle to turn from one roadway to the other while encroaching on both shoulders. A 40-foot median width was established to allow a large truck to turn from one roadway to the other while encroaching on both shoulders. A 60-foot median was established as the minimum median 2 width to allow all standard legal vehicles to turn from the inner lane of one roadway to the inner lane of the other roadway (AASHO 1954). Most early divided highways were designed with one of the above median widths, with one major exception. A four-foot median width, which was considered adequate to allow pedestrians to seek refuge while crossing a roadway, was used as a design criterion where pedestrian traffic was a consideration. As traffic volumes at these median openings increased, vehicles attempting to negotiate a left turn or U-turn would back up in the inner lane. The vehicle stoppage in the inner or fast lane presented an impediment to the traffic flow, thus increasing the potential for accidents to occur. The immediate solution was to provide a refuge or storage lane in the median. There were three problems with this solution: first, the storage capacity was limited; second, opposing turns became interlocked; and, third, the width of median available to accormnodate turns was effectively reduced. Traflic attempting to cross the intersection would occasionally become trapped in the through lanes because cars waiting to make a left turn were standing in the bidirectional median opening. The congestion worsened when left-turning trucks that were too long for the median opening occupied part of the through lanes. A solution developed by the Michigan Department of Transportation (MDOT) was to provide directional median openings back-to-back at a mid-block location, not opposite another street, and at the same time provide a storage lane for vehicles waiting to make a left turn (Figure 1.1.12). These directional cross-overs would often be used for accommodating left-turning vehicles from nearby signalized intersections. 3 There are several areas of concern regarding the operation of directional/bidirection mid-block cross-overs. The primary problem is the adverse travel distance required by the vehicles using the cross-overs to make the left-hand turn or reach the destination adjacent to the opposite direction of travel. Another problem is weaving maneuvers the turning vehicle may have to negotiate with thru traffic to reach the desired destination. This includes merging with the higher speed traffic in the inner lane while completing the turning movement. Also, lack of familiarity with the indirect turn procedure may also cause confusion resulting in accidents. A survey of fifty states and major Michigan counties (as a part of this research) found that only four other states and three Michigan county road commissions are known to provide median cross-overs as a means for making indirect left-turn movements. The questionnaire is shown in Figure 1.1.1 and the aggregated responses are shown in Figure 1.1.2. Because only four states use this geometric feature, the amount of experience with directional median cross-overs is severely limited and subsequently limits the data available for examination. An advantage of the boulevard cross-section is that lefi turns are limited to a few locations. The boulevard cross-section also decreases the opportunity for head-on collisions, as well as acting as a refuge area for pedestrians crossing the roadway. Most businesses do not feel there is an access loss and resultant business loss when a boulevard section is used (Metcalf 1989). STATE RESPONDING: ............................................. CONTACT PERSON NAME: ........................................ ZIP CODE: .................................................. PHONE NO.( ) ............................................ FAX NO. ( ) ............................................... 1. Does your state utilize directional cross-overs in your free-access, median separated roadways? Yes No 2. Are these used to accommodate left turns normally accommodated at an intersection? Yes No 3. If they are, what is the distance they are spaced from the center of the intersection to the center of the directional cross-over? feet 4. What is the minimum width median in which they are constructed? feet 5. Have you conducted any before and after accident or capacity analysis of roadways where you have used this type of facility? Yes No Figure 1.1.1 Survey of State-of-the-Practice Letter 6 The boulevard cross-section has the potential to reduce system-wide delay, increase roadway capacity, provide safety for pedestrians, create a positive impact upon development, and create a more aesthetically pleasing corridor. Some potential negatives include higher construction costs than for the CCLTL cross-section, reduced local access, weaving and merging with the high-speed lane, and increased travel time for left-turning vehicles (Hartman and Szplett 1989). The features of various roadway cross-sectiontypes can be examined in Figures 1 .1 .3 through 1.1.8. A rural roadway with a boulevard cross-section with flush shoulders and residential driveways is shown in Figure 1.1.3. The cross-overs in the median are bidirectional; i.e., the traffic can travel in either direction through the cross-over. The inherent, adverse distance that must be traveled and the weaving that is experienced by residents to access their homes are evident. An urban roadway with a boulevard cross-section with flush shoulders and directional cross-overs placed at either side of the intersection is shown in Figure 1.1.4. The directional cross-overs are intended for negotiating lefi turns that are prohibited at the intersection. Drivers traveling from left to right across the photograph wanting to turn left at the crossroad would first travel through the intersection to the directional cross-over, make a U-turn through the cross-over, travel from right to left to the crossroad, and turn right at the crossroad, thus completing the left turn desired. Similarly, a driver desiring to access business es adjacent to the roadway in the opposite direction would use the cross-overs to change direction and gain access. Evident in these maneuvers is the distance traveled and the weaving that must take place to complete the maneuvers. .8qu figogmom one 390-380 ~Eo§oh2m 5S9 nouoom Ego—non 35m MAJ v.5»:— a a :r. W $5.. I Mali; '1! ~. ~ News: w 028% 5;? 390-820 E03025 5:6 groom “5.638— 535 v44 «Burn .41 r413... 9 An urban boulevard section with curb and gutter and bidirectional cross-overs is shown in Figure 1.1.5. Drivers desiring to access the drives along the opposing direction of travel can use the bidirectional cross-overs to reverse direction and access the driveways. A similar urban roadway with curb and gutter, but with directional cross-overs and intersecting streets that do not cross the median, is shown in Figure 1.1.6. Drivers desiring to access the streets that intersect with the opposite direction of travel must use the directional cross-overs to obtain access. Drivers entering the roadway from the intersecting streets, desiring to travel in the opposite direction, must use a cross-over to obtain access to that direction of travel. A five- lane urban roadway with curb and gutter and with a continuous center lane which is intended to accommodate left-tum movements is shown in Figure 1.1.7. Access to driveways adjacent to the opposite direction of travel is obtained by moving into the left-turn lane prior to completing the turn movement. A similar roadway with curb and gutter and seven lanes of pavement, which includes a CCLTL intended to accommodate left-turn movements into a major cross-street, is shown in Figure 1.1.8. The operation is the same as that used in Figure 1.1.7, with vehicles needing access to the major cross-street being accommodated in the center left-turn lane. Heading up the lefi-tum movements prevents interlocking of left-turning vehicles. The current MDOT standards for median cross-over designs are shown in Figures 1.1.9 and 1.1.10. The standards that are used for designing bidirectional cross-overs are shown in Figure 1.1.9. The standards that are used for designing directional cross-overs are shown in Figure 1.1.10. Newly proposed MDOT standards for designing directional cross- oversare illustrated inFigures 1.1.11 through 1.1.13. 325.38 a: U u 2805.5 5.3 nouoom Ea>2=om as . . D Wu.“ 0.5»:— r.. is .. 1. 3.31.1. . . 3 ms'b't‘S-m‘ET 33933.3 - 5,: 330-380 2838on 30363.03 5 B .8308 BBS—non 52.5 e44 9:.» h 33?, u ,....3§§m§wwz . . 5:08.895 :3 Banana HS 00 83 3E :35 54.“ «Eur.— .Htaarviauama... 13 8308.380 053 ashamed 5250 0:3 -=o>om =35 w .2 2:5 14 935% “90-380 ~826§Em Susannah... .«o 8053qu ng£2 «J; «Eur.— _ fl... _ a... awn-EUR! .hiegbufilg $5388.! .bvxscgbuciaia. Samaria-«E4133... Eggfilscggg 111 Al AI .. IIIIIIIIIIIIIIIIII NH IIIIIIIIIIIIIIIIIIIIIIIII N.“ 111 My» .. Q. a: -IIIHW IIIIIIIIIIIIIIIIIIIIIIIII H IIIIIIIIIIIIII IV IV IIIIIII _ 3A, _ a... «BEER... $383525»; gimme“! .EBREuE-na Egcggg Bagel-calms... Al Al .. IIIIIIIIIIIIIIIIIIIII My” IIIIIIIIIIIIIIIIIIIIIIIII My”!!! .8.. 37 u teas W7 “Va: . 35 h Fig .8 a (a g -IIHWIIIIIIIIIIIIIIIIIIIIII IIHW IIIIIIIIIIIIIIIIIIIIII IV. 1V 15 uggm :o>O-mmo..U 3:28th Engages; .«o 325.339 53:32 3.3m ensure 'I'll-I'll|||||I|[||||l||'l|l|II «NI III Tm .fUHIIIIII gazdmNIIfiIrI\III|III_ szhg l6 283m “30-380 isotonic 6880.5 5335938.“. no 825.3qu Sumfioaz =44 oar.— .. i 8.6.. onTEom n 3232:. . Sm. 5282 onTEom u 8.9. x03: :5 29:5 N am Loocmmmom u a IIIIIIIIIAIIHI. 33 I IIIIIIIII . NJ ml. “.0me .ON .Om . .mm .Vm .NN .NL— Il.m b\ M" OCOJ .|\ . .H I Co... I 05.. lllll 4!. .5 .N@ .2 .mm .5 .Nm .~___ .m .~ EM MN 0”“; .I\ _. IIcmlIIII :0.— . ocog IIIII I )III. .55. .vm .Nw .Dm .mu .3. .N.—_ .N CVV MN ouooq . II . IIIIHIM :3 .2 .mm .9. .on .m— 30£o> c 300 go 5.9.0.. Efimfihzfi 8-9, _ mam _ am _ m Lo>aocoz +0 on».— . 3030‘, cmnnou Lou v00» .. coguoe no 5.2.. .5: u z wCLaTD Lo+ mcofioa EDGE—2 17 mmmflmflmm.) *mmumsmamnm, WIMWMWWWMWI \ Mar .1.) 6‘” I Figure 1.1.12 Michigan Department of Transportation Directional Proposed Cross-Over Standard 18 2355 B>O-mm80 38qu 3:28on coautoamqfih mo “magnum 39:22 24; Pin; .Bugggfigs .aaéghggiémssegégfiag .géghggsg ngegg gfiégéiggé 559.255.3338.. .E.§.E§ug 5.3.agaugiws Igniguazffiea I 58§§§s§.§ 7E§§§E Jam IIIIIIIIp I . ~ 19 1.2 THE PROBLEM MDOT began constructing boulevard roadways with directional cross-overs after 1960. An examination of the literature pertaining to cross-overs and left-turn movements has not yielded a formal study comparing divided highways featuring directional cross-overs with highways having CCLTL and evaluating the impact of either cross-section on capacity, safety, and level of service. Roadway construction with a boulevard cross-section is more expensive than a roadway with a CCLTL. The boulevard cross-section requires more right-of-way and traflic signing. There is concern that accidents may occur when vehicles attempt to cross traffic while trying to reach driveway or cross-street destinations adjacent to the opposite roadway. The type of cross-over needed to provide access to the other direction can become a controversial issue, as it may result in businesses becoming concerned with property access for their customers. For Michigan trunklines, the alternative to a boulevard or divided highway is a multilane highway with a continuous two-way center left-tum lane. This nondivided cross- section is usually provided as a five- or seven-lane facility. In addition to the acknowledged aesthetic benefits of a divided highway, a need exists to determine whether such construction yields sufficiently fewer accidents to justify the additional costs. 1.3 OBJECTIVE The objective of this research is to quantify the safety differences between divided highways and multilane facilities with CCLTLs. If there are benefits to be realized from a 20 divided highway, it is important to determine if they are sensitive to the median width. This is important because the right-of-way cost for a divided highway is directly related to the median width. It is also desirable to determine whether factors such as number of driveways, type of median cross-overs, number of intersections, and speed limits significantly affect the decrease or increase in traffic accidents. 1.4 RESEARCH APPROACH The method for comparing the boulevard cross-sections and the cross-sections is outlined as follows: 1. Review the literature and the two roadway types to determine which features have been reported to have an impact on roadway safety. 2. Conduct a thorough review of Michigan trunklines to establish the location and control section numbers of segments of the two roadway types. 3. Devise a method for data collection. 4. Collect and code the data in a manner suitable for computer analysis. 5. Apply statistical analysis techniques to the data to determine if there is a difference between the two roadway types under consideration and examine roadway features to determine their contribution to any differences detected. 6. Interpret the statistical data, draw inferences, and develop conclusions. 7. Formulate recommendations. 1.5 ORDER OF PRESENTATION The dissertation consists of the following parts: 21 The literature review and hypothesis are presented in chapter 2. The research approach and the details of the experimental design specifications are presented in chapter 3. The statistical analysis of the collected data and the accident data and their interaction are presented in chapter 4. This includes an evaluation of the independent variables, data type, and accident data. The inferences, conclusions, and suggested subsequent research are presented in chapter 5. The appendices contain the research data and statistical results. CHAPTER 2 The literature was surveyed to determine what previous investigations have been conducted on boulevard sections and multilane roadways with center left-turn lanes, especially with regard to accident experience. These findings are summarized in the following sections. 2.1 MEDIAN WIDTH A guide was found for the determination of roadway width and median treatment in semiurban areas. The guide delineated methods of dividing directions of travel which are varied. At one extreme is a pair of parallel one-way streets separated from each other by rows of city blocks. At the other extreme, all that separates the two directions of travel is a double yellow paint line down the middle of the roadway. In between these extremes are medians of varying widths, with and without curb and gutter, and those multilane facilities in which the opposing directions of travel are separated by a flush center lane which is used for left-turning movements only. Medians that are wider than 30 feet allow for the construction of directional cross- overs which, in effect, limit the number of crossing points. One of the disadvantages with 22 23 a median is that it results in an increased travel distance for those desiring to reach a property on the opposite side of the median. Similarly, those wanting to make left-hand turns at intersections where turns are prohibited would need to complete the movement using a directional cross-over. A positive aspect is that the left-turn capacity of an intersection that uses a directional cross-over can be nearly double that of a normal turn lane (author unknown 1 963). It further stated that CCLTL roadways with an 11- to lS-foot wide median do not provide the same traffic separation as roadways with wider medians, especially those with curbs. They do provide continuous access to abutting property on either side of the roadway. The author also stated that very high volumes of turns at intersections must be handled by phased signals, with a special phase for left turns to prevent the problem of interlocking. The heading up of this left-turn maneuver is a characteristic of a roadway with a CCLTL. It was reported that a road with a CCLTL provides for left-turning vehicles being stored out of the traffic stream. It does not provide the same security level for pedestrians as a divided roadway. Snow plowing is a consideration, as snow cannot be stored in the center left-turn lane as it can in a wide median (author unknown 1963). In a paper by another unknown author, assumed to be the same author (1963), the following characteristics of various medians were summarized: l. Widths of 16 to 22 feet are undesirable because U-turns are not feasible. However, these widths provide sufficient room for construction of left-turn lanes. 2. Widths of 23 to 39 feet are undesirable as they do not provide for U-turns. This width provides room to store passenger vehicles crossing an intersection, but it does 24 not provide adequate room for larger vehicles. (While the author uses the word vehicles, it is understood that what is meant is vehicles in excess of 1 6 feet in length.) 3. Widths of 40 to 59 feet cannot accomodate U-turns for all legal-sized vehicles. Only passenger cars and single unit trucks can negotiate the tight U-turns. 4. Widths in excess of 60 feet are most desirable, as the U-turn is feasible for all legal vehicles. Figure 2.1.1 provides guidelines for street width designs and median treatments. These were produced by the same unidentifiable source, and provide some insight to the thoughts on the subject at the time. It is interesting to note that the various median treatments, in the opinion of the author, do not significantly effect capacity. Although, the unknown author suggests that the number of accidents increased as the width of the median increased. He suggested in number 4 above that wider medians were more desirable. Cribbins, et al., in their investigation of median widths and openings in North Carolina, found that median widths are one of the factors having the least effect on accidents, although the accident rate for commercial vehicles did decrease as the median width increased. In another study in North Carolina, no correlation between accident rate and median width was found (Cribbins, Horn, Beeson, and Taylor 1966). The distance between the two stop bars on a minor road crossing of a divided highway was found to be a contributing factor to an increase in accidents (Van Maren 1978). It appears that the increase in exposure experienced, when crossing a longer distance, is a contributing factor to the increase in the accident rate. 25 as: .5553: 353: 20582:. 5:02 Ba “:23 .800 00 8855.20: 5 02:0 2." 2:0...— 80:80.08: 0.98:0:20:00 259,8: 2 0.85 0.825 :05 002 2 2:00:00 8:80 .8282: 53-3 2> 8 8:00.222 222:0 8: 80280.: .820 0808088 E85 8 802: - 83:2: :0:_>_: .080 830000: 25. 028500.. 2220088 08:3 88302 .co 0:2: :02 .53-: 008:: - Em: 8.: 2w: 2:050:02 E35 85 082 0.8 0:25-: 2028.805 .259, 8: :2: 28:0: :8888 to 0:8. :04 2 82:80: 2.8200: 002 0 2 22:02 .3 - .3. .2898 E8 :2 2 :022: 0n .2208: :8 2 0.: :2: :8 o 0: EB 8:282:22 32—2902. 8: 80:80—08: .08: 00 0: 6:2 25 .8 00.: 8:20 .8 0:80 02: 0:283:82 :8: 8 2 0E3 02:22.: 0:25 :2 08:3 08:: >8 .0» ‘3 8 2 0.: w2:80:o 328:2": =0 8 :23an c2 :8 :2288 880:0 .mEE c2 :8 00:2 2 2 2 8% a :028: xmomaao 0:2 0: 5:0 _ :5 0:25 :0..— .0w:.8._ 5:80.: - 0.8332: 8 8020 :280200 0:22 :S :3 :2: :2 :8: :280222 8 80:5 8.500: :8 0:308 8: 0009 0:222: :0:_>_::D .2288 53:2 2 :028: 0: .2208: :8 2 0.: :2: .8.: c 0: EB 8: 80:80—08: 20: 0 0: 6:2 25 .8 00: 8:20 .8 8:288:82 32—828: :8: 8 2 0:2: 02.222 0:2: :2 08:3 08:: 3&8» 3 8 2 gw22008 088 2:: 8280222 :2 :8 :2288 880% .083 :2 8.: 00:2 2 2 2 Saw a :022: 8080 0:2 328:2»: =0 8 :022208 023 .c - .v 82:08 - 08335 8 8020 8:00:80 0:22 .25 > :3 :2: :2 :8: :280222 8 8020 0: :28 .23 0:2: :01— 038282 8 :02: 0:0 22:00 .8 0m: 08: 80.: 226: :8: 8 2 0:2: :0.— .0830>t: 8 8020 .2880 53-02 83 as? 80 8:20 .8 08:: 30 .0» .3 8 2 0.: 22:80:: 2 28w 9 :022: 8280 0:2 22:00 :8: 2280202 8 80.20 8:288:22 82—220 0223 280: 8 3:222: 09 2 08: 8:. 0:2: :04 €22 :2 8.: 0:2 .8808 288:2: s :38 .2 - .: 80:28: 0: :00 80000 802: :2? 0.88.8.8» 082 .8 0000 2 80020 53-: :5 :88: 00 .mF—a 02 0:020: as? 80 82.05:. 88.: 2> 8 :280282 a 32:. - .330: 82% to was :3 .53-: 3:5 - 20: :2: 38:0: :8888 to 0:8: :02 3:3 a .8 mice: t. 2 80.: 83-28 6:26»: :2 03:00.. :80 .:2000.&08 8:20 .8020 .3580 .3 00: - ES? :2. :280282 8 8020 2 223:8: 8:288 00:22.— 0825 830:0 0:289:35 5050: Ms: :a‘<8_2:0k% aaaiozha 3 8000< a .00 00 08:28.. < 8.83:8:— 8 0:25. :04 8.830.: 80.5.8.5. 22:02 26 €28.82 2." £00.... 2.802: :0200 0: 8:8 80.80808. :0038: 80:80.08: 002.3 :00: 0: 8:80 82800.: 2:: 68.88:... 80:30:: :00 8202.0 .080 0 08:3 800:0 2:800:02 00080 8 00.082 88.20.. 8800.00 58-8. 0: :02:2. 0: ::0 2 0.: :2: .8.. o 0: .23 .080 00 m2 .000. 0.5 .8 00: 028.28 08:: :0. 08:3 .088 :0. .8.. 00:0. 0. 2 :280800 0:28 .25 own 20:»... .8 000:: 3&8.» .2: .8 2 0.: M28008 2 2.0m : :02:2. .mm0280 .28. 02.3.8. 80.: 00.80222 8 8020 80.80822 :0038: :0200 0: 8:8 822.208: 0.823 0.:0200::D .083 .2 2 .0 02.802: :38... 0: 8:8 80.80822 :0038: 80:80.08: 002.3 :00: 0: 8:80 80:80.: 2:: 68.88:... .:0:.>08 0: :00 08.0 .8 82:02.0 8.00% 002.3 80280. .0882 00.080 8 00.0 2 88-201. 5.00:8 53-02 3.0 .05.... 80 ..0:w.m .8 000:: 32.0.2 .3 8 050.: w2800:0 2 0:0w an :028: b.0800 00.00.: :000 0 2000.0 2 :02:2. 2.080220 08.0 2 000% own—08 - 00280808. 8 08.0 88-20. 0. > 80.80222 :0038: :0>._0m 0: 8:8 828.208: 002.3 2:02.008: Moran .NN I .Cu .008:.0> :2: 080. M23 8:02:88 2 .2 ..0.00%0 8:282:22 :0038: :0>:0m 0: ...3 0288:0w 2&0: 80:22:. :05 088 00023 :00: 0: 8800 8.82:2. 0.5 .8 08000:- :0>:0:08 :2: .3 028.200.. b0> .Am .20. a: 00:83 088 :0. 0: 80.3% 8: ea 02 05022.. 550., 80 .050: 0.008 ::0 00:0. :2: M2082: :: 0.00. 00:: :0. .:0:000x0 8.0080 0.823 .0.:0..0>0 000% 0w088 .2: :02:2. @8080 0:28:20 0.00.008. m0 000.: 00.8 088 :0: 2:00.83 8: 08% :02:2. 382.00 8828. -28: 00: 088-3 0000088820 880% b0> :0::: 80020 0302008: .2828 :0..08: 00m 82:08 8m - .mN .328 I. b. .00.. 203.08 .8203: .8: 03:00.. :80 08:50 .3 00: - ..:S:> ooh 80.80808. 8 802m. 8280808 8:20 a. 0.8.8: 5:285 80:20 300.5 203.80 80.800008. 00038.. 08302.5 0. 0.0003. 8:0 :%.0.:0m% 0580 00 20:28.. < :28888— «a 8.59 «.81— 9.0083— 808895. 02:02 27 As a result of several years of study, the State of Illinois adopted a median width of 40 feet for rural highways and a median width of no less than 6 feet in urban areas where turns are allowed (Hutchinson 1966). Hakkert and Ben-Yakov raised the issue of pedestrian delay at signalized intersections. They state that the issue is important because, as pedestrian delay increases, the pedestrians increase the amount of crossing on red signals, thus decreasing pedestrian safety. They propose that when the signal cycle length is minimized the pedestrian delay can generally be reduced. They also recommend that for divided highway sections the minimum green period be extended to allow the crossing pedestrian to begin the second leg when the phase is still green (Hakkert and Ben-Yakov 1958). As previously noted the divided or boulevard cross-section does provide a refuge for the pedestrian in the median, thus allowing the pedestrian to cross the roadway in stages during more than one signal cycle. A median accident study was conducted on several sections of highway in Long Island, New York, to determine the effect of median design on accident rates for urban, multilane, and free-access highways with average daily traffic (ADI) volumes up to 44,000 vehicles. The median designs were classified as either deterring types or nontraversable types. A deterring median was defined as one which discouraged deliberate crossing by obstructions; such as flush and raised with intermittent shrubbery, mountable double curb, or earth medians with flat cross slopes. A nontraversable median was defined as one which prevented crossing from one roadway to the other by obstructions; such as, concrete posts or earth medians with steep cross slopes and guardrail. The results of the study are as follows: 28 The ratio of night-to-day accident rates between intersections averaged 3.5 to 1 for the nontraversable group and 2.0 to 1 for the deterring group of medians. The night traffic volume was determined to be 25 percent of the 24-hour traffic volume. The accident rates between intersections increased linearly with traffic volumes for deterring-type medians. For deterring-type medians, there appears to be no correlation between median width and accident rates even though the accident rate peaked for the 15-foot median with medians less than 15 feet and more than 15 feet having a lower accident rate. The authors did not have sufficient data to determine the relationship between median width and accident rate for nontraversable medians (Figure 2.1.2). The overtaking type of accident accounted for more than 70 percent of accidents between intersections for both deterring and nontraversable median types. In the deterring-type median, the grass-flush median had the lowest rate for all accidents between intersections; the curbed-type median had the highest rate. When the effect of volume was taken into account, the results did not change. In the deterring group, both the earth and miscellaneous features median had the smallest percentage of severe accidents. For the nontraversable type, the double guardrail had the smallest percentage of severe accidents, while the concrete posts had the highest percentage of severe accidents. The accident rate, property damage frequency, and severity of accidents were all much higher for medians with concrete posts. The frequency of all accidents involving the median was higher for the nontraversable type than the deterring type. On highways with deterring curbed 29 if I I I ! /\ WI" l f” / nI 34V | ""1""- *- 3’1...» a so do no 3 l l g maximum 5'” 3 E \ an \- __ °1333I¢ to so «a 45‘ mm 0' not“ I. "I? Figure 2.1.2 Relationship of Median Widths to Accident Rates (Billion and Parsons 1953) 30 medians and nonillumination, the nighttime accident rate at intersections was twice that of the day rate; whereas on highways with illumination, the night and day intersection accident rates were the same. The study also investigated the accident rate between intersections with and without illumination; and it was found that illumination apparently has no effect on accident rate between intersections (Billion and Parsons 1953). In a California median study, Telford, et al., investigated traversable, deterring and nontraversable medians. They found that the accident rate for traversable and deterring-type medians did not vary much for medians above 10 feet in width. The accident rate for nontraversable medians increased at a rather uniform rate, as the median width increased for medians greater than 25 feet. It was also found that there was no relationship between accident severity and median width (Telford, et a1. 1953). In both the Long Island, New York, and the California median studies, the significance of the difference in accident rates for the various median widths was not discussed. In a synthesis prepared by the Transportation Research Board, it was reported that "The median width is unrelated to the accident rate." (TransportationResearch Board 1978). 2.2 CROSS-OVER TYPES As stated in the AASHTO design guides, highway efficiency, safety, speed, operation cost, and capacity depend on the general highway intersection design. The main objective of intersection design is to reduce the number of accidents and to provide a more convenient route of travel. Human factors, traffic considerations, physical elements, and economic factors must be considered when designing at-grade intersections (AASHTO 1990). 31 MDOT studied the effects of replacing a bidirectional cross-over with a pair of directional cross-overs at the intersection of US-l 0 and Opdyke Road. US-10, a major route in Michigan, intersects at a tee with Opdyke Road. At this intersection, US-10 was four lanes in each direction separated by a 40-foot median with cars traveling at 50 miles per hour (mph). Opdyke Road is a two-lane, two-way roadway with a speed limit of 35 mph. The following changes were made to the intersection between May 1, 1967, and September 1, 1967. The existing bidirectional median cross-over which was aligned with Opdyke Road was closed. Improvements to the intersection included the construction of two directional, median cross-overs northwest and southeast of the intersection and the construction of median, deceleration lanes for northbound US-10 traffic at both cross-overs. An additional lane was also provided for northbound US-10 traffic exiting at Opdyke Road and for channelizing the intersection. Accident data for the year following the improvements were compared with accident data from the year before the improvements. Total accidents decreased 62 percent, from 34 to 13, and injuries decreased 70 percent, from 20 to 6. Median accidents were completely eliminated. There were no far-side, right-angle accidents at the north cross-over and only two right-angle accidents at the south cross-over. The project met its goal of reducing the accident hazard, especially median and far- side patterns. It also increased the efficiency of lane usage on northbound US-l 0. The cost of the improvements was $67,000. The benefit for the first year alone was calculated to be $47,300 (Hoffman, et a1. 1960). 32 The accident data for this example were limited to one year, but they did provide a view of the results of one of the first directional cross-overs constructed in conjunction with other improvements in Michigan. 2.3 CROSS-OVER SPACING A study performed in California found that the lower-volume crossroads had a higher accident rate than higher-volume crossroads. The author concluded that increasing the crossroad spacing within reason, which in effect reduces the total crossroads so that volumes that would be concentrated at the remaining crossroads would be beneficial (McDonald 1953). This indicates that it may be beneficial to limit the number of cross- overs; therefore, increasing the volumes on the cross-overs. It should be noted that the accident count was accidents per year, not accidents per million vehicle miles. Cribbins, et al., in a study that surveyed median cross-over spacings, that varied from 1 cross-over per mile to 27 cross-overs per mile, found that as the median cross-over spacing increased the accidents increased. The North Carolina median placement policy is to space median cross- overs at a minimum of 1,300 feet (Cribbins, et al. 1966). It was also concluded in this work that a precise optimum median cross-over spacing could not be determined. A Federal Highway Administration publication on the Safety Effectiveness of Highway Design Features, published in November 1992, cited a report to the United States Congress that addressed access control. The report to Congress, which analyzed accident data from 30 states, concluded that full access control was the most important single design factor for accident reduction. This conclusion was sustained by the results of comparing interstate performance with performance on primary highways that allow free-access. Many 33 of the findings agree with those of other studies, that the frequency of median openings directly affects the accident rate. The authors conclude in this work that access should be minimized wherever possible (Cirillo 1992). 2.4 DRIVEWAYS It was found in a study by Cribbins, et al., that, as access points in North Carolina increased, the number of median openings had a significant effect on accident potential. It was also determined that, as access points increased, all types of accidents increased (Cribbins, et al. 1966). The literature concerning the relationship between accidents and driveways for roadways with boulevard cross-sections is limited. The balance of this section reviews literature concerning driveways on arterial and two-lane highways. While these findings may not be directly applicable, they do give an indication of what might occur to roadway operations and accident history as a result of this roadway feature. Schoppert found that the driveways and intersections were directly related to the number of accidents at all ADT levels (Schoppert 1957). In a work by Box, it was recommended that criteria be established for driveway radii and driveway design for one-way and two-way operations. Since the publication of the article 27 years ago, many of the criteria have become part of state and local standards (Box 1969). In another study in 1969, Box compared driveway street connections to intersections, suggesting that most driveways can be considered as very minor intersections. Using the rationale that 83 percent of all driveway accidents occurred on major streets, Box carried out 34 a two-year study (1961 and 1962) on 39.7 miles of major streets. Box's research found an advantage to roads that have barriers (medians) as opposed to those without barriers in terms of frequency and severity of accidents. Since 60 percent of the accidents involved lefi- turning movements, this statistic would explain the lower accident rates when barriers (or medians) were used. Commercial driveways represented the primary hazards (Box 1969). In a 1970 Louisiana study, a correlation was found between accident rates and pavement cross slopes and conflicts per mile, the conflicts being defined as drives and intersections (Dart and Mann 1970). The variation in accident rates for pavement cross slopes was small with no test for significance being provided. It was found that driveway accidents accounted for 4.2 percent to 13 percent of the accidents reported, in a report by Uckotter. The purpose of Uckotter's research project was to determine the significance of driveway accidents on commercial arterial highways; to examine roadway, driveway, and land use characteristics related to driveway accidents; and to determine which factors had significant effects on driveway accidents. Data for the study were collected from 14 roadway sections. Accident data during a three-year period from January 1, 1969, to December 31, 1972, were obtained from the Indiana State Police accident records. Stepwise regression analysis was performed to measure the level of interaction among factors used to predict driveway accidents. A model was then developed to predict the number of driveway accidents per mile per year of a roadway section. The ADT, the number of medium plus high volume driveways per mile, and the driveway volume per mile explained 82.6 percent of the variability of driveway accidents per mile per year. 35 Driveway accidents were found to represent one-third of all traffic accidents on those sections of arterial highways serving commercial land uses. Driveway accident rates were notably higher at driveway volumes between 250 and 800 vpd. Above 800 vpd, driveway accident rates were lower than those driveways with lower volumes. It was postulated that the main reason for a lower accident rate at higher volume driveways was that the entrances and exits were better designed (Uckotter 1974). In a report of a study by J. C. Glennon, et al., an outline for use by highway agencies concerned with controlling direct access on arterial highways was developed. A discussion of the problems that are associated with control of direct access and the circumstances around driveway conflicts was given. Some driveway accident statistics from 1970 were included in the report and were broken down into several different classifications. The legal aspects and current practices used for direct access control were also discussed. The second part of the Glennon, et al., report was intended to orient the user to the identification and analysis of access control techniques. First, the techniques were identified and classified; there were over 70 classifications. Multiple evaluations then followed, such as legal, technical, economic, operational effectiveness, cost effectiveness, and development of warrants for implementation (Glennon, et al. 1975). McGurick and Satterly stated that full access control is not the solution to reducing the number of accidents that occur on urban arterial highways, as access must be provided. The report identified some of the characteristics of driveway accidents and driveway-related accident occurrence to various physical and environmental features of the roadway and traffic characteristics. A statistical analysis of the accident data showed that the driveway accident rate tends to decrease as the spacing between two driveways increases and the 36 spacing between a driveway and an adjacent intersection increases. Multiple regression analyses identified the following relationships: the driveway accident rate decreased as the number of commercial driveways per kilometer (km) decreased, as the number of through lanes decreased, as the number of total intersections per km increased, as the number of total driveways per km decreased, and as the traffic volume on the arterial highway decreased. This study was intended to give the public official or engineer a tool to better understand the circumstances related to driveway accidents, to predict driveway accident rates, and to estimate the effectiveness of measures to reduce driveway-related accidents (McGurick and Satterly 1976). In a study addressing driveway access control to arterial streets, Bochner stated that some of the long-term effects of poor access management are a high accident frequency and erosion of roadway capacity. Bochner's work represents an ordinance-based approach to arterial driveway access control. Bochner stated that "the capacity of a four-lane arterial street with a 45 mph speed limit will be reduced by over one percent for every two percent of the traffic that turns between the right lane and driveways at unsignalized driveway locations." Statistics for several accident types that were introduced by the conflicting traffic movements of driveways were presented. Included in Bochner's report was a table showing the recommended minimum driveway separation for different speed limits on arterial roads. There are several factors that should be considered in ordinance provisions, they included: corner clearance, number of driveways, number of lanes per driveway, prohibition of turns, and left-tum lanes. The author concluded that each ordinance provision would require 37 technical justification as the legislative process is used to adopt ordinance provisions (Bochner 1978). A National Cooperative Highway Research Program synthesis of studies on the cost and safety of highway design elements resulted in the following conclusions: 1. "As the presence of roadside features increases, the intersection accident rate increases." 2. "As the total width of driveways within 200 feet of the intersection increases, the intersection accident rate increases." (Transportation Research Board 1978). In a Texas study based on accident data from 1975 to 1977, Rogness and Richards found that 16 percent of all traffic accidents were driveway-related. It was suggested that a significant proportion of accidents are influenced by driveway design features, vehicle operating characteristics, and uncontrolled traffic movements at driveways. Approximately 95 percent of the driveway-related accidents occurred in urban areas. The severity was found to be 34 fatal accidents in every 10,000 driveway-related accidents. Motorcycle and truck accidents made up 72 percent of all fatal driveway accidents (Rogness and Richards 1981). 2.5 TURN-RELATED ISSUES Left-turn accidents at driveways comprised the largest percentage of driveway accidents. It was reported that conflicts from entering vehicles caused those accidents (Box 1970) Before-and-after accident information for isolated intersections where directional cross-overs were constructed were evaluated by Savage. The construction of directional 38 cross-overs appeared to have a significant effect in reducing the number of accidents, particularly right-angle accidents. It was recommended, however, that directional cross- overs be used only where there are significant operational problems that need to be corrected (Savage 1974). A National Cooperative Highway Research Program synthesis consisting of studies on the cost and safety of highway design elements, produced the following conclusions: 1 . "The presence of left-turn lanes at two-phase signalized intersections can increase the accident rate." 2. "The accident frequency at divided highway grade intersections is inversely proportional to the median width." (Transportation Research Board 1978). In a study by Higgins, et al., where the objective was to identify the safety and operational impacts of median acceleration and storage lanes and to assess the cost- effectiveness of current design applications, the following aspects were considered: geometric design practices, effects on traffic operations, safety impacts, driver understanding and acceptance of the designs, and costs. The categories that were closely examined were: median deceleration, storage, acceleration lanes, and CCLTLs. A survey of highway and traffic agencies in Canada and the United States was conducted. Among the survey results was the conclusion that median acceleration lanes were not extensively used in Canada or the United States. However, the CCLTLs were found to be used extensively in the United States and were also quite popular in Canada. In CCLTL design, the speed limit ranged from 25 to 60 mph, and 12 feet was clearly the most popular width. There was a marked decrease in accidents where CCLTLs were implemented. Accidents involving left turns from the main road decreased by 36 percent. 39 Accidents involving left turns onto the main road decreased by 22 percent. Accidents involving a vehicle turning off with one turning on decreased by 36 percent. It should be noted that total accidents, not accident rates, were cited in the research. According to the data obtained from the survey, there was a decrease in delay and congestion in areas where CCLTLs were implemented. Some guidelines for CCLTL usage suggested that they were best suited for strip development but required effective signing and marking to reduce the vehicle operator’s misunderstanding, indecision, and misuse (Higgins, et al. 1984). AASHTO's 1990 edition of the WWW provides a national guideline for accommodating turning movements. The median width needed to provide for efficient and safe indirect lefi-turns/U-turns depends on the required turning paths of the design vehicles. If a road has a small median, a jug-handle type ramp or a loop design can be used to improve efficiency and safety around the turns. Highways without medians must use local streets with interconnecting patterns, special median openings, jug- handle type ramps, or an at-grade loop to make indirect left-turns or indirect U-turns. Highways with wide medians that use a cross-over for these turns can be confusing, but do reduce congestion. U-turn median cross-overs are used for left turns in high traffic areas. CCLTLs should only be used in urban settings with low speeds and no more than two lanes in each direction. Median left-turn lanes are auxiliary lanes for storage or speed changes. They are necessary in reducing accidents, inconvenience, and lost efficiency in high traffic or high speed areas (AASHTO 1990). 40 2.6 TRAFFIC VOLUME Cribbins, et al., in a study conducted in 1966 in North Carolina, concluded that "in the development of a predictive equation for accidents that ADT and roadside access combined with the number and type of median openings are the most important predictors of accident rates." In a boulevard intersection accident study performed in California, McDonald found that divided highway roadways with less than 5,500 ADT were unusual, and the roadways with the lower volume were removed from the study (McDonald 1953). Accidents on lower volume roads with less than 5,000 ADT were found to have little correlation with any roadway element (Schoppert 1957). It was found that accident predictions were more accurate for roadways with 3,000 ADT or greater (Schoppert 195 7). 2.7 RAISED MEDIANS In a California study performed by Telford, et al., it was found that traversable medians had a lower accident rate than deterrent medians (curbed being included in this category) for traffic volumes less than 15,000 ADT. When the traffic volumes exceeded 15,000 ADT, the nontraversable median proved to be more effective. This was generally true for all levels of severity (Telford, et al. 1953). Metcalf discussed the pros and cons of both raised and flush medians which are paraphrased in the next two paragraphs. Raised medians are permanent, and this permanence has both positive and negative aspects. For instance, if the left-turn lane is too short for the required storage, it is costly to correct the problem by lengthening the turn lane. With flush medians, correction can be 41 made by repainting. Raised medians also separate traffic flows reducing, but not eliminating, the possibility of head-on collisions. Raised medians force left-turn movements at predetermined locations, thus helping in access management. They can be made much more aesthetically pleasing and provide a safe location for both pedestrian crossing, traffic signal and light poles, and other traffic signs. Improper placement of signs on raised medians, however, can be a maintenance nightmare. Adjacent businesses often do not like raised medians because they feel they are being "cut off', which is not necessarily true. A raised median will not discourage the motorists from making the necessary movements to go where they want to go. Flush medians cannot completely control left turns; they are not "self-enforcing" like raised medians. Flush medians are not aesthetically pleasing. Although flush medians are more economical to install, the concern for maintenance and the safety of the maintenance crews should be considered. Flush medians are difficult to see in rain or snow, or when covered with dirt or sand from snow removal and ice control operations. Metcalf concluded that flush medians are probably best in rural areas and raised medians are best in urban settings where speeds are lower (Metcalf 1989). 2.8 CCLTLs Ohio State University carried out a project to evaluate the effectiveness of CCLTLs as an access control measure. Results of a nationwide expert opinion survey and field studies were included in this report. The following is a brief summary of the report: CCLTLs included in this project were mostly common in commercial land use areas. Positive effects of these left-turn lanes on traffic operation were evident throughout the 8,000 42 to 31,000 ADT range. Once CCLTLs were installed, reductions in accident rates were observed. The speed limit ranged from 25 to 45 mph. One, two, or three through-lanes in each direction were found acceptable, with the five-lane section most common. These findings are consistent with the W W, except that AASHTO does not recommend a center left-tum application for situations with three lanes in each direction. The recommended width was that of a normal left-tum lane. CCLTLs could extend directly up to minor intersections, but separate left-tum lanes should be placed at major intersections. Total accidents and injury accidents decreased after the addition of CCLTLs, especially rear-end, side-swipe, and mid-block left-turn collisions. Also, CCLTLs have not been shown to lead to an increase in head-on collisions. Seventy traffic engineers with practical experience with CCLTLs responded to the survey. They reported that CCLTLs significantly increased the traffic flow quality and reduced accidents. Fears that the lane would produce severe head-on collisions had been proven unfounded. Public reaction was favorable. The effectiveness of CCLTLs increased with the appropriate signing and was also enhanced by a positive effort to educate the public. General guidelines called for operating with 10- to lS-foot lanes, with an optimum width of 12 feet. The use of CCLTLs was applicable in residential, as well as commercial areas, and cross-tum storage was considered a proper function of CCLTLs. Three sites were used in the field study. US-20 between Painseville and Mentor was changed from a four—lane to a three-lane roadway. The beneficial effects of the CCLTL in terms of traffic flow were offset by the elimination of one traffic lane in each direction. Travel time increased along with weaving, and there was a reduction in accidents. SR. 264 in Cincinnati was changed from a four-lane to a five-lane highway. Travel speeds didn't 43 change significantly, but at higher volumes the benefits of the CCLTL were more obvious. Braking conflicts were reduced. US-42 north of Mansfield changed from a two-lane to a three-lane highway. Average total travel times and average delays were reduced. There were improvements in running speeds, as well as a reduction in accidents. Braking conflicts were reduced 37 percent and weaving was reduced 80 percent. The Ohio State University study demonstrated that the two-way, left-turn median lane is effective in improving traffic flow and in reducing left-turn-related accidents. It also lessens the severity of accidents (Nemeth 1976). In a 1983 paper, David P. McCormick and Eugene M. Wilson compared the operational effects of roadways having continuous two-way, left-tum lanes with four-lane sections and five-lane, Z-pattem sections. The purpose of their work was to determine under which circurnstancesa particular alternative would produce the best possible results from the standpoint of movement efficiency and safety. A rather small sample size of seven locations was used. In their conclusions, McCormick and Wilson state that CCLTLs had a substantially lower mean conflict rate than either the four-lane or the Z-pattem treatment. U-turns were virtually eliminated with the use of the CCLTLs. The recommendation of McCormick, et al., for use of the CCLTLs along strip development agrees with other studies (McCormick and Wilson 1983). 2.9 ROADWAY GEOMETRY Cribbins, et al., determined that in North Carolina: 1) accident rates could be accurately predicted for boulevard roadways; 2) the number of rear-end accidents is less 44 where storage lanes are provided; 3) rear-end accidents occur more frequently at unsignalized intersections than at signalized intersections; and 4) 35 percent of accidents between intersections involve median openings (Cribbins, Avey and Donaldson 1967). An accident analysis was conducted by B. Beukers and J.H. Jenezon on three types of intersections in the Netherlands. The three types of intersections are: MA: An intersection where the main road is not divided by any median and does not provide a left-turn bay. M: An unsignalized intersection where the main road is divided by a median and has left-turn bays. 133212: Same as Type B except the intersection has a traffic light at the intersection. The accident analysis, which included 18 intersections of Type A, 20 intersections of Type B, and 13 intersections of Type C, was performed over a period of two years. The ADT for the intersections were 7,500, 8,000, and 19,000 vehicles for Types A, B, and C, respectively. For the 24-hour periods studied, Type B intersections had the highest percentage (69.5 percent) of collisions that involved crossing or oncoming traffic, while Type A and C intersections had 46.9 percent and 41.4 percent, respectively. As expected, the ratio of rear- end collisions were highest at Type C intersections with 35.1 percent. There were 26.0 percent and 7.6 percent at Type A and B intersections, respectively. However, when the lights were turned on to the flashing mode for Type C intersections, rear-end collisions dropped to 18.8 percent and collisions at the intersections involving crossing or oncoming traffic increased to 55.0 percent (Beukers and Jenezon 1972). 45 A study was conducted in the early 19703 to determine the relationship between roadway geometry and accidents. Kentucky rural accident and traffic volume data from 1970 to 1972 were collected. The highway sections studied were divided into the following categories: two-lane, three-lane, four-lane undivided, four-lane divided (no access control), and interstate and parkway. The following observations were made: 1. Four-lane undivided highways had the highest accident, injury, and fatality rates. This was attributed to the higher volumes and number of conflict points associated with this roadway type. When the highway was divided, the accident rate dropped sharply and injury and fatality rates declined. This was attributed to the fewer conflict points that exist on this roadway type. When access and at—grade intersections were eliminated, such as on interstates and parkways, the accident rate reached a minimum. The critical accident rate is defined as that rate above which a hazardous situation is believed to exist. Two-lane highways had the highest critical rate in terms of million vehicle miles (MVM) of travel. Four-lane undivided highways had the highest critical rate in terms of accidents per mile per year. Rear-end or same-direction sideswipe accidents were the most frequent types of accidents for all highways as a group. For three-lane, four-lane divided, and four- lane undivided highways, rear-end accidents were most common. Single-vehicle and rear-end accidents were most common on two-lane roads as well as on interstate routes and parkways. Interstate routes and parkways had a significant number of head-on and sideswipe accidents, while four-lane divided and undivided highways had a significant percentage of angle accidents (Agent, et al. 1975). 46 A 1979 investigation by R. J. Taylor cited accident statistics for a highway which had been converted to a CCLTL roadway. The total number of accidents was reduced by 35.9 percent. Rear-end accidents declined 45 percent; left-turn 20 percent; head-on 67 percent; opposite-direction sideswipe 100 percent; parking 79 percent; angle 40 percent; same- direction sideswipe 50 percent; pedestrian/bicyclist30 percent; and fixed objects 65 percent. The run-off-the-road accidents increased by 86 percent and backing-related, 125 percent. 2.10 LITERATURE SEARCH SUMMARY The literature search found general agreement among researchers on the subject of boulevard roadway operation. There was agreement that the median width did have an effect on the accommodation of vehicles making U-turns through median cross-overs. The authors differed on how to group median widths for analytical purposes, but generally it was at 0 to 30 feet, 30 to 60 feet, and over 60 feet of median width. There was disagreement as to whether accident rates increased or decreased as median width increased. In two studies, opposing results were obtained. One study found that the accident rate increased with increasing median width and the other found that the accident rate decreased with increasing median width. One study reported that the directional cross-overs were effective in reducing accidents, but the study that indicated these results had only two years of data, which may be insufficient to support that conclusion. The use of CCLTLs appeared to improve the operation of urban arterials. The use of the boulevard roadway section with indirect left turns appeared to improve operations. No comparison between the two roadway types was found in the literature. 47 The studies examined also indicated that roadways with embed medians had a higher accident rate than roadways with uncurbed medians for traffic volumes under 15,000 ADT. On roadways with higher traffic volumes, the roadways with curbed medians had a lower accident rate. No published research was found which compared the Michigan-designed boulevard with five- and seven-lane CCLTL facilities. Studies where low traffic volumes were used resulted in unreliable statistics. The defrnition of low traffic volume was 3,000 ADT in one study and 5,500 ADT in another study. Several studies dealt with the relationship between signals and accidents. The general conclusion is that the stop-go signals did not reduce accidents, but merely changed the type of accident that occurred at the intersection. 2.11 RESEARCH OBJECTIVES The literature reviewed in sections 2.1 through 2.9 suggests that there may be a difference between the accident rates of roadways with a boulevard cross-section and those of roadways with CCLTLs. The research hypotheses to be tested in this study are formulated as follows: 1. The accident rate for a roadway with a boulevard is different than the accident rate for a roadway with a CCLTL. 2. The accident rate for roadways with a boulevard cross-section is influenced by geometric features and roadway environment; i.e., median width, posted speed, and average daily traffic. 3. There is a difference in the accident rate for a boulevard with directional cross-overs and a boulevard with bidirectional cross-overs. CHAPTER 3 3.1 THE VARIABLES OF INTEREST This dissertation is part of a cooperative research effort between Michigan State University and MDOT. MDOT was interested in investigating the safety aspects of divided highways with respect to the geometric design features. MDOT was also interested in the ability to predict accident rates for boulevard highways as a function of the geometric features so that they could develop cost-effective designs. The literature search revealed extensive research and investigation of roadway features that attempt to explain the difference between the accident experience on roadways with CCLTLs and that on roadways with a boulevard cross-section, but information on the Michigan design was limited to only one location studied in the 1960s. The conclusions of the researchers were not always in agreement even when the same roadway features were the subject of the research. The hypotheses regarding the safety aspects of a boulevard cross-section will be tested by examining the relationship between the roadway elements and the accident data. As there are several roadway features that could affect the accident rate, multivariate 48 49 analyses will be performed to test the relationship of the independent variables with respect to the dependent variables. The analyses require four types of data: 1) the physical roadway features; 2) the accident data as recorded by the local jurisdictional authorities or Michigan State Police; 3) traffic data; and 4) traffic control data. The details of the proposed analyses are found in section 3.8. 3.2 THE RESEARCH DATA Roadway data were obtained from the MDOT roadway inventory. The records for all five-lane, seven-lane, and boulevard roadways in the State of Michigan were obtained. The values of the independent variables present on each of the roadway sections were obtained from the following: MDOT photo-log system, plan of record, Planning and Traffic and Safety data files, and a field check combining video tape and measurements. The published official traffic volumes and information concerning presence of a curb were obtained from the MDOT W113 manual dated 1990. The Michigan county population figures were obtained from the State of Michigan Library compact disc records of the WWW 3.2.1 DATA INDEX MDOT uses a numerical system to inventory the state highways. Each control section is further subdivided into mile points which are measured in hundredths of a mile and the numbering increases from south to north and from west to east. The state roadway inventory shows the control sections of the state highways that fall into the CCLTL and boulevard categories. 50 MDOT photographs all of the state highways on a periodic basis (until recently). The original 35 mm film process has been upgraded to a videotape process and all records are now saved on an optical disk. These records, with additional videotaping and field measurements, were used to define the roadway features in this research. The equipment and procedures are described in section 3.6. The accident data are formatted with standard data fields by year. The method of categorizing the different routes has been to assign a (MALI) prime nmnber to each roadway segment. The prime ntunbers and mile points may or may not correlate with the photolog mile points. The values used to report accidents are derived from maps and other horizontal measurements, while data derived from the photolog are based on driven distances. Thus a small error is introduced and the error accumulates as the distance from the origin increases. To resolve this discrepancy, the English name of each intersection is coded from the photolog and subsequently matched with the corresponding name from the MALI. A simple linear transformation is used to provide each accident with a photolog-based mileage point. The data were assembled into working computer files while data collection was proceeding. Each data set was inspected and errors that were detected were verified on the photolog or in the field and then corrected. The accident data, obtained from the University of Michigan TransportationReswch Institute, were converted from a tape format to a compact disk format. This enabled quick access to the data via a desktop computer which was especially convenient for both statistical analyses and plotting. The roadway data were then merged with the accident data into one computer file for ease of inspection and analysis. 51 3.2.2 ROADWAY FEATURES The following roadway features have the potential to affect the accident potential of roadways with CCLTLs and boulevard cross-sections and can be a practical consideration in a roadway design: 1. 2. 10. Intersection density: The number of intersections per mile of roadway. Signal/no signal: The presence or absence of stop-go traffic signals. Signal density: The number of signals per mile of roadway. Driveway type: The classification of the driveway, generally defining its primary use such as residential or commercial. Driveway density: The number of driveways per mile. Lane width: The width of the individual traveled lane measured in feet. Shoulder/curb: The method of treating the area beyond the edge of pavement either with a curb or a shoulder. Median width: The width of the median measured from edge of metal to edge of metal. Cross-over type: The type of cross-over provided in the median. Directional cross- overs permit travel only in one direction from one roadway to the other. Bidirectional cross-overs permit travel in either direction from one roadway to the other or across the median. Cross-over density: The number of cross—overs occurring per mile. 52 3.2.3 TRAFFIC CONTROL DATA The following traffic data have the potential of affecting the accident potential of CCLTL and boulevard roadways: 1. Traffic volumes: The number of vehicles on a particular roadway could directly represent the increased potential of conflict between vehicles which are traveling along, attempting to enter or perform some other maneuver on that same roadway. 2. Percent commercial vehicles: The percentage of commercial vehicles represents a portion of the motor traffic which requires certain accommodations, such as adequate turning radii, to participate in the traffic stream. Commercial vehicles include all buses, single-unit trucks, and truck combinations except for light delivery trucks. 3. County population: The population of the county in which a roadway is located. This may be a surrogate measure of development intensity that is present along roadways in a particular geographic area, which in turn could reflect the cross-street volumes. Data on cross-street volumes are not available. 4. Speed limits: Speed limits are traditionally based on the 85th percentile running speed of the roadway in question. Speed limits may also be a surrogate measure of developmental density and/or traffic congestion. 3.2.4 THE ACCIDENT DATA The accident data were fiirnished on magnetic tape by the University of Michigan Transportation Research Institute. The data furnished were for 1970 through 1991. 53 The accident types are described as follows: Head-on: The collision of two vehicles traveling in opposite directions with the fronts of the two vehicles impacting. Head-on left: The collision of two vehicles when one of the vehicles is turning left and the front end of both vehicles impact. Angle-straight: The collision of two vehicles traveling at 90 degrees to each other prior to the time of impact. Rear-end: The collision of two vehicles traveling in the same direction with the front of the trailing vehicle impacting the rear of the leading vehicle. Rear-end right: The collision of two vehicles traveling in the same direction with the front of the trailing vehicle impacting the rear of the leading vehicle when the leading vehicle turns right. Rear-end left: The collision of two vehicles traveling in the same direction with the front of the trailing vehicle impacting the rear of the leading vehicle when the leading vehicle turns left. Angle-turn: The collision of two vehicles traveling in opposite directions when the turning vehicle turns into the path of the other vehicle. Sideswipe same direction: The collision of two vehicles traveling in the same direction with one of the vehicles impacting the side of the other vehicle with the side of the vehicle. Sideswipe opposite direction: The collision of two vehicles traveling in opposite directions with one of the vehicles impacting the side of the other vehicle with the side of the vehicle. 10. ll. 12. 13. 14. 54 Angle-drive: The collision of two vehicles when one vehicle is turning into a driveway or out of a driveway with the front end of one vehicle impacting the side or front of the other vehicle. Rear-end drive: The collision of two vehicles traveling in the same direction with the front of the trailing vehicle impacting the rear of the leading vehicle when the leading vehicle turns into a driveway. Dual left-turn: The collision of two vehicles traveling in the same direction and turning left along side of each other with the side of one vehicle impacting the side of the other vehicle. Dual right-turn: The collision of two vehicles traveling in the same direction and turning right along side of each other with the side of one vehicle impacting the side of the other vehicle. Other accidents: Other accidents not categorized above, which include accidents such as deer, pedestrian, and parking-related accidents. Accident severity is defined as follows: Type "A" injury: An incapacitating injury, any injury, other than fatal, which prevents the injured person from walking, driving, or normally continuing the activities which he/she was capable of performing prior to the traffic accident. Type "B" injury: A nonincapacitating injury, other than fatal and incapacitating, which is evident to any person at the scene of the accident. Type "C" injury: A possible injury reported or claimed which is not a fatal, incapacitating, or nonincapacitating evident injury. Type "K" injury: A fatal injury that results in death within 90 days of the accident. 55 3.5 SITES OF INVESTIGATION The project includes all five-lane, seven-lane, and boulevard highways in the State of Michigan, representing a variety of sites that include urban and rural settings. These locations are distributed around the state and consist of 366.3 miles of five-lane highways, 55.8 miles of seven-lane highways, and 512 miles of boulevard highways. A complete listing of these roadway sections is provided in Appendix Table A1 with the control section, common state route number, and limits of data acquisition indicated. A State of Michigan map showing MDOT Districts is shown in Figure A]. 3.6 EXECUTION OF THE EXPERIMENT The highway data for the experiment were collected on the basis of the features listed in Section 3.2.2. All the data were collected during the period from January 1994 to October 1994. The forms shown in Figures A4 to A9 and A13 were designed for ease of use by the data collectors. A set of instructions in Figures A2, A3, and A10-A12 was developed by the author for use by the individuals collecting the data. It was determined that multiple viewing passes of the photolog would be made, with similar data being collected during each pass. Information identifying the control section (location of the data being collected), the date that the video or film was recorded, the direction of travel, the variable being collected, and the mile point were common to all data sets collected. The forms also provided a location for recording the common given route number, the name of the individual recording the data, the name of the individual verifying the data, and the sheet number. 56 The data were collected in MDOT offices located in Lansing, Jackson, and Kalamazoo. The equipment consisted of the following units: 1. Sony optical disk player (Figures 3.6.1 to 3.6.4) 2. Bell and Howell 35 mm film viewing machine 3. Sharp Slim Cam video camera 4. Canon AB] 35 mm camera 5. Distance measuring device Field measurements were performed using a measuring wheel and a video camera to field-check data acquired from the photolog video player and the 35 mm projection screen and to obtain certain other ground data. Separate instructions were developed for the field verification and collection of additional data. Space was provided for any unusual circumstances or observations that may have been noted during the data acquisition and field verification phase. The forms and their corresponding data are as follows: 1. Form A: number of lanes 2. Form B: intersection mile points and names 3. Form C: cross-over types and mile points 4. Form D: drive mile points and characteristics 5. Form B: unusual features such as lunes or bus turn-outs 6. Form P: type of roadway 7. Form G: median width, lane width, and speed limit 57 Figure 3.6.2 Loading Optical Disk Into Photolog Machine Figure 3.6.3 Operating Photolog Equipment "‘ I.I..1.a.,u.1un I’m-min" f , . /~ . Fill! Map Uttfllll'l Window»; 1 :Whigcn lrunuiur ‘ a“ N1 .1 (fr: Ditutian’ O Mainline : ‘EtVf‘EEQ M. 1 I ( «I Figure 3.6.4 Photolog Image of Road Section 59 8. Form K: parking restrictions; not used in the study 9. Form L: turn restrictions; not used in the study 3.7 DATA INSPECTION A visual inspection was made of the data set utilizing scatter diagrams and histograms. This was performed so that any obvious coding errors, correlations, or trends could be recognized. 3.8 UNIT OF ANALYSIS The unit of analysis is a section of highway with a unique control section with beginning and ending mile points (the basic MDOT referencing system). Originally the segments of highway were divided into unique element for every change in the number of lanes, median width, lane width, curb and gutter/shoulder,and/or speed limit. However, the control sections did not exhibit a change in element as often as anticipated which resulted in a number of segments of considerable length. Thus, a uniform segment length of one quarter mile was chosen. Segments less than one quarter mile in length were not used in the statistical analysis. 3.9 STATISTICAL ANALYSIS Discriminate analysis was utilized to determine which independent variables explained the difference in accident rates for boulevard and CCLTL roadways. A branching process was the applied utilizing discriminate analysis. The discriminate analysis was 60 continued until a sample size of 30 values or larger was no longer available or no significant difference was indicated at the 0.95 significance level. The statistics computed include the mean, standard error, maximum, and minimum for seven accident types. These statistics were obtained for all segments, unsignalized segments, and signalized segments. The Student’s t-test was then utilized to determine if the difference between the means for boulevard vs. CCLTL data sets were significant. The one-way analyses of variance (AN OVA) utilizing Tukey’s B was performed for the following data sets: 0 three categories of boulevard widths. - three categories of boulevard posted speeds. 1' three categories of boulevard ADT. - three boulevard median widths for two cross-over types. - three categories of CCLTL posted speeds. 0 three categories of CCLTL ADT. 3.10 CHAPTER SUMMARY The independent variables determined to be relevant to the validity of the hypothesis are: traffic volumes, percent commercial vehicles, county population, speed limits, intersection density, signal density, driveway type and density, lane width, shoulder and curb, median width, cross-over type and density. The dependent variables were selected to be: accident rate expressed as accidents per hundred million vehicle miles of travel and accident severity. 61 The data for the project were collected through the use of photolog techniques and field measurements. These were merged with available records and were then converted into computer data files for input in computer-supported analyses. The analytical unit of analysis was a quarter-mile section of highway with uniform roadway features. CHAPTER4 4.1 INTRODUCTION The distribution of the data is presented in Section 4.2. The frequency distributions present an indication of which independent variables have a relationship with the dependent variables described in Chapter 3. The results of discriminate analyses utilized to determine which independent variables explained the difference in accident rates for boulevard and CCLTL roadways are presented in Section 4.3. In section 4.4 accident statistics for seven accident categories are developed for the independent variables that appeared to be important. The Student’s t-test and AN OVA with Tukey-b test were utilized to determine whether there was a significant difference between the means for the groupings that were being compared. The conclusions drawn from the data analysis are discussed in Chapter 5. 4.2 DATA INSPECTION The fiequency distributions of the data for the independent and dependent variables were investigated. These frequency distributions provided an indication of distribution types 62 63 (normal, uniform, or Poisson) skewed data, coding errors, and outliers. These frequency distributions are presented in Figures 4.2.1 to 4.2.15. The occurrence of different types of accidents and their distribution for boulevard and CCLTL highways are of interest. Figure 4.2.1 illustrates the accident data distribution as a function of the accident types for boulevard and CCLTL highway segments. Thirteen accident types are shown with the remaining accident types (such as, animal-related accidents) being grouped into the "other" category. Boulevard highway segments were found to have a higher percentage of rear-end and angle-straight accidents than the CCLTL highways. Rear-end and angle-straight accidents are associated with intersection conflicts. CCLTLs had higher percentages of other, angle- drive, rear-end drive, and head-on, left-turn accident types. It is not surprising that divided (boulevard) highways have a relatively lower percentage of head-on accidents as the directions are separated by a median. This reduces the number of potential head-on accidents. Other than vaulting the median, this type of accident would only occur if a wrong-way movement takes place. It is not intuitive that a boulevard should have a relatively lower percentage of driveway-related accidents. However, as will be seen later on Figure 4.2.9, the driveway density is generally greater for highway segments with CCLTLs. Rear-end accidents represent the largest component of accidents--accounting for 34,021 or 56 percent of the boulevard accidents and 29,582 or 47 percent of the CCLTL accidents. The second most frequent accident type was the angle-straight accident accounting for 8,784 or 14 percent of the boulevard accidents and 5,442 or 9 percent of the El Boulevard I Center Left umr 1116121 lend U-"U. U31 1300 11 uo peaH d ddo adIMS apws ,_ eAua pua Jeaa eAuo a|6uv 1 Jaluo ‘ .LH pue may n pue Jeaa ‘ ewes adIMs ems urn; a|6uv pus-lees 111618119 alfiuv uo-peaH 60 0.4:. O O O V M N sauw to efimuearad 50 10 Accident Percentage for each Type of Accident, All Segment Lengths Figure 4.2.1 65 CCLTL accidents. Head-on left- and angle-tum were the next most frequent accident types. Severity is a measure of the seriousness of the injuries sustained by individuals involved in an accident. In Figure 4.2.2 the distribution of accident data is illustrated as a function of accident severity for boulevard and CCLTL highway segments for four levels of accident severity. The classification of the level of accident severity is as follows: Type A represents accidents where disabling injuries occurred; Type B represents accidents where nondisabling injuries occurred; Type C represents accidents where no visible injuries occurred but complaints of injury were voiced; and Fatalities represents those accidents where deaths occurred. The bar chart indicates that there is no apparent difference between the boulevard and the CCLTL highway segments for the four severity categories. It might have been expected that a higher percentage of severe injury accidents and fatalities would be associated with the CCLTL segments as the direction of travel is not physically separated. The Type C accident was the most common accident type for both roadway types with 9,581 or 64 percent of the boulevard injury accidents and 13,112 or 63 percent of the CCLTL injury accidents occurring in this category. The interesting observation is that the percentage of each injury type is almost identical for each roadway type. Traffic volumes vary for individual highways, depending on the location and the demand for highway use. Figure 4.2.3 illustrates the distribution of total miles as a function of traffic volumes for boulevard and CCLTL highway segments. Twenty-six categories of ADT volumes, ranging from 4,000 vehicles per day to over 104,000 vehicles per day in increments of 4,000, are provided in the bar chart. 66 35293 =< .2 38:33.; 3:33 EmEoo< " «.mé 2:9“. mega—an. Ell. ca.— ..oacoo I 232:3 B 3:863 co ergo...“ m 8C 0 on»... < 2;... or am on ov om cm on stuepgaov to accrueorad 20 18 to P V ,— 67 El Boulevard I Center Left N O Q F P seam to abetuemed OOOVOI -000001 00096 00026 00088 000V8 00008 -~ 0009L 0002L 00089 00079 00009 00099 00029 00087 000?? 0000? 00098 00028 00082 000V2 00002 - 0009i 0002i 0008 000? Traffic Volume (ADT) Traffic Volume Percentage for All Section Lengths Figure 4.2.3 68 Boulevard highway segments had a higher percentage in the 4,000 vehicles per day to 8,000 vehicles per day range and from 56,000 vehicles per day and higher. CCLTL highway segments with 12,000 to 52,000 vehicles per day had higher percentage than boulevard highway segments with similar levels of traffic with few of the volume category exceptions. It would appear that the lower voltune roadways are associated with rural situations where boulevards are more prevalent and that CCLTLs are associated with urban situations. The exceptions are the very high volumes which can be accounted for by boulevard roadways such as 8 Mile Road in the Detroit Metropolitan area. This is illustrated in Figure 4.2.6 which indicates a higher percentage of boulevards in the 55 mph category. The CCLTL percentage is higher for the lower speeds except for the 40 mph category. The traffic volume for the boulevard cross-section with the largest percentage of miles was an ADT of 36,000 vehicles per day, accounting for 31 miles or 12 percent of the total miles of that roadway type. The traffic volume for the CCLTL cross-section that accounted for the largest percentage of miles was an ADT of 16,000 vehicles per day, accounting for 25 miles or 15 percent of the total mileage for that roadway type. The highest actual volume for the boulevard cross-section was 104,000 vehicles per day. The highest volume for the CCLTL cross-section was 76,000 vehicles per day. In general it appears that the boulevard cross-section has more roadway mileage with the higher volumes. The portion of traffic on a roadway that can be attributed to commercial traffic may be a factor in the accident occurrences on that highway. The distribution of the surveyed roadway mileage as a function of percent commercial for boulevard and CCLTL segments is illustrated in Figure 4.2.4. 69 36:3 :ozoom =< .352 .o «cached m> 33.55500 «5ch " v.~.v 959... 3.9.3500 2.3.8.... w h o m v m N _. :3 .2on I 955.25 B .. sauw Io afierueored 70 The commercial percentages of volume are provided for 11 categories. Except for the 1 and 4 percent categories, the CCLTL segments dominated the roadway with lower percentages of commercial traffic. This is the opposite of what may have been expected as the CCLTLs are found in the lower speed, urban areas which would be expected to have a higher commercial vehicle presence. The higher commercial vehicle traffic categories are dominated by the boulevard segments. This also was unexpected but could be accounted for by the boulevard roadways in southeastern Michigan. The highest commercial percentage for'the boulevard cross-section, 2 percent, accounted for 48 miles or 20 percent of the roadway mileage. The highest commercial percentage for the center left-turn cross-section, 2 percent accounted for 46 miles or 28 percent of the total mileage. The mileage for both the boulevard and CCLTL roadways appeared to be concentrated in areas with low percentages of commercial vehicle traffic. It is believed that population could be a surrogate measure of development. In Figure 4.2.5 the distribution of roadway miles as a function of the county population is illustrated for boulevard and CCLTL highway segments. Values in the bar chart are provided for 22 population categories ranging from 100,000 to 2,200,000, in increments of 100,000. Boulevard segments had a higher percentage of miles in less populated counties which would be expected as the boulevard segments were more prevalent in the higher speed areas (Figure 4.2.6). The CCLTL segments were more prevalent in the middle population areas. The boulevard segments were more prevalent in the higher population area. 71 E3 Boulevard I Center Left 30 25 N ‘- seuw to efimuemed ,' I ‘I I I O m o m o r 0000022 0000012 0000002 0000061 0000081 00000L1 0000091 0000091 00000?1 0000081 0000021 .3, 0000011 0000001 000006 000008 00000L 000009 000009 00000? 000008 000002 000001 Population Lengths IOI’I Population Percentage for All Sect' Figure 4.2.5 72 It is quite evident that the boulevard sections are located in both the least populated and most heavily populated areas while the CCLTLs are located mainly in the sparsely to moderately populated counties. Twenty-nine percent of the boulevard cross-section mileage (69 miles) is located in Wayne County. Twenty percent of the C CLTL cross-section mileage or 34 miles is located in Oakland County. The posted speed limit can be an indication of the amount of development, and the rural or urban nature of a roadway. Figure 4.2.6 (Speed Limit Percentage for all Sections) illustrates the distribution of segment mileage as a function of posted speed limits for boulevard and CCLTL highways. The seven posted speed categories (from 25 to 55 mph in 5 mph increments) are represented in the bar chart. The boulevard mileage appears to be linearly distributed. The greater percentage of CCLTL mileage is concentrated in the 35 to 50 mph category. The greatest mileage of boulevard cross-section, 123 miles or 51 percent, has a 55 mph speed limit. The greatest mileage of CCLTL cross-section, 54 miles or 32 percent, have a 45 mph speed limit. The boulevard cross-sections are concentrated in the 55 mph speed limit areas and the CCLTL cross-sections are concentrated in the 35 and 45 mph speed limit areas. It was found in the literature that intersection density can have an effect on the accident rate. Figure 4.2.7 illustrates the distribution of segment mileage as a function of intersection density for boulevard and CCLTL highway segments. Nine categories of intersection density from five intersections per mile to greater than 40 intersections per mile in increments of five intersections per mile are presented in a bar 73 to.— ..3:00 I Em>£=om n. mm 2.253 =< .8 $5528; 33m n odd 2:9". 9. 39.5 team ov mm on mm or ON on ov om ow sauw Io afieruaarad 2.55.. 5.63 ..< .3550 5.33.8... .o 38:33.". 2.5. u find. 2:9". 3...: .2. 20.83.25. 3.230 5.83.3... OVA ov mm on ma cm 2 or m l a IIHHH II, or 1 11 cm la on 11 ow 11 on ow seam to afietuearad 75 chart. The large number of intersections in some cases is due to the presence of short segments in the sample. These short segments were removed during the analysis phase. The largest portion of the boulevard segment-miles is concentrated in the five- to ten-intersection per mile categories. This is inconsistent with the population distribution which indicates that the boulevard segments are concentrated in more rural situations where a lower intersection density would exist and high population density areas where higher intersection density would be expected. The largest portion of CCLTL segments appears to be concentrated in the five- to ten-intersection per mile categories, indicating that these segments are found in a more urban setting than the boulevard segments. The highest category of intersection density for a boulevard cross-section was 58 percent of the total or 142 miles, with an intersection spacing of five intersections per mile. The highest percentage of intersection density for CCLTL cross-section was 46 percent of the total or 79 miles for an intersection spacing of ten intersections per mile. The intersection spacing appears to be concentrated in an intersection density up to 15 intersections per mile. It is possible that accidents are related to signal density. Figure 4.2.8 illustrates the distribution of signal density for boulevard and CCLTL highways. The signal density has been divided into categories ranging from one signal per mile to greater than ten signals per mile, in increments of one signal per mile. The largest percentage of boulevard mileage is concentrated in the category of one signal per mile which, with the intersection density, correlates with the concentration of boulevard mileage in rural areas. The majority of the boulevard mileage is concentrated in the two through four signal per mile category. The balance of the boulevard mileage is distributed in the five to greater than ten signal per mile 76 «598.. 5.83 ..< 5.9.8 .255 .o 3852?. 2.5. u Qué 9.sz 3A 2. IllTlui .25 .2. «.25.». 3.2.3 .22» m h o 11% cm.— ..mEoO I 296.38 0 .3 tom tom f9. tom om sauw go afieiuazuad 77 category. The CCLTL mileage is concentrated in the one through four signal per mile category with no uniform distribution. The balance of the mileage is distributed in the five to greater than ten signal per mile categories. It appears that there is a higher concentration of the CCLTL mileage in the more urban areas, consistent with the higher signal concentrations and the distribution of mileage by speed limit. The highest category of stop-go signalized intersection density for boulevard highways was 54 percent or 131 miles with a signal spacing of one signal per mile. The highest category of stop-go signal density on CCLTL highways was 37 percent or 64 miles for a signal spacing of two signals per mile. The stop-go signal spacing appears to be concentrated at a signal density of one signal per mile for boulevard cross-sections and two signals per mile for CCLTL cross-sections. The amount of roadside development and the number of access points were shown to be related to the accident frequency in the literature search. Figure 4.2.9 illustrates the distribution of segment mileage as a function of driveway density for boulevard and CCLTL highways. Nine categories of driveway density ranging from five driveways per mile to greater than one hundred driveways per mile, in increments of five driveways per mile, are provided in the bar chart. The boulevard segments appear to be concentrated in the five driveways per mile to thirty driveways per mile category with the greatest concentration in the five driveways per mile category. The CCLTL mileage is concentrated in the 30 to 60 driveways per mile category. These coincide, once again, with the indication from the mileage/speed limit distributionwhich indicated the majority of boulevard mileage is in rural areas. This is indicated by a lower driveway density. The majority of CCLTL mileage is in more urbanized areas which is indicated by a higher driveway density. 78 co.— ..oEoO I .5538 .u 0.59.3 EoEmom ..< 5.9.3 >u>>o>to M.o 35:83.. 2.5. u add 2:9“. co—k mm mm ms 3...; .2. 232625. 3.2.00 >a3a>to mm mm mv mm mm m_. O F In P O N mm on sauw go afimuamad 79 The boulevard cross-section driveway concentration with the highest mileage percentage is five driveways per mile for 66 miles or 27 percent of the total mileage. The CCLTL driveway spacing has the highest concentration at 50 driveways per mile for 12 percent of the total mileage or 22 miles. It appears in general that the higher driveway concentration for the boulevard cross-sectionis in the 5 to 25 driveways per mile range and the CCLTL concentration is in the 30 to 80 driveways per mile range. It was indicated in the literature search that the driveway access type could affect the accident rate. Figure 4.2.10 illustrates the distribution of segment mileage as a function of commercial driveway density for boulevard and CCLTL highway segments. The commercial driveway density is divided into 16 categories ranging from five driveways per mile to greater than seventy-five driveways per mile in increments of five driveways per mile. The largest portion of the boulevard mileage is concentrated in the five driveway per mile category. This would coincide with the boulevard speed distribution, which indicates that the boulevards are concentrated in the more rural areas. It could also be an indication that the number of commercial driveway access points on boulevard roadways are limited, with collector roads being utilized to supplement the access.The balance of the boulevard mileage is distributed fairly evenly between 10 commercial driveways per mile and 45 commercial driveways per mile, and between 50 commercial driveways per mile and over 75 commercial driveways per mile. The distribution of CCLTL mileage is concentrated between 5 commercial driveways per mile and 45 commercial driveways per mile, with the larger percentage between 35 and 45 commercial driveways per mile. The boulevard cross-section, commercial driveway density with the highest percentage of mileage has five commercial driveways per mile, for 133 miles or 55 percent 80 9.65.. 5.83 ..< 5.23 >u3¢>to 3.955560 .0 693523. 6...... u 3&6 6.59.“. 3...: .2. «.33th 3.205800. 3.250 $33.5 3.805500 2% 2. 2. mm on mm on me ow mm cm mm cm m. o. m r-w1111.1 .1. c3538- ! 8 Em>o_=om n. [T ov om sauw JO afieruamad 81 of the total mileage. The CCLTL section commercial driveway density with the highest percentage of mileage has 45 driveways per mile, for 23 miles or 13 percent of the total mileage. It appears that other than the concentration of the low density of commercial drives on boulevard cross-section roadways, there is no trend. While information was not cited in the literature, it was decided to investigate the distribution of lane widths. Figure 4.2.11 illustrates the distribution of segment mileage as a function of lane width for boulevard and CCLTL highways. Four lane width categories ranging from nine feet to twelve feet in one foot increments, are provided in the bar chart. The largest percentage of mileage for both boulevard and CCLTL mileage is concentrated in the 12-foot lane category with CCLTL segments having a higher percentage of the mileage in this category. This coincides with the AASHTO guidelines which recommends that l2-foot lanes are desirable. The larger percentage of CCLTL mileage in this category could be attributed to the historically more recent nature of this construction type. This is also reflected in the next largest category which indicates that boulevards have a larger percentage of ll-foot lanes than the CCLTLs. The narrower widths could also be an indication of restricted right-of-way conditions. Twelve-foot lanes are prevalent for both the boulevard and center left-turn lane cross- sections. The boulevard cross-section has 12-foot lanes on 153 miles or 63 percent of the total mileage. The CCLTL cross-section has 12-foot lanes on 134 or 77 percent of the total mileage. There was an indication in investigations by other researchers that curbs could have an effect on accident rates. Figure 4.2. 12 illustrates the distribution of segment mileage as a function of curbed and noncurbed roadway sections for boulevard and CCLTL highways. 82 msumcod :ozuww =< .50.; man.— u0 amflcuewm 2.5. . Sand 2:9". AE 52>) 2.3 NF _. r or m to.— ..250 I Ew>o_=om n. or am on 0? on on on on sauw 10 asexuazuad 83 2.9.3 8.88 __< .83 023.5 .o 83:83“. 2.... n «E... 2:9“. £30 02 or 0N on ov om om on ow om sauw go afieiuazuad 84 The curbed and noncurbed roadways are divided into these categories in Figure 4.2.12. The predominance of boulevard mileage is in the noncurbed roadway category as might be expected since the majority of the mileage is in rural areas. AASHTO guidelines recommend that no curb be placed adjacent to roadways where the posted speed limit is above 45 mph. The majority of the CCLTL mileage is in the embed category, which is another indication of the primary location of these roadway types in an urban environment. The majority of the mileage of the boulevard cross-section, 147 miles or 61 percent, is not curbed. The majority of the mileage of the CCLTL cross-section, 139 miles or 85 percent, is curbed. The literature indicated that there could be a relationship between the median width of a boulevard highway and the accident rate. Figure 4.2.13 illustrates the distribution of boulevard mileage as a function of median width for boulevard highway segments. There are values in the bar chart provided for three categories of median width: 0 to 30 feet, 30 to 60 feet, and greater than 60 feet. The highest percentage of mileage is in the 30 to 60 feet medians. The category of 30 to 60 feet, as found in the literature, is desirable for passenger vehicle operation. The 60-foot and greater category is desirable for truck operations. The 30- to 60-foot median is the most common type of median with 135 miles or 56 percent of the roadway mileage in this category. The next most common median width is the greater than 60 feet with 98 miles or 40 percent. The least prevalent median width is less than 30 feet with only 4 percent in that category. MDOT has been replacing bidirectional cross-overs with directional cross-overs. Figure 4.2.14 illustrates the distribution of cross-overs as a function of cross-over type for boulevard highway segments. '!#1.J!.OOIIQ.OOOOOIUI‘ .e- u no. 1.. .0. on. n..- u o u n .... n no no a o - .. e o a c... n a. IOIOCOIQOOOunion-olooacloono-o - o u - 60 50 O O O V ('0 N seam to “among 10 0 Median > 60 ft Median 30-60 ft Median < 30 ft Median Width Width, All BLVD Sections Lengths Mile Percentage of Med' Figure 4.2.13 86 25:3 5580 0>..m ..< in: 33320 no 0085950 .56... n 3.3. 950.”. 5339.0 .0 on»... 35.82.05 3.8.82.0 or am on 0? on on sJaAossma mo; )0 efiewazuad 87 The two cross-over types, directional and bidirectional, are graphically represented in Figure 4.2.14. The two types of cross-overs are fairly evenly distributed with approximately half of the cross-overs in each category. Directional type cross-overs are located in 1,119 locations or 49 percent. Bidirectional cross-overs are located in 1,173 locations or 51 percent. In order to determine the mix of section lengths, a chart indicating section distribution by length was prepared. Figure 4.2.15 illustrates the distribution of section miles as a function of section length for boulevard and CCLTL highways. The highway segments are divided into 13 categories of section length in increments ranging from one to thirteen miles, in increments of one mile. The majority of both boulevard and CCLTL sections are one to four mile lengths, with the balance of the miles distributed between four-mile and nine-mile length categories. A larger percentage of the CCLTL sections than the boulevard sections are in the shorter categories of one and two mile lengths. The most common section length for boulevards is three miles, which represents 54 miles or 22 percent of the total boulevard mileage. The most common center left-turn lane cross-section length is 2 miles, which represents 43 miles or 25 percent of the total mileage for that roadway type. The longest section is a boulevard cross-section roadway which is 13 miles long. The longest center left-turn lane cross-section roadway is eight miles long. The disproportionate length of these roadways compared to the preponderance of roadways with short sections has caused problems with the confidence intervals computed for accident rate vs. roadway type. It was therefore decided to reduce the segments longer than 0.25 mile to multiple segments of 0.25 mile for the remainder of the investigation. 88 to; .2000 I 2132.50 fl $305.. .3303 ._< .598. 3:30 .0 ems—32o“. 2.5. ” Emmi 950.“. .85... 59.3 5.88 2 N. : o. a a h o m w m N seuw }O afiewezued 89 SUMMARY OF DATA INSPECTION An inspection of the data indicates that it is not normally distributed for any of the data sets reviewed. Some of the data sets appear to have a Poisson distribution. The rear-end, angle-straight, angle-turn, head-on-lefi turn, and driveway related accidents are the most prevalent types of accidents detected. Accident severity does not appear to be a factor when comparing boulevard with CCLTL roadways. The traffic volume, intersection density, and population for boulevard and CCLTL roadways have similar distributions. The boulevards are prevalent in lower and higher values for these three categories and the CCLTL roadway are more prevalent in the middle ranges. The distributions of the speed limits followed the same pattern except boulevard and CCLTL are similarly distributed for low speed limits. Signal and driveway density distributions exhibit the same patterns except boulevards and CCLTLs are similarly distributed for high signal densities. Commercial driveway density is distributed similarly for boulevards and CCLTLs for the lower and middle ranges but the CCLTLs are more prevalent in the higher driveway density ranges. There is a higher percentage of CCLTL roadways with curbed segments. These distributions are indicative of where these roadway types are located. The boulevard roadways are traditionally utilized in high density and rural areas. The CCLTLs are found in the medium to higher density areas. The balance of the distribution comparisons do not indicate any similarity between boulevards and CCLTL roadways. 4.3 DISCRIMINATE ANALYSIS Discriminate analysis is a procedure that is utilized to identify the variables that are most important for distinguishing among groups and can be used to develop a procedure to predict group membership for new cases. This analysis was used to determine which independent variables may explain the variance in accident rates for boulevard and CCLTL roadways for all accidents, rear-end accidents, and drive-related accidents. Certain criteria should be met when using this procedure; i.e., each group should be from a multivariant normal population, and the population covariance matrixes should all be equal. While all of the conditions are not necessarily met for this procedure, it was determined to be the most appropriate technique to identify the variables worthy of further analysis to determine their role in explaining the variance in the accident rates. The discriminate analysis was progressively applied through a branching procedure. The data set was progressively divided into two data sets which produced the most significant difference in accident rates. The dichotomous split of the remaining data set continued until a sample size of 30 value or greater was no longer available or until no significant difference (with an a = 0.05) was found between the two cells. The Student’s t- test was utilized to determine the split between the values of the dependent variable that provided the most significant difference between the mean accident rates. The variables selected to help explain the variance in the accident rates selected were: 1. Average Daily Traffic 2. Commercial Drive Density 3. Curb/ Shoulder 4. Driveway Density 91 5. Intersection Density 6. Lane Width 7. Number of Lanes 8. Percent Commercial 9. Population 10. Signal Density 11. Posted Speed 12. Median Width (Boulevard Roadway only) 4.3.1 BOULEVARD ROADWAY ALL ACCIDENTS The discriminate analysis procedure was applied through a branching procedure to the overall boulevard roadway accident rate. The results of the discriminate analysis are displayed graphically in Figure 4.3 . 1. Signal density was identified as the dependent variable that explained the greatest difference in overall accident rates. The accident rates for the segment groupings were tested utilizing the t-test to determine which signal category would produce the most significant difference in accident rates between the groupings. This resulted in subdividing the segments into those that had no signals and those that did have signals. The mean accident rate for those segments without signals is 221 accidents per million vehicle miles (A/MVM) compared to an accident rate of 691 A/MVM for signalized segments. These rates are significantly different at the 0.95 significance level and resulted in a variance reduction of 31 percent. The segments without signals were analyzed for the variable that would explain the most significant difference in accident rate in these segments. It was found that intersection 92 $5203.. .3333. pagfiom =< .8 3.33. mafia: 3.3.8.085 fin... Paw...— omd n «M 83%.? mm 8459.2. mm: Emig— _ ”magma... . Tumgdsl _ “cacao: _ _ . _ _ 80.838» «3.5.78 9808. 7mm 08.3.6.3 9m. .20... 8a. can... an. 37...: EN... 3...: 23% :28 £sz 58x 28:: 228 28520.. mum 3.3.... as... 3:sz was 3:sz was 2:5sz 20:89.sz ouEmzmo 20:89.5. #5 5:33 is...» . m f . 4 _ . W «8.22"...» .8358» Egan”... 48.3.83 81. 21k in. sank. .8»: ..:fi 2.... sank 20:538.. £sz max: .6sz 20:89.5. 2.... E 228 2288228 1.288 :28 9A...» 859. $5 on.» oEwOasS . _ . F . . 8.5.2....” ”8.20.0.3 2i. «8...... 81. Emu 235 228 8.... 8:8: .....__o @502 1»:sz zonommmwz. 1.»:sz 20:89.32. _ _ . _ flags-mm $828.8 5?. fink m8»: 3%.... 205.58.. Esz 20.5ng2. 3,:sz €20.» 3:.sz 229m A m _ §.§.~9.$ 8?: «SR .628 3.6.... 93 density best explains the difference between those segments with O to 4 intersections per mile, having an accident rate of 174 W M, and those segments with an intersection density greater than 4 intersections per mile, having an accident rate of 336 A/MV M. This grouping explained an additional 9 percent of the sample variance. The segments with signal density greater than zero were analyzed and population was identified as the variable which explains the most significant difference in accident rates. Further population analysis did not indicate a significant difference in the sample groupings for the various populations. Thus, the cells were not split further. The discriminate analysis procedure was continued until the sample size was smaller than 30 values or there was no significant difference. This branching procedure resulted in an R2 = 0.36. The signal density, with an R2 of 0.3 1 , explains 86 percent of the difference with all of the other variables shown in the chart explaining 14 percent of the remaining explainable difference. It would therefore appear that significant differences in boulevard accident rates are primarily signal related. 4.3.2 BOULEVARD ROADWAY REAR-END ACCIDENTS The discriminate analysis procedure was also applied through a branching procedure to the boulevard roadway rear-end accident rate. The results of the discriminate analysis are displayed in Figure 4.3.2. The variable that explained the greatest difference in overall accident rate was signal density. The accident rates for the segment groupings were tested utilizing the t-test to determine which signal category splits would produce the most significantly different accident rates between groupings. This resulted in subdividing the segments into those that had no signals and those that did have signals. The mean accident BaoEoo< .3333. uB>oBom EMEBM .8 3.38m mare: 8.208085 NM... 953% 94 ovd u «M mmm.mmm.¢~ mm 83.8.2 mm: mum m4n2._._w2m0 r_ 8.... 200 20 2.200 2.. .252 588$ 8.» one: =25 .20 0.0 02 80.8 A :2 80835508.. 8.2.8 A 5... 80.85.9588 w — _ _ — — _ L 5.8.0.00 90.08.?00 08.00.00 60.05.00 .09... Sew 850.”. 00.... of.“ 8A.. NA.“ 5.. RE 0 A 200 .000 E. o u 200 500 E. o A 200 .2. 0 A 200 E. _ _ F _ ~ _ «8.0.00 8.8.1.00 00»: Sum. can: 0km 200 500 .2. 200 .800 02. .20 0.0 oz 00 A 880 088.. 00.880 0882qu _ . _ _ 0.2.30.8“. 00 08.80.00 802 8!.xl .9»: 04mm: 200 .2. 880 088A. 0 A 2288 .290 o A 3.88 .290 — __ 80. Eton-00 42A,. 89x 2288 .ocum 102 that did have signals. The mean accident rate for those segments without signals is 149 A/MV M compared to a mean accident rate of 490 A/MVM for signalized segments. These rates are significantly different at the 0.95 significance level and resulted in a variance reduction of 29 percent. The segments without signals were analyzed for the variable that would explain the most significant difference in accident rates in these segments. It was found that posted speed limits best explains the difference with those segments with posted speeds fiom 25 to 50 mph having a mean accident rate of 176 NW M and those segments with posted speeds greater than 50 mph having a mean accident rate of 37 A/MVM. This grouping provided the most significant difference and explained an additional 8 percent of the sample variance. The segments with signal density greater than zero were analyzed and intersection density was identified as the variable which explains the most significant differences in accident rates. Further analysis of intersection density did not indicate that there is a significant difference in the sample groupings for the various intersection densities. Thus, the cells were not split further. The discriminate analysis procedure was continued until the sample size was too small or there was no significant difference. This resulted in an R2 = 0.33. The signal density, with an R2 of 0.29, explains 88 percent of the difference with all of the other variables shown in the chart explaining 12 percent of the remaining explainable difference. It would, therefore, appear that the most significant differences in CCLTL rear-end accident rates are primarily signal-related. 103 4.3.6 CCLTL ROADWAY DRIVE-RELATED ACCIDENTS The discriminate analysis procedure was applied through a branching procedure to the CCLTL roadway drive-related accident rate. The results of the discriminate analysis are displayed graphically in Figure 4.3.6. The most important variable identified that explained the differences in drive-related accident rates was the presence or absence of curbs. The accident rates for those segments with curbs have a mean of 151 A/MV M compared to a mean accident rate of 116 NW M for uncurbed segments. These rates are significantly different at the 0.95 significance level and resulted in a variance reduction of only 0.6 percent. The segments with curbs were analyzed for the variable that would best explain the differences in accident rates in these segments. It was found that signal density best explains the difference with those segments with signal densities equal to O signals per mile having a mean accident rate of 118 A/MV M and those segments with signal density greater than 0 signals per mile having a mean accident rate of 188 A/MV M. This grouping providing the most significant difference in accident rates and explained an additional 4 percent of the sample variance. The segments without curbs were analyzed for the variable that would explain the most significant difference in accident rates in these segments. It was found that posted speed best explains the difference with those segments with posted speed from 35 to 50 mph having a mean accident rate of 191 A/MV M and those segments with a posted speed greater than 50 mph having a mean accident rate of 25 NW M. This grouping provided the greatest diffemeces and explained an additional 2 percent of the sample variance. The discriminate analysis procedure was continued until the sample size was smaller than 30 values or there was no significant difference. This resulted in an R2 = 0.13. The 0.000.000. 202.08% 0.5.7.004 00.000 000250000 00.200. 03.5 00.. 0.300% 0.03.090. 0.0025805 0 m w 0.5»:— 104 n _ .0 u «m «20.0 _ 0.2 mm 82.8.28. H I i i .3. 30.00 8280.00 .882 .00 ”830:... a... 8.... .0... 8..lx 81.2.02: 200>00l 28 000! 020 000001 2.0.. 0.. 20.. 0.6.20 0.... 020 Flall. .IJL 80.500 328.000 .2... 8%.. 2.1.8...“ 28 00.. 200 00 .228 00.0 0820.930 828%.. 8285.08.82. 00.0 0.20 0.220 F _ _ _ r 22.8.00 8280.00 . . 820.700 .0... .2... .1. 8w... “WWW...” .1. sum. 200 00 .2218 28 00A. .9. I 200 80000.2. 0000820320 .68 3000 00.5.2.2. 2.0.2.932. 00.0 0.220 .320 — _ % h _ u _ _ 3.08.800 80.80.00 0.208.100 30080.00 8.108.“ 8.1;...V .8... 87V 8.... caumu 200 2.20.0 200200228 2.0.2. 025 200 0200 00.000.303.20 0....0 0.002 0.2.0.3025 2.0.2.932. 8350.. §~..v.0<580~ — — u H L — n — 8.8.00 32.02100 5.80.30 .882900 0.... Nil .0... .0...xl 2."... 87M. 232.... 2008.228 2008.228 25.2.02... .9 I 860000 0030000030. 0A 200 0.0 0.200 0.0 F _ _ — _ — 0808.800 5.08.0700 3.... 0...x| 0.0.... EAR 0000 00.000 200 2.20.0 0.5002 050 _ _ _ 08.80.0800 3.... a...“ 050 105 curb/no curb with an R2 of 0.01 explains 5 percent of the difference with all of the other variables shown in the chart explaining 95 percent of the remaining explainable difference. It appears that there is a weak relationship between the dependent variables and the CCLTL drive-related mean accident rate. 4.3.7 DISCRIMINATE ANALYSIS SUMMARY The discriminate analysis process was utilized to investigate the factors that account for the differences in accident rates for all, rear-end and drive-related accidents for both boulevard and CCLT roadways. The main discriminating factor for all and rear-end accidents is the presence of signals for both boulevard and CCLTL roadways. Signalization explains from a low of 84 percent of the explainable difference for all types of accidents for CCLTL to a high of 92 percent of the explainable difference for rear-end accidents of boulevard roadways. 4.4 ACCIDENT ANALYSIS The accident statistics were examined for the boulevard and center left-turn cross- sections for various conditions and the means between similar groups were compared utilizing the Student’s t-test with an a = 0.05. The results of the statistical analyses are summarized in Tables 4.4.1.1 to 4.4.4.1. These results are also graphically illustrated through the use of bar charts in Figures C.1 through C.27 which are located in Appendix C. The statistical analysis for significance of difference between the means is summarized in Tables D.1 through D5 which are located in Appendix D. The results are summarized as follows: 106 4.4.1 BOULEVARDS AND CONTINUOUS CENTER LEFT-TURN LANES ACCIDENT ANALYSIS (Table 4.4.1, Appendix C, Figures C.1-C.3, Appendix D, Table D.1 - 13.5) 4.4.1.1 TESTING OF FIRST HYPOTHESIS The first hypotheses was stated as follows: “It may, therefore, be postulated that the accident rate for a roadway with a boulevard is different than an accident rate for a roadway with a CCLTL.” To test this hypotheses, the accident rates for CCLTL and boulevard highways were compared. 4.4.1.2 SELECTION OF RANGE OF INVESTIGATION Table 4.4.1.1 (Boulevard and CCLTL Accident Rates Per 100 Million Vehicle Miles) tabulates the accident statistics for seven accident categories for boulevard and CCLTL highway segments. These statistics are tabulated for all segments, segments without stop- and-go signals, and segments with stop-and-go signals. Appendix C illustrates graphically the mean accident rates for the seven categories of accident types. The statistical significance of the difference between the mean accident rates are summarized in Tables 1 and 2 in Appendix D. 07 1 wmhg ._.zmn=00< sz 25:. Em.— mmhzmo Dz< nm<>m.=.-Om .366 m4m..m .50_0>._m .50 I50 .50 330] 00%. 02.50200 20:. .00.. 220. 5022.0 020 .5... 202.0 20 9mm. 0Polz< 00.0.2.0 0.002 mud—s. m.._0=._m> 2034.5 2:. mm... mhzmn=00< 108 4.4.1.3 INVESTIGATION OF ALL SEGMENTS (Appendix C, Figure C. 1 , Appendix D, Table D. 1) CCLTL highways had a higher overall mean accident rate for all accident categories than boulevard highways. The difference between the means expressed as a percentage of the larger mean accident rate and the significance of the difference as determined by the Student’s t-test was as follows: Percent Assidsntl’xpc 12mg: tics! All 53 Significant Rear-End 49 Significant Angle-Straight 24 Significant Angle-Turn 58 Significant Head-On Left 75 Significant Driveway-Related 85 Significant Other 30 Significant The intersection-related accident types of rear-end, angle-straight, angle-turn, and head-on left-turn are reduced by directing turning movements through cross-overs on boulevard-type highways. The driveway-related mean accident rate is significantly lower for boulevard highways when compared to CCLTL highways. There are more CCLTL highway segments with higher driveway densities which helps to account for the difference. There is no reason to believe that the mean accident rate for boulevard highways is the same as for CCLTL highways. Boulevard highways have a lower mean accident rate for all the accident types investigated. 109 4.4.1.4 INVESTIGATION OF UNSIGNALIZED SEGMENTS (Appendix C, Chart C2, Appendix D, Table D2) The CCLTL highways had a higher mean accident rate than boulevard highways for all seven categories of accident types for unsignalized segments. The difference between the mean accident rate expressed as a percentage of the larger mean accident rate and the significance of the difference as determined by the Student’s t-test was as follows: Percent Accidcflxps 2:132:00 figs; All 5 1 Significant Rear-End 54 Significant Angle-Turn 68 Significant Head-On Left 47 Significant Driveway-Related 87 Significant Other 1 8 Significant The only category that was not significant was the angle-straight accident type which is an intersection-related accident where through and crossing traffic movement conflict. Apparently, this type of accident is unaffected by the roadway type, as only the left-turning movements are removed from through intersections on a boulevard highway. The crossing traffic conflict remains the same for both roadway types. Rear-end, angle-turn, and head-on left-turn are intersection—related. The removal of the lefi-tums from the major boulevard intersections can be presumed to account for the difference in the mean accident rate along with the continuous friction that exists at the interface of the CCLTL and through traffic. The left turns from the CCLTLs provide a continuous opportunity for turning movement-related accidents to occur. 110 The driveway-related mean accident rate is inordinately higher for the CCLTL highway segments and could be partially attributed to the higher density of driveways on CCLTL highways, but also may be attributed to the movement from the CCLTL into driveways across opposing lanes of traffic. 4.4.1.5 INVESTIGATION OF SIGNALIZED SEGMENTS (Appendix C, Table C.3, Appendix D, Table D3) The CCLTL highways had a higher mean accident rate than boulevard highways for all seven categories of accident types for signalized highway segments. The difference between the means expressed as percentages of the larger mean accident rates and the significance of the differences as determined by the Student’s t-test are as follows: Percent Accidentlxnc 1332mm tics! All 43 Significant Rear-End 3 1 Significant Angle-Straight 16 Significant Angle-Turn 39 Significant Head-On Left 75 Significant Driveway-Related 78 Significant Other 32 Significant All accident categories exhibited a significant difference between the mean accident rates. The intersection-related accidents of rear-end and angle-turn exhibited a smaller relative difference than the unsignalized highway segments, except for head-on left-turn accidents, which exhibited a higher relative difference. The angle-straight mean accident 111 rate exhibited a significant difference for signalized highway segments where it did not for unsignalized highway segments. The difference between the mean accident rates for driveway-related accidents for signalized segments was less than for unsignalized highway segments which could be attributed to lower speeds in the signalized highway segments. 4.4.1.6 INVESTIGATION OF BOULEVARD ACCIDENT RATES: UNSIGNALIZED VS. SIGNALIZED SEGMENTS (Appendix D, Table D4) The signalized boulevard highways exhibited a higher mean accident rate for all seven accident categories when compared to the unsignalized boulevard highways. The differences between the means, expressed as percentages of the larger mean accident rates and the significance of the differences as determined by the Student’s t-test, are as follows: Percent Assidsnflxnc Diffmncc £1531 All 68 Significant Rear-End 80 Significant Angle-Straight 74 Significant Angle-Turn 75 Significant Head-On Left 68 Significant Driveway-Related 70 Significant Other 35 Significant The difference in the mean accident rate for signalized vs. unsignalized boulevard highway segments is significant as measured by the Student’s t-test in all accident categories. As there are more signals in urbanized areas, it may be that the increase in driveway-related accidents is due to an increase in driveway density in urbanized areas. The intersection-related accidents increased for the signalized segments as would be expected. 112 4.4.1.7 INVESTIGATION OF CCLTL ACCIDENT RATES UNSIGNALIZED VS. SIGNALIZED SEGMENTS (Appendix D, Table D5) The signalized CCLTL highway segments exhibited a higher mean accident rate for all seven accident categories when compared to the unsignalized C CLTL highway segments. The differences between the mean accident rates, expressed as percentages of the larger mean accident rates and the significance of the differences as determined by the Student’s t-test are as follows: Percent All 63 Significant Rear-End 70 Significant Angle-Straight 78 Significant Angle-Turn 53 Significant Head-On Left 85 Significant Driveway-Related 46 Significant Other 46 Significant The difference in the mean accident rates for signalized vs. unsignalized CCLTL highway segments is significant, as measured by the Student’s t-test, in all accident categories. The increase of the mean accident rate for driveway-related accidents would be expected as stop-and-go signals are located in more urbanized areas where driveway density is higher. The opportunity for accidents is thus increased with the increased driveway density. The intersection-related accident types of rear-end, angle-turn, and head-on left- tum had higher mean accident rates in signalized segments. This would be expected as the opportunity for rear-end accidents and turning movement accidents increases in signalized segments. 113 4.4.1.8 SUMMARY It is evident from numeric and graphical representations that the accident rates for CCLTLs are higher than those of boulevard roadways in all categories and that there is a significant difference between the mean accident rates in all categories except one. Also, it is evident that there is a significant difference between the mean accident rate for signalized and unsignalized segments for both boulevard and CCLTL segments with signalized segment accident rates being much higher. 4.4.2 BOULEVARD ACCIDENT ANALYSIS FOR VARYING MEDIAN WIDTHS (Table 4.4.2.1, Figure 4.4.2.2, Appendix C, Tables C.4—C.10, Appendix D, Figure D.1) 4.4.2.1.TESTING OF SECOND HYPOTHESES The second hypotheses was stated in part as follows: “The accident rate for roadways with a boulevard cross-section is influenced by geometric features and roadway environment; i.e., median width, posted speed, and average daily traffic.” In this section the median width portion of this hypothesis will be tested. The results of the statistical analysis for accident rates by median width are shown in Table 4.4.2.1. The method of analysis that was used was the tukey-b multiple comparison test. The tukey-b multiple comparison test adjusts for the number of comparisons that are being made. This test protects from finding that the differences between two means may be significant as determined by the Student’s t-test while in reality, the means are not significantly different. The tukey-b multiple comparison test was utilized for all of the comparisons that involved more than two means. The results of the tukey-b multiple 114 Ibo—3 250m: >m amt». hzmo.00< O¢<>waaom r.N.V.V mam > 0 0 0 0 0 w 0 _ mum; «NED PIOZKNO m..02< Ozm gm. much 4.? and... EDEN) 203...: 8r mum m...zma.00< 115 comparison test are shown in Figure 4.4.2.2. A sample of the tukey-b multiple comparison test is shown in Figure D. 1. 4.4.2.2 SELECTION OF RANGE OF INVESTIGATION The literature indicated that there are three categories of median widths for Operational purposes: median widths less than or equal to 30 feet that do not readily accommodate passenger vehicle U-turns (Category 1); median widths greater than 30 feet but less than 60 feet, that accommodate passenger vehicle U-turns, but do not readily accommodate larger truck U-tums (Category 2); median widths greater than or equal to 60 feet that accommodate all larger legal vehicle U-turn movements (Category 3). These three categories were, therefore, selected for analysis. 4.4.2.3 INVESTIGATION OF ALL SEGMENTS, UNSIGNALIZED SEGMENTS, AND SIGNALIZED SEGMENTS The difference discussed in this and subsequent sections within this chapter refer to significant differences in the statistical sense (a = 0.05) of the mean accident rate of two groupings of data being compared. The references to roadways refer to "boulevard" roadways. Roadways with Category 1 medians exhibit higher mean accident rates when compared to roadways with Category 2 medians for both head-on-lefi-turn and other types of accidents. This is true for both unsignalized and signalized roadways. In addition, these roadways exhibit a higher mean accident rate for the overall accident category for unsignalized roadways. 0.0.. E0280 5.0.: 02.9.. o... .00.. 00:00:. .023 o... 2.0 00:20.0... Swami»... 0 003 20... 00.00%... +m .000. 0002000 :00... him... 05 .00.. 00:00.: .0305: 0... VS. ought... 28053.0 0 003 20... 0800...... w+ .N .0.... . 09.0..» 0003.2. .30. mod 0... .0 agent... 3005:»... 0 080000 :w: 0.003003. E03300. .0.. 300. 5 00000005. 0523 5.002 mEb0> 00.. 00.003. 0>o=< 3.3-0.5 «NV... 0.59% m+ m+ m+ m+ +m m+ mmm: MmIPO +m ><3m>20 m+ m+ m+ m+ w+ m+ m+ ZMDFme... ZO-Dm mmhg ...sz.00< 055.500 ...».V.V mag... 00 0.. 00 00 0.. 00 00 0.. 00 0.. 00 00 0.. 00 00 0.. 00 00 0.. 00 .080. 0000 0000000.. 000.0000 00.. .0.0 0.00 0... 000.0000 .00. 0000 0000. .000 0..0. 000.0 0.... 0.000 0..00.00..0 :50. 000 .00 000 000 .00 00. .00 ..0 000 00. .00 00. .00 000 000 000 000. 00.. .0.0 0000 0000 50506.2. 0 0. 0 0 0 0 00 0 0 .0 0 0 0 0 0 00. 0 0 00 0 0 0.0525.— 00 00 00. 0. m. 00 00 00 00 00 .0 .0 00. 00 00. 000 000 000 000 000 000 8000,00 90000.0 0. 0 .0 0. 0 0 0. 0 0 . 0 0 0. . 00 .0 0. 00 0. 00 .l0 8:0 2020.0. .0. .0. 000 00 00 00 00 00 00 00 .0 00 0.. 00 00. 000 000 000 00. 000 0..i 08.2. 000.220.0— 000 00. .0 000 00. .0 000 00. .0 000 00. .0 000 00. .0 000 00. .0 000 00. .0 .500 00000 00.0. 00.0 00000000 000 0000 0.0. 0.0 80000.0 000 0000 00.0 00.. ..00. 000.0 0000 0000. .0000 0000 500. 0.0 0.0 .00 000 0.. 0. 000. 0.. 00. 00. 00. 00. 000 0.0 000 00. 0.. .00 0..0 .00. 000. 205.002 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 05.0.5.2 00 00 00. 00 00 00 00 0. 00 .0 00 00 00 .0 .0 0|. 0.. 00. 000 .00 000 800.50 0.00000 0 0 00 . 0 0 0 . . . . 0 0 0 0. 0 0 00 0. .. 0t. 3:0 2020.0. .8. .. 00. . 00 00 0. . 0. . 0. 00 00 00 00 00 00. 00. 00. 000 .00 :8: _ 0.2.20.0 02 E 0.0 00 000 0.0 00 000 0.0 00 000 0.0 00 000 0.0 0.0. 000 0.0 00 000 0.0 00 .0000 3000 00000 .000. 0000 0000. .00. 0000 0000 000. 0000 0.0.. 000. 0000. 00000 .0.0 .080 00000 0.00. 00.00. 0.00.. 00000 .000. .0.0 0.0 000 000 .8 00. 000. ..0 000 00. .00 00. .00 000 000 000 000. 00.. ...0 .000 0000 50506.2 .0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 05.0.5.2 00 00 0.. 00 00 00 0. 00 00 00 00 00 0. 0. ... .0. 000 000 000 0.0 .00 800.30 2020.0 0 0 00 0 0 0 0 0 0 . 0 0 0 0 0. 0 0. 00 0. .0 00 8:0 0.00030. 00. 00. 0.. .0. 00 ...0 0. 0. 00 .0. .0 00 00 00 .0 00 .00 000 000 .00 0.0 08.2.. 0.2 Ifllmllflunlfilflumlmu . |0||0| 0 . ....Il0ll0uln 0 .0 0.00m... m00m00m00m00m0 00%. 002.0 ><30>Eo 0:. 0000 20.9.02 20:. 0002... 020250 0002.. 020 «<00 000C .3.. mud! magzg 203...! 8.. ~30. mhzmo.00< 120 .20.. .5280 S008 .232 on. .00.. 50:00:. 0023 05 0:0 vacant... .fiomawmm 0 003 0005 00.03.05 +m .0.0. .5280 S008 .832 05 .00.. 50:00:. 003800: 05 .050 00:20b6 anomawfi 0 003 0.05 00.00%... m+ .354 35 05 .0 02.20050 2.03.5.0 0 00.00.00. ..m: 0.003.000m Eggnom 8.. 3&8 E ©00000ax0v 0.0025 00.000. wfiba> .0.. 0.300% 0>oc< 0003-020 «.096 95»..— 0+ 0+ +0 0+ 0+ 0+ 000?. 00.0.8 0+ 0+ . 0+ 0+ 0233.00 2050 0+ 0+ -000... 20-30.: 0+ 0+ 0+ 0+ 53-000200 0.5.3.00 0+ 0+ 0+ 0+ 0+ 01.0200 0+ 0+ 0+ 0+ 0+ 920-050 0+ 0+ 0+ 0+ 0+ 0+ 0+2 00nA0 00km 00V0V00 00uA0 00nA0 00V0V00 00km 00km 00V0V00 00005 00V0V00 0030 00uV0 00V0V00 00uV0 0030 00V0V00 00uV0 0000 .0005 00202000 0.0202000 000300290 000.00.20.02: 0.0202000 30. 121 4.4.3.3 INVESTIGATION OF ALL UNSIGNALIZED AND SIGNALIZED SEGMEN TS Urban roadways have a higher mean accident rate for the overall, angle-straight, and “other” accident category types than both the suburban and rural roadways. In addition, urban roadways have a higher mean accident rate for angle-tum and driveway-related accident types than rural roadways. These differences in mean accident rates are also true for unsignalized roadways. In addition, rear-end accidents are significantly higher on suburban roadways than rural roadways. When signalized, the urban roadways have a higher mean accident rates for “other” accident types than suburban roadways. They also have higher mean accident rates for head-on left-turn and “other” accident types than rural roadways. Rural roadways have a lower mean accident rate for all accident categories except head-on left-turn and other accident types when compared to suburban roadways. The same is true for unsignalized roadways except for angle-straight and other accident types. When comparing roadways with the same two posted speed categories for signalized roadways, the only significantly different mean accident rate was for head-on left-turn accidents with the suburban roadways having the lower mean accident rate. 4.4.3.4 SUMMARY The clear trend is that roadways with lower posted speeds exhibit a higher accident rate. This is true for unsignalized roadways also. Signalized roadways have only a few accident types which exhibit significantly different accident rates at different posted speeds, 122 but these accident types also have the same trend of higher accident rates for lower posted speed roadways. These trends of higher accident rates for lower speed may not necessarily be cause and effect, but may be the result of lower speed roadways being located in areas of higher development with resultant increases in points of conflict and congestion providing for increased opportunities for accidents. 4.4.4 BOULEVARD ACCIDENT ANALYSIS FOR VARYING AVERAGE DAILY TRAFFIC (Table 4.4.4.1, Figure 4.4.4.2, Appendix C, Tables C.18-024) 4.4.4.1 TESTING OF THE SECOND HYPOTHESES The second hypotheses will be tested in this section with regard to the relationship of boulevard roadway accident rates to average daily traffic. The hypothesis was stated in Section 2.10. The results of the statistical anaylsis for accidents by average daily traffic are shown in Table 4.4.4.1. The method of analysis that was used was the tukey-b multiple comparison test as discussed in Section 4.4.2.1. The results of the tukey-b multiple comparison test are shown in Figure 4.4.4.2. 4.4.4.2 SELECTION OF RANGE OF INVESTIGATION The frequency distributions were divided into three volume categories: 0 to 30,000 vehicles per day (category 1), 30,000 to 60,000 vehicles per day (category 2), and greater than 60,000 vehicles per day (category 3). The accident records were, therefore, aggregated into these three categories. 123 Din—5h >.=< >0 mm...m430m p.966 MAB; 2 8F NF. 2 RF 3 2 8F we 2. 8F 8 2 8F 8 2. RF 9. 2 RF 9. F58. NSF 088 88 NRN 88 89 83 8% N08 mNoN No3 FNNN Fm? 8mNF NSF SEN wk? «82 mNoNv 03.8 R3. 53. own new NNN FmN mam FNN 3: SN EN 3: FFVN moF RN N8 wNm N8 8F: 82 3? 8R FmFN 535%:— m o FF o o o o o o n o o o o o 8 o o 3F 0 No 535.52 8 m: FNF mm mm. Fm N 8 2 N 9. 9 Na Fm 9: BF 3N FNN F8 0:... E. :25st 2856— m 9 NF m m N n m 9 m n o N w FN N N mm 9. NF. 8 35 2835. 2F 03 8F 9. 8 NF” NF MN 8 8 8 8 8 8 m2 o8 N8 9.» m8 8... one 522— mm EF 8». mm 3F 8m mm 3F 8m 8 EF 8m 8 EF 8m 8 EF 8m 8 EF 8m 9% ovoNF 88F. FNNF :5 88 8.. mo? 53. 82 8% No; 92 :8 8:: 8N» 2N8 88F N82 N99. N33 58. mNN me «Fm Na 8N NmF Fm NF 32 mm m2 9: NF EN mFm n; m; we. m8 NE: mFNN Easing o o o o o o o o o o o o o o o 2 o o 8 o o .5525. cm R 8 «N 8 8 a NF 9F 3 N N 8 8 E m: vNF FN 8N mmN MN 8.638 285$ N o F. N n F F N F. N N F F. n F. B 2 m 8 FN NF 35 2856. 8 Nm 9: N N m N a F F NF 8 m N N N 03 8F FF. NNN RN 9.: c8:— 85.22052:— mNF 8N 3F. mNF SN 3... NF 8N 3% mNF SN 3% NF SN 34. NF SN 3% NF SN 3% F58 NB: EN» 88.. 38 58F 89. Fm: 89 St New» 28 8mm 85 8:: 2:: vaFm 83m NSNN F88 882 New: 55 8..” m3 NNN FmN mam FNN x: «N 82 3F FFVN 8F RN N8 mNm N8 8F: 82 v3: 8% SN sagas. o o o o o o o o o o o o o o o o o o o o o 5355.2 8 SF 8 8 we 8 8 8 Fm FN mm N .3 t. Fm 6F 3N 8F on F9. F8 8550 2356 o o m n n F N N F. N N F m F. F. 9 F: N N 0N 0F 35 2386 Fm NFF NFF Fm mm 2 NF 2 NF N 8 NF 9. mm Fv 3N NFN fl BF. 98 EN :82. .3. M C V V 8 V M 8 V JV 8 V V 8 V V 8 V V 8 W a m a a m a a m a a m a a m a a m a a o ,a q u q u q u q u q o u q a a m ... 9 = 9 = 9 = 9 = 9 = 9 = 9 > = w e m w e w m. w w e w e w w v m m a m m a m m a o o a m m o m m o m w a u m a u w o u u u u m a. w a. a. w k a. m m m m m m m 0 0 $9» 3:5 ><>>m>Eo zma km: 20.93: 2%» mFoz< .EozEm whoz< ozm «(mm mmat .:< mud—2 m..0=..m> 2034.2 8w «mm mhzmn=004 124 .88 EoEooa F808 F233 05 was 5:58 53>» 05 28 8:20va 58¢?me m 33 205 3:365 +m .88 60208 :85 6.32 05 was F8608 5305: 05 can 855th 58¢?me m 33 22: 8329: m+ Agog wad of 3 8:80wa tsunami a 8853 ..m: mxaaumom Ego—8m Now 9% Fun 832? E wommoaxov can; 3ND owfio>< wash; F8 $38M «>22 $3-020 N666 9.5»:— m+ m+ m+ m+ m+ m+ mun—i MMIHO +m +m +m +m ><3m>aa 53%qu m+ m+ ZO.D 3203025 F8 333% 68; ".m66 935. on": on 7: EN": o3": NNmn: Emu: me m4m2<3m>20 ZMDPLLmJ o 8.». S. 2 «Nd co. _- 6. 0 M26 mm. _. ON 2 ZO.Q; .226265 35:99.5 :3 N "53>; 35:09.85 35:09.5 zah N 33>; REFEREE 38:09.5 E>S<< :32 $522 522 2>2\< =32 memEOmm QmN_1_h Hip—Z ~NMVW‘DI‘NG A. Z vaQmo am: 50m 0: 33°.— flomuovm _O.n—=Oo 147 ..O.. .85 533:5 3.5 3.4.. 2.3: mo o: 327. .3 Bat? 83 .3 “388:3 5% ad. mm mm nm em mN 3N mm mm 3 ON 2 E 2 8 w _ E M 9—4 N —i ~ ~ c I—t ~NMVWOF°¢© 325.2200 H.592 .E mdz ¢<> 5: ~53 0mm ED 3388 ..0: SEC..— o: 8:8 nozoom 35:8 148 no 2. .02: vaQmo :9: 50m Panda: 3 02:0 0: Dun-O.— ..n. 325 3332.5 3.5 "3. as»: 3 38:8 a... an an 5N on vs ms an a a— a— h— e— v— n— N- = :0303 .9580 149 .3.. 325 538:5 3.5 54 2.35... .3 ea .8... 3 .353» a... 2 038:8 a... an on an 5 on ma 3 an an 5 an a. 2 S 3 n— v— n— n— = E a a h w v. v n n _ 8.5228 EEO E 3.2 E ”3:: 32> :5 ME:— Om: .qu ll UZEZH DZ—ZZ—Umm vaQmo am: SEC.» 0: 3:8 5:08 .0250 150 a? .85 258:5 5.5 £4 25:. ... 2. .8.: E 35.3 3a.. 3 38:8 3% an an an S on mm vs an «N a an a— m— h— o— n— v— m— N— = 6 II! "NMQWVOI‘NG Bin—2:50 ”E: E 3.2 .E 3.2 55 :3 ”.53 ug 4% 3335: @255 02—225”:— vaQmo fir. SEC... 2. 8:2 8:08 35:8 151 PHOTOLOG USE BY CONSULTANT PERSONNEL When the Traffic & Safety photolog facilities in the Transportation Building are made available for use by consultant personnel, the following work rules shall apply. Any consultant requested changes to these rules shall be approved by the MDOT Project Manager and the Photolog Supervisor prior to implementation. Two types of viewing equipment may be available - laser disc and 35 mm film viewing machines. Current film is available for all districts except District 5 (Grand Rapids) and District 9 (Metro Detroit Area). The most current photolog for these two districts is only on disc. W Work hours for consultants personnel shall be 7:30 am. to 1 1:45 am. and 12:45 pm. to 4:30 pm. Monday through Friday. No work is allowed during the normal lunch hour (1 l :45 am. to 12:45 pm.) In deference to state employees in the photolog/graphics area. No work will be allowed on Saturdays, Sundays, or holidays. Disc and film viewing equipment shall remain where it is currently located. NO MOVING of equipment except for minor table angle adjustments shall be made without the assistance of photolog personnel. The 35 mm viewers are not as subject to motion disturbance as the laser disc system, but are much older and should be considered fragile. Consultant personnel shall be responsible for finding and properly loading the appropriate disc of reel of 35 mm film. Photolog personnel will instruct consultant personnel in the proper loading or hanging procedures and will be available for assistance if something goes wrong with either system. In case of the laser disc, if it locks up and cannot be restarted via a “hard reboot,” photolog personnel will try to resolve the problem. If the problem cannot be fixed in a short period of time, the consultant employee affected may be required to stop work until the equipment is repaired. If the 35 mm film rips, stop the viewer immediately and notify photolog personnel. Do not take the film off the machine. Repairs will be made by photolog personnel. If there is any question regarding use of the equipment, DO NOT ASSUME!!! PLEASE ASK!!! Figure A.10: Work Rules 152 MEDIAN CROSS-OVER STUDY VIDEO LOG INVENTORY INSTRUCTIONS 1. Record control section, date, direction, variable being recorded one time, unless a change occurs, then record it only when the change occurs. 2. Record one event per pass. G) First pass record median width, lane width and speed limit. E I D . . Use the comment area to note unusual circumstances, such as closed drives or merged areas that are treated as intersections. Note any problem areas for resolution. Record all events for both directions. Use only codes provided. Make a decision on the situation, code it accordingly and note any problem or observations in the comment area only. Figure A.ll: Field Data Collection Instructions 153 MEDIAN CROSS-OVER STUDY VIDEO LOG CODES W Five digit numeric xxxxx DAIE Six digit numeric xxxxxx DIRECIIQN One digit numeric 1 northbound 2 southbound 3 eastbound 4 westbound YARIAELE One character alpha G median width H lane width J speed limit MILLEQINI Five digit numeric xxx.xx note: decimal indicated by digit placement W Five digit numeric xxx.xx note: decimal indicated by digit placement LANEMDIH Four digit numeric xx.xx note: decimal indicated by digit placement SEEEILLIMII Two digit numeric xx Figure A.12: Field Data Collection Codes 154 :0: 3025 553—30 San— um~.< ohswmh .3 2. .3.—m .3 But? 82. .3 382.8 33 an an an 2 cu mm vs an «N a an a— w— 2 e— m— 3 n— n— = G '— F'NHQWVOl‘aa EBB EBB :2... .E 3.2 E 5.: 5S :3 “ES 8m .56 H724 ZSGm—E Gama—m UZ—A—Zm CZ—ZZ—Omm vegmo ..0: $20..— 2. 2:8 8:08 35.8 APPENDIX B DISCRIMINATE ANALYSIS RZADHAY=L in groups defined by RALLRB 866 Q 866 D I S R I M I RAL v U REVISED DATA GROUPED 155 N A N T A N A L Y S ’Unweighted) cases were processed. of these were excluded from the analysis. (Unweighted) cases will be used in the analysis. Number of cases by group Number of cases Weighted RALLR3 Unweighted 1 289 2 288 3 289 Total 866 3roup means RALLR3 N Total RALLR3 N Total 26235 36713 43933 35626 I 2 6 10 6 ADTB .29412 .19444 .56401 .09700 NTDENS .31142 .19444 .72664 .41109 289. 288. 289. 866. O 0 0 Label 0 TO 130 130 TO 368 >368 COMDRDEN 14 .52249 .3888? .56055 .82448 11. 11. 11. 11. 65744 60069 54671 60162 H H M CURB .77509 .56944 .39100 .57852 NUMLN .19377 .68056 .99308 .62240 Figure B.l: Discriminate Analysis Example DRVDENS 7.72318 15.01389 21.05190 14.59584 PERCOMM 5.00000 4.42708 3.47751 4.30139 ROADWAY=1 RALLR3 Total 154477 130571 174878 153335 POP .43253 .86111 .46713 .50924 Group standard deviations RALLR3 UUH Total RALLR3 Total HUM“ Tota 17601 22876 23271 22583 ADTB .23102 .37659 .62707 .48494 INTDENS 160242. 116574 175316. 153687. .31083 .41904 .8841? .91089 POP 78061 .51671 47267 16687 Figure B.1: Continued SIGDENS .20761 .76389 3.23875 1.40416 COMDRDEN 5.07790 14.78916 22.02848 16.46848 .47539 .62770 .72077 .61725 SIGDENS .94931 1.64427 .75412 .33707 '0 k! 156 SPD 52.54325 50.00000 46.15917 49.56697 CURB .41825 .49602 .48882 .49408 NUMLN .53732 .90414 .91284 .86788 SPD 4.75282 6.12017 7.33302 6.68988 MED 53.00346 53.70833 55.57093 54.09469 DRVDENS 11.16198 22.14113 28.41167 22.42017 PERCOMM 3.02765 3.00897 2.52632 2.92850 MED 28.77131 26.52114 24.25034 26.56988 157 ROADWAYai Pooled within-groups correlation matrix ADTB COMDRDEN CURB DRVDENS INTDENS LNWD ADTB 1.00000 COMDRDEN .14684 1.00006 CURB -.56956 -.27471 1.00000 DRVDENS .11588 .86395 -.27409 1.00000 INTDENS .24290 .15561 -.22098 .07544 1.00000 LNWD .01947 -.09211 .20623 -.04309 -.46449 1.00000 NUMLN .70026 .25502 -.66765 .24277 .26882 -.17414 PERCOMM -.48779 -.05578 .32788 .03613 -.26707 .06084 PO? .06020 .12669 -.07972 .08754 -.10415 .06359 SIGDENS .33366 .14332 -.25808 .12876 .27008 -.01102 SPD -.48876 -.29715 .59742 -.25463 -.31533 .19341 MED .07872 .18209 -.04336 .15261 .04368 .03860 NUMLN PERCOI‘H P0? 5 IGDENS SPD MED NUMLN 1.00000 PEROSWH -.30893 1.00000 PO? -.01057 -.25579 1.00000 SIGDENS .33812 -.14013 .13891 1.00000 SPD -.57100 .29892 -.11707 -.32115 1.00000 MED .08721 -.03526 .03406 .04825 -.17415 1.00000 wilks' Lambda (U—statistic) and univariate F-ratio with 2 and 863 degrees of freedom Variable Hilks' Lambda F Significance ADTB .89625 49.9516 .0000 COMDRDEN .89495 50.6524 .0000 CURB .89888 48.5409 .0000 DRVDENS .94079 27.1590 .0000 INTDENS .81060 100.8248 .0000 LNUD .99462 2.3327 .0976 NUMLN .85606 72.5524 .0000 PERCOI‘H .95393 20.8395 .0000 POP .98611 6.0769 .0024 SIGDENS .68155 201.6157 .0000 SP0 .84578 78.6795 .oooo MED .99833 .7198 .4871 Figure B.1: Continued ROADWAYs; On groups defined by RALLR3 Analysis number Direct method: - - D I S C R I M I N A N T .J RALL REVISED DATA GROUPED BY 3 158 A N A L Y S I 8 all variables passing the tolerance test are entered. Minimum tolerance level .................. .00100 Canonical Discriminant Functions Maximum number of functions .............. 2 Minimum cumulative percent of variance... 100.00 Maximum significance of wilks' Lambda.... 1.0000 Prior probability for each group is .33333 Classification function coefficients (Fisher's linear discriminant functions) RALLRB - 1 2 3 0 TO 130 130 TO 368 >368 ADTB -.0002802 -.0002767 -.0002911 COMDRDEN .1230054 .1353073 .1731393 CURB .2282162 .1427605 .0608063 DRVDENS ~.O725388 -.O703095 -.0889914 INTDENS 2.2167106 2.2871628 2.3320329 LNUD 42.4198558 42.8175570 43.0803928 NUMLN 21.7449300 22.2543821 22.114655? PERCOMM 1.1911056 1.2132113 1.1703634 POP .0000289 .0000280 .0000284 SIGDENS -1.7871039 -1.7971094 ~1.1815041 SPD 2.3229964 2.3241538 2.2818160 MED .0664825 .0634075 .0606841 (Constant) -338.9193751 -345.0062050 -346.3490705 Figure B.l: Continued 159 Canonical Discriminant Functions Pct of Cum Canonical After Hilks’ Pcn Eigenvalue Variance Pct Corr Fcn Lambda Chi-square df Sig 0 .578511 469.308 24 .0000 1' .6332 91.55 91.55 .6226 : 1 .944801 48.690 11 .0000 2‘ .0584 8.45 100.00 .2349 ' Marks the 2 canonical discriminant {unetions remaining in the analysis. Standardized canonical discriminant function coefficients ADTB COMDRDEN CURB DRVDENS INTDENS LNHD NUMLN PERCOMM POP SIGDENS SPD MED Figure B.l: Func l Func 2 -.14306 .27392 .41843 -.o7793 —.03921 -.03175 -.21005 .29512 .39954 .48962 .1982? .23555 .11370 .6104? -.04227 .15400 -.01987 -.20839 .66724 -.69915 -.14508 .15719 -.07655 -.06814 Continued ROADWAY=L Structure matrix: 160 9ooled within~groups correlations between discriminating variables and canonical discriminant functions (Variables ordered by size of correlation within function) Func 1 SIGDENS .85278. INTDENS .59716' SPD -.5332S' COMDRDEN .42427' CURB -.40624' DRVDENS .30272' PERCOMM -.27513' LNHD -.08968' MED .05131' NUMLN .48333 ADTB .40384 PO? .09694 Func 2 -.34098 .36721 -.19823 .24184 -.37004 .29010 -.O7937 -.07334 -.00427 .58838' .46262‘ -.37311' ' denotes largest absolute correlation between each variable and any discriminant function. Canonical discriminant functions evaluated at group means (group centroids) Group Punc 1 -.81954 2 -.25558 3 1.07424 >Harning I 43 Func 2 -.23295 .33290 -.09879 >MXMEMORY (the maximum amount of memory that can be allocated dynamically) >has been reached. To increase this value, use the SET MKMEMORY command. Figure B.l: Continued 161 ROADWAY=1 Classification results - No. of Predicted Group Membership Actual Group Cases 1 2 3 Group 1 289 227 48 14 0 TO 130 78.50 16.6% 4.8% Group 2 288 123 111 S4 130 TO 368 42.7% 38.5% 18.8% Group 3 289 40 55 194 >368 13.8! 19.0% 67.1! Percent of 'grouped' cases correctly classified: 61.43% Classification processing summary 866 (Unweighted) cases were processed. 0 cases were excluded for missing or out-of-range group codes. 0 cases had at least one missing discriminating variable. 866 (Unweighted) cases were used for printed output. ROADWAY-1:51GNAL DENSITY-O Number of valid observations (listwise) - 609.00 Variable RALL TOTAL ACCIDENT RATE Mean 221.463 8.8. Mean 9.915 Std Dev 244.691 Variance 59873.819 Range 2717.808 Minimum .00 Maximum 2717.81 Sum 134871.132 Valid observations - 609 Missing observations - 0 Figure B.l: Continued APPENDIX C DATA ANALYSIS RESULTS 162 3noE3< 2:5 8:53qu 5.50 can Ego—..em ..8 8.3— «doEooe. 3.0 95“:— . 2553 a3 8.80 I .5383 3 we; EB :26 85oz :3 EB 5.5m Bu Essa. so 3o: .65. 22¢ 53. __< mhzmn=00< mzj 21:... mmhzwo oz< o¢<>mJDOm KO". mhg .—.2m0_oo< $31M 310lH3A NOHTIW OOI» 83d 31.“ LNEOIOOV 163 monas- Eshéou— .5250 can mango—..em gum—.3325 ..ea 6.88% «5293.. "MO 959..— 5303 “—0 wt: msz zxah hum; zuhzmo oz< wn¢<>w430m emu-22010.23 KO"— wmhg hzmn=00< sa'uw 313IH3A NOH1IW 00L 83d 31“ .LNBOIOOV 8.3.— Eahéoq ..850 can mango—56m ovum—2&5 new 8.5— .dou_oo< "MO 2:»:— wmzj 21:... hum... zmhzwo 02¢ mo¢<>un50m amN_._<20_m mo". mmbg hszUO< O .8. $31M EWOIHEA NOITIIW 00L 83d 31“ .LNBOIOOV 165 38282. 3. as 62.5 .582 E as. .8263 2.32:8 "so 2.5... 95:58 86.893 88.292:- 8553. .2- 5.932599. 8:: 8:78 8.12 whzm900< .._._< KOu Ibo—3 ZSONS >m mh<¢ ...zmo_00< o¢<>musom 83TH 3"IOIH3A NOITIIW 00L 83d 31“ lN30IOOV 166 38282. sadism 5 £23 58: E and .8282. 2.58: "no 2.5..— 385 3.8m- 31.6m- Mkzmeonz nzm ¢m m._.<¢ ...zmn__OO< a¢<>mnsom 931W! 310|H3A NOITIIW ML 836 31“ MEOIOOV 167 88263. 585.265. 5 £23 5:52 E as. .8263. Ens—=5 6.0 2.6:— whzmn=00< PIG—(mhm mu_02< mou— Iho=s 250m: >m mh<¢ ...zmo_00< o¢<>musom 831M 319lH3A NOITIIW 00L 83¢! 31W 1N30|OOV 168 3828... Essa—us. as 52.5 .582 a. 3.2 .828... 2.32:5 .5 2.5... 6.21.1... . mamas .3... «.. . MW. . ...... ...a»... an. «A... .. whzw900< zzah 5024‘ 10m FPO—3 25am! >0 mk<¢ PZMQOO< o¢<>mJDOm § 0 § .55 § 3 ° 831M NOIHSA Momm 00! 83:! 31W LNSOIOOV § 3 169 35263. floa— eoueaofl 5m 523 5:55 .3 Sam EoEoo< Ego—56m £0 9...»; 8558. .2- m...zmn.00< hum.— 20 3: mon— :._.0=$ 250m: >m mh<¢ hzwn=00< o¢<>mJDOm 831M S‘lOIHEA NOI'I1IW 00L Bid 31.11! lNSOlOOV 170 38293 5525 5 £23 5:52 3 as. .5293. E5225 ad 25:. £39m 35.8ng 823w van—a583- fiasuom __<>>m>Eo :0...— E93 250m: >m mh<¢ hzmo_00< n¢<>m.50m §§§§§§§ 831!" IIGIHSA NOITIIW Ml 83d 31W LNEOIOOV 171 3:0293‘ no non: .550 he 52>? 5:52 3 85— EoEoo< ESE—som— 5—0 9.5»:— 8nn2 8v§von onuvz .8. .33 831W BTOIHM NOITIIW 00L 83d 31W LNEOIOOV «#2369. ".0 $9; $5.05... :5? 233.2,xnwhgpzwewuc..amciio , 8,“ 172 3523< =< ..8 :35 18am .3 83— 23293. Ego—son "3.0 95w:— 88893 3?ng 35:53 83.222: I 355$ .25 EEJOWEW mmunm mmvmvmn mhzmQOO< ...: KO"— 223 Dmmcm >m mh<¢ hzmn_00< D¢<>mnsom 35m $31M E'IOIHBA NOI'HIW 0M 83d 31“ 1NEOIOOV 173 3523< cum—Laom .Su =5: 3on 3 Sam EoEuo< Egg—gm "2.0 0.5“:— ...—530mg mmuAI mmvmvmm mMHVm mhzm900< 02m ¢m m._.<¢ h2m900< Guam-50m 931m anon-Ian NOITIIW ML 83:! 31W LNBOIOOV 174 3528< Emmahméuna he an: 3on .3 Sum «5293.. Ego—gm 23.0 9...»:— goon 88...:ng gnaw 85.898:- gaom __0 m._.<¢ hzmn=00< DEE-500 175 35293. 55.2»: E gs: .525 E 3.2 .5292 .5338 ”2.0 2:3,.— whzmo.00< 283k m402< mon— ._._E_4 nwwnm >m whim hzmo_00< n¢<>m1500 '- SB'IIW 310IH3A NOf'I'IIW 0“ 83d 3].“ LNSCIIOOV 176 35292 :3 goes: 5 :3: Beam E 35— .5293 2.32:8 .20 as»: 35:58 ..(I mhzm900< hum.— 20 3: KO". ...—2... ammmw >m mh<¢ ...2ma.00< D¢<>MADOQ 831W! 313|H3A NOITIIW 00L 83d 31“ LNBOIOOV 177 3538< .mgBoEn ..8 :85 68mm .3 Sam EoEoo< Ego—som— "30 Earn tSJOwwn—w mmunw mmvmvfi mnnvm mhzm900< ><>>m>_¢a mo". .22... Damn—m >0 m._.<¢ hzmn.00< o¢<>m.50m $31M roman NOI'HIW 001 83d 3.“!!! .LNBOIOOV fluoEoo< .3 893. ..0—3O he :3: .5on .3 33— 2.0293. ...—«>22.»— 05.0 PEME 178 2:253 8%:ng 85:96 85.292:- acosuom __0 whim ...zmo_00< a¢<>mJDOm 331W! 31OIH3A NOHTM 00L 83d 31W .LNECIIOOV 179 $5235. =< .3.“ 953,—. .32— ow§o>< .3 83— .523< Ego—gm "EU 95w:— acoEmwm 85.8% D 35595 3.222: I 355$ .25 gnu—(K... >420 m0<¢w>< 08.8 «A ho< 80.8 v ko< v 80.8 80.8 uv Ho< whzmn=00< ._._< m0“. Din—(KP >.=< >m m._.<¢ ...zmo_oo< o¢<>mnsom W8, con $31M E'IOIHEA NOITIIW 00L 83d 3.“!!! .LNSGIOOV 180 35299< tum—Lao"— ..8 953,—. .32— uuauo>< .3 Sum «5303.. Ego—son "3.0 unaut— Qu—E >420 m0<¢w>< 08.8 "A ._.D< 80.8 v .22 v 08.8 whzm900< azm 3! KO". Din—(m... >.=< >0 m._.<¢ ...sz_00< o¢<>mnsom $31M TIOIHBA NOITIIW 00L 83d 31“ .lNEOIOOV 181 35293 .fi_§m.o_u=< he 953. £5 09:93. E 8.2 .5293 23225 58 an»; 35.298:- aieaow =420 §g< 08.8 «A ...0< 80.: v ._.n_< v 80.8 80 8' uv .E( .8. § § § 331!“ 310lH3A NOITIIW 00$ 83:! 31W 1N30IOOV fi m...zma.00< PIG—(chm WJOZ< KO". Gin—<0... >.=<0 00<¢m>< >0 mh<¢ ...zmo_00< 0¢<>m.500 182 8523< 559.2»: ..8 052,—. .350 ou§o>< .3 8nd Equo< ...—«>250 And 0.5»:— QE >3 030‘: 96.8 «A h0< 96.8 v .52 v 95.8 96.8 Iv ho< 00.2090“: 2%.: 0402< 00n— OEE... >n=<0 mG<¢m>< >0 05.4”. ...zmo_00< o¢<>04300 831M EWOIHSA NOITIIW 00L 83d 31“ MEOIOOV 183 BOULEVARD ACCIDENT RATE BY AVERAGE DAILY TRAFFIC FOR HEAD ON LEFT ACCIDENTS 810 gas '— .- 831"! FIOIHEA NOITIIW Ml» 83d 3].“ LNBOIOOV 30,000 < ADT < 60.” ADT >8 W.” AVERAGE DAILY TRAFFK: ADT 3 ugw>< 28.8 IA ...0( 98.8 v #9 v 80.8 95.8 Iv ...n.< m...zmo.00< ><30>.¢0 mo". Din—<8... >.=<0 wD<¢m>< >0 uh: ...zmn.00< 055.500 831" E'IOIHSA NOITIIW 005 83d 31“ lNBCIIODV 185 85234. no 89:. .659 .3.. oEF—H >55 09293.. .3 8am =523< Ego—som— "vnd 0.5»:— E >430 wgm>< 08.8 "A Po< 80.8 v ._.D< v 80.8 80.8 «V ...o< 85:96 in whzmQOO< no mun—>5. mmEO m0". Ding... >.=< >m uh<¢ tame—004‘ o¢<>mn50m $31M 313lH3A NGTIIW 00L 83d 31“ MBOIOOV 186 fio>OémEU ..8 85— EoEua< ...—551:5 "mad «...—arm 3.29m .9583oI 85.292: .93826B 8&ng _%._o._m I 85.298: gseéml #2303502: m¢m>0wm0¢0 m0“. whim bzmo_00< o¢<>m4300 831W! 310M3A NOITIIW GM 83:! 31W lNSOIOOV 187 9.05-830 35:99.5 .5.— 85— 30299< Ego—gm GNU 95»:— ENQOSSE m¢m>0mw0¢0 ..(20F0m25 KO". mh<¢ .—.sz_00< O¢<>m.50m 831M 3'IOIH3A NOITIIW 00L 83d 31“ LNBCIIOOV 188 9.36-395 3:339:65 .8.— mouam EoEoo< Ego—som— KNU 95w:— h2m900( “.0 us: «.553 85.235 88:58 Bugaal ....coeuom _.0mmO¢0 A<20....Omm_ou_m mo“. mmhé hzmo.00< O¢<>mn=aOm $31M NON-EA NOITIIW out 83:! 31W LNBOIOOV APPENDIX D SIGNIFICANCE TESTS 189 3.8.» Aha—UUBhagiom 3A— 035. 00. 00. 00. 00. f 00. 00. r 8. =82 .856 a 00. 0 00. 0 8. v 00. v 8. 0 00.0 00. 0 =86: alvbn: mm. mm. no. m0. mm. mm. . 00. =3uoco .0035 u— 00. . 00. 0 00. v 8. 0 00. . 00.0 00. 0 .3600 «TV 1 0m: . 0N9? k; T No. w. Nm. - 00.0w. 02V... «80. 00.0me 00.35 00.3: 8.000 8.00vw 00.0FNF 8.00: F 00.0 00.0 8.0 8.0 8.0 00.0 00.0 00:20.50 :82 a . a. I 0062 00.80 00.35 00.000 00.35 8.80 00.5 00.000 00.3w 8.000 8.35 00.08 8605 8.000 acct->530 «0.802. VN. F05 wwfimmmn VNNm: whvwoow v0.v8n N0. 5km nmdmo wad—mo HQROO on. 3:00 no.mnmov «QBNva 8.000an oocItu> 84.09 00.00— de: mo. rN made 3.2 00.0? 5.94. owdw ONOV 8.00m 00H: anion cudon :82 .2. 000 n3 .3 .2. 000 Q> .5 H500 Q> .5 .2. .000 Q> 40 .t. 400 Q> .3 .E 400 Q> an .E 400 Q. 40 wmnik EWILO >S$M>EQ Llfliodfil 2&3.“ .302v. IQEthGZv. QZWicflk j< _ E 305:; .5025 0:_E:mn< oiEamIgF 38...-.— .8. 350935518 190 .8... 0003003. 000.2000: J.H.—00.0.5280 ”0.: 0.000 8.. 00.. 8.. 8.. 8.. 8.. 8.. 3.02.8000. 8.0 8.0 .0.0 8.0 00.0 8.0 8.0 8.0209 0 00. 00. 00. 00.. 00. 00.. 00.. .308 .8000. 8.0 8.0 8.0 8.0 8.0 8.0 8.0 3.08.? 0 00.0. 00.0.- 00.0- 00.0. 00.0 00.0. .0.0. .80 0 8.000 8.0.... 8.00. 8.0.0 8.000. 8.000 8.0.0 .0 8.0 8.0 8.0 8.0 8.0 8.0 8.0 09330.0 08: 000.858»: 8.0.. 8.000 8.9 8.000 8.000 8. .0.. 8.000 8.9 8.000 8.0.. 8.08 8.9 8.80 20002800 7.0.00.0. 00.0000 00.00000 00.0: 00.00... 00.0.00 8.0.. 8.0000 00.0.00 00.0008 00.0000 80.08. 0.00000 8000.5 _00.0.. 00.8 .000. 0..0. 00.0. 00.00 00.0. 00.00 00.00 0..0... 00.00 .0.0: 0....00 08: _ 00%. «mic $38,000 00320.90: 20.: 0:0,? 103000.305. ozméfim ...? _ .85.; .308: 00.503. 0.0.50.3. in»... w><>>0<0m DmN3m4300 191 what—vac“ fiONu—aflM—m hawk “meal“ AFN—Doghgro—flcm "M.A— o—DNH. 8. a. S. 8.. 8., 8. 8. 8. . 8. . 8. . 8.o 3.0 8.. 8... no. 3. 8. 8.. B. F 3. 8. .320 .826 ._ 8. . 8. . 8. . 8.0 8.0 8. . 8... 3.25 any? a... 3i. 8. F- Rd. 8..- 8c. «2:. 52 8.5.. 8.0mm 8.3m 8.8m 8.8m 8.8m 8.8... 2 8... 8.0 8.0 86 8.0 So 85 See-so =82. . I I; 8.8» 8.3a 808 8.5a 8.8m 8.53 8.8a 8.3“ 8.8m 8.3a 8.8m 8.3a 8.8a 8.5“ 83.238 2.8% :aoa— 2.208 «0&3 8.8:. 8.8%. 8.33 8.89 No.83 3.9.5 2.3%: 9.808 $839. 83:." 8.5.; «SE 8.3. 2.5. 3.? «25 3.8 8% 8.3 an: 8.3 «18¢ 3.8" 2.3“. 8.8» :8: .58 915 .500 Sin .500 o ..m .58 913 .58 9.3 .58 min .58 93m 8%: mqu ><§m>io tfléodfi: sz.w..oz< 53:». 5 nzm. .m? _ w _ _ F r 88E .995 2.533 4.5.on 3.5 w><>>om430m 192 £936an coganflmEon—gufiub .8.— »83 Ego—som— "vd 93:. hm; no; 8 F ha. . um. . ha 1 so. . =36): 3025.“.— oo. v 8. u 8.0 8. v 8. u 8. u 8. . =86} AWN—IE— mo. . mm. Y mm m mm. . no. . no r 00. =8.oco ioatn «— 00; 8Q 80 8.. 8.9 83 8. mm. Y mm. W. 3 m- 00.me 8.6.. 82. mv. T 8M3 ooNNm 8.3 86mm 8.08 8.8m 8.. 5 oo. o 8.0 8.0 8.0 8.0 8.: 8.5mm 8.80 8.5mm 8.80 8Nmu oomoo 8.5mm 8.08 8.5.3 8.80 8.591. 8.000 8.5mm 8.3 K. NwFNF 8.3 «m. FNmN 8.9m nodmnu so. an; gnaw no. 5v 8.9%: ~08va Mr. 08% MN. 0N8 8.3555. 5. whoa NOV: no.3 mndv mfimw hmdm NN.o_. 8.3 mud? 8.5m mafia Bonn mafia 8mg ovéwn 05 052: 03 052: 9m 0523 05 052: 05 0523 GB 0523 06 0B2: mmnik km: 0 > §m>EQ k 0% 219750 < stkam-Es QZMig j< _ ~ m><>>0<0m DMN_I_<20_w\omN_I_m.50m 193 33.8»— efifiuflmfifiaaia E .8..» 4.500 "md 2.3 oo 8. nm.— 8. no .3 F hm =86). 303:0“ 8. v 8. v 8.0 8. v 8. v 8.0 8. 3 =86} culpa— mm. mm mm w mm. mm. 8.? m0. =8.oco .82.";— 8. v 8. v 8.0 8. v 8. . 8.0 8. . =86co ouvbn. mud. mm. 3. and T mm. 7 as. J 8.9 5 5 «80“ 8.58 8. 5 0°.an 8.2.0 8.8m 8.31 8.8V E 8.0 B. 8.0 8.0 8.0 8.0 8.0 09.2050 :90: a . 1 . H 8.000 8. 5v 8.00m 00. 5V 8.8M oo. 5? 8.80 8.23 8.8m 8. 51 8.8a 00.an 8.8m ooénv ncoaaZoonO afiooovu vmfinvuw 9.28“” owdnhon 8.03: 50.68? owdusn 8.039 «99.39 8.8mm o fivvwn: nNKnQon 3.8 "NOV cad Eva— NQNwN Rum: 9.52 5.09 NndN— nud— ondn 2.00 vwd: mafia «1.00? 9.9;. nwfpuw «No: 0.0 03023 05 02023 0.0 03023 03% 0.023 05. 02023 030 03023 030 03023 mum: $1.50 >EQ §.20.Q<>>0<0m DmN360 Valid Cum Value Label Value Frequency Percent Percent Percent 1.00 97 11.2 11.2 11.2 2.00 423 48.8 48.8 60.0 3 00 346 40.0 40 0 100 0 Total 866 100.0 100.0 Valid cases 866 Missing cases 0 Figure D.1: Continued ROADWAYzl — - - - - Z N E W A Y - - - - - Variable RALL TOTAL ACCIDENT RATE By Variable M533 medians 0—30; 30 to 60; >60 Analysis of Variance Sum of Mean F F Source D.F. Squares Squares Ratio Prob. Between Groups 2 1345628.?87 672814.3935 4.4160 .0124 Within Groups 863 131483774.7 152356.6335 Total 865 132829403.5 Standard Standard Group Count Mean Deviation Error 95 Pct Cont Int tor Mean Grp 1 97 375.3881 467.7201 47.4898 281.1217 TO 469.6546 Grp 2 423 323.0373 380.8345 18.5168 286.6407 TO 359.4340 Grp 3 346 406.5319 377.9345 20.3179 366.5694 TO 446.4944 Total 866 362.2604 391.8673 13.3162 336.1245 TO 388.3962 GROUP MINIMUM MAXIMUM .0000 2717.8082 .0000 2598.996? .0000 2380.7274 051:6) {16% u N H .0000 2717.8082 u 3 E3 Figure D.l: Continued 197 ROADHAY-l -----oN2wAY----- Variable RALL TOTAL ACCIDENT RATE By Variable M503 medians 0-30; 30 to 60; >60 Multiple Range Tests: LSD test with significance level .05 The difference between two means is significant if MEAN(J)-MEAN(I) >- 276.0042 ' RANGE ' SORT(1/N(I) o l/N(J)h with the following value(s) for RANGE: 2.73 (0) Indicates significant differences which are shown in the lower triangle 'UHO '1 H M H U Mean M803 323.0373 Grp 375.3881 Grp 406.5319 Crp UHN Figure D.1: Continued 198 ROADWAY- l -----ONEHAY----- Variable RALL TOTAL ACCIDENT RATE By Variable M503 medians 0-30; 30 to 60; >60 Multiple Range Tests: Tukey-HSD test with significance level .050 The difference between two means is significant if MEAN(J)-MEAN(I) >- 276.0042 - RANGE - SQRT(1/N(I) + 1/N(J)) with the following value(s) for RANGE: 3.33 (0) Indicates significant differences which are shown in the lower triangle 0 c c I'I'I' P P P 2 1 3 Mean M803 323.0373 Grp 2 375.3991 Grp 1 406.5319 Crp 3 - Figure D.1: Continued 199 ROADHAY-l -----ONEHAY----- Variable RALL TOTAL acetnssr RATE 5y Variable M203 medians 0-30; 30 to 60; >60 Multiple Range Tests: Tukey-B test with significance level .050 The difference between two means is significant if MEAN(J)-MEAN(I) >- 276.0042 ‘ RANGE ' SORT(1/N(I) o 1/N(J)) with the following value(s) for RANGE: Step 2 3 RANGE 3.07 3.33 (') Indicates significant differences which are shown in the lower triangle '0'10 '0'10 '0'10 2 1 3 Mean M203 323.0373 crp 2 375.3881 Grp 1 406.5319 Grp 3 0 Figure D.1: Continued REFERENCES REFERENCES Agent, K. R., and R. C. Deen, 1975, Relationships Between Readway (jegmem'cs and Aggidents, Transportation Research Record, Bulletin 541: 1-11. Agent, K. R., 1983, WWW, Transportation Quarterly Volume 37, New York, New York: 99-114. American Association of State Highway and Transportation Officials, W 11W, (Washington, D.C.: American Association of State Highway Transportation Officials, 1990): 817-836. American Association of State Highway and Transportation Officials, Akfirade Intersections, Ch_ap_ter 1X, (Washington, D.C.: American Association of State Highway Transportation Officials, 1990): 669-670. Author Unknown, 1963, lhe guide for the Determinatign Qf Rgadway Width and Median Ireatment in Llrbg Semi-Emmi Areas, (February). Basha, P. E., and S. L. Ramsey, 1993, WWW, Institute of Transportation Engineers Journal (May): 30-35. Beukers, B6 and J. H. Jenezon, 1972, WW nge lypes of Single-level Intersegtign, Traffic Engineering Division, Ministry of Transport, Netherlands, International Study Week in Traffic Engineering and Safety: 1-4. Billion, C. E. and Parsons, NC, 1953, WI; New York State Department of Public Works, Highway Research Board, Bulletin 308: 64--79. Bochner, B. S., 1978, MW, Institute of Transportation Engineers Journal: 115-121. Box, P. C. et al., _-. l HUI 1.... in! - 161161 - 'r ’oelati._ ' 0 uh: M020, (1970): Chapter 5- Box, P. C., 1969, Wigs, Public Safety Systems, (May): 18-22. 200 201 Box, P. C., 1969, 'v w ide an V l t dies- III ei 11 Public Safety Systems, (September/October): 16-19. Brenner, R., and J. H. Mathewson, 1957, I exe 0t r Vehi 1e Ac i e Highway Research Board Bulletin 161: 1-8. Brodsky, H. and A. S. Hakkert, 1983, Highway Accidehi Rages _a_nd Rahal Iravel Densities, Accident Analysis and Prevention Volume 15, Oxford, England: 73-84. Cirillo, J. A., 1992, W131, Federal Highway Administration No. Federal Highway Administration-RD-9l-O44 (November): 1-9. Cribbins, P. D. ,J. W. Horn, F. V. Beeson, and R. D. Taylor, mwmmiansand Mied 1ah Qpehings eh Divided Highw ways is NQllh Caren 1ha, (North Carolina: North Carolina University: 1966). Cribbins, P. D., J. M. Arey, and J. K. Donaldson, 1967, Effeets ef Seleeted Headway md WWW TranSPortation Research Board Record 188: 8-25. Cribbins, P. D., J. W. Horn, F. V. Beeson, and R. D. Taylor, 1966, Mediahflmhingsgm Ir’vid 1', w. ‘ ei Effc u .110- ,.Lae, - . _ 'v- o__ - i e, Highway Research Record 188: 140-157. Cribbins, P. D. and G. S. Summer, 1964, 1 ti 0 'd R t Design Campenents ef Varieus Types ei Highways, Raleigh, North Carolina. Crosby, J., 1960, 0 - ‘d-il ‘ itte 20‘1' ‘ u -_ ‘ HWJ ‘ , Ju- ,Highway Research Board Bulletin 266: 63--77. DarLO K., andL Mann 1970 W Wane, Highway Research Record 312: 1-16. Garner, G. R., 1970, Aeeidehis at Median Csoss-Overs, Highway Research Record 312: 55-63. Glennon, J. C., Valenta, J. J ., Thorson, B. A. and Azzeh, J. A., Techhicai giaideiines fer Centre! at Direci; Assess, (Washington, D.C.: Federal Highway Administration Publication 1975). Hakkert, A. S. and Ben-Yakov, Y., 1985, W090 Highways, Institute of Transportation Engineers Journal (March): 42-45. Hartman, J. P. and D. B. Szplett, 1989, ' s' t ' : ' ed e i W, Institute of Transportation Engineers Compendium of Technical Papers: 59-62. 202 Higgins, K., M. Martin, M. Parker, M. Van Aerde, S. Van Winkle, M. Walton, T. Wetrnore, G. Wolff, and S. Yagar, ec 've e s fMedian and c l r ' Lefl-Idmihg Vehicles, (Washington,D.C.: Institute of Transportation Engineers 1984). Hoffman, M. R., A. A. Lampela, and A. A. Nasart., 1969, Evalaaiien at a Safety Ezejeei: "W" o_ 'ic'hircin o -O'vr Wit _’.1_ ._ Dr t'o.l -o -O'vr Michigan Department of State Highways TSD-SS-116-69: 1-11. Hutchinson, J. W., T. W. Kennedy, and H. E. Surman, 1966, HMMediah ve 1,Transportation Research Board Record 105: 50- 84. Mak, K. K. and P. H. Wright, 1976, ' l V hi 1 i ' ' , Traffic Engineering Volume 46, (January): 16-25. McCormick, D. P. andE. M. Wilson, 1983, WW. Iwe-Way Lefi- 1am Lanes, Transportation Research Record 923: 86- 90. McCoy, P. T., Malone, M. S., 1989, WWW Readyyus, Transportation Research Record 1239, Washington, DC McDonald, J. W., 1953, lai n B tw en MW, Transportation Research Board Bulletin 74. V7.-17 McGuirk, W. W. and Satterly, G. T., 1976, Evaldan'eh ef Easiers Ihfldeheing Dn’veway Aeeidents, Transportation Research Record 601: 66-72. McMonagle, J. C., 1953, el i n f r 1 i n l ers cti c' , Michigan State Highway Department, Highway Research Board, Bulletin 74 (January): 46-51. Metcalf, G. W., 1989, Rg'sed Medians @019: Flash Medias, Institute of Transportation Engineers Compendium of Technical Papers: 116-118. MDOT, State Highyyay Henge] SeeLieh Atlas, MDOT Report number 42, (Lansing, Michigan 1990). MDOT, WWW (LanSing, Michigan 1990)- Morin, D. A., 1967, ' ' n it' n e t t c' en Data, U. S. Bureau of Public Roads, Highway Research Record 188: 72-79. Moskowitz, K. and Schaefer, W. E., 1958, Qalifemia Median Shady, Highway Research Board Bulletin 266: 36-62. Nemeth, Z. A, 1976, evel ideline f h A licai n nin u wo- w -T edian Lanes, nOhio State University (July): 1-205. 203 Oppe, S., 1991, vel me r 1 an Traffic af ' l rend Ehichiaiiehs, Accident and Prevention Volume 23, Oxford, England: 413-422. Oppe, S., 1991, v 1 1 and raffic afe ' i D vel Accident Analysis and Prevention, Volume 23: 401-412. Radwan, A. E. K. C. Sinha, andH. L. Michael, 1979, WWW , u an or-' v.11. De _- anan 1 1t 1. ve o ' '°ot 0 Min {cad W 1M“ __ lHi_1wa s: 1 tin o.'11_-_-.'n uu , Purdue University, Indiana State Highway Commission (July): 1-29. Recht, J. L., 1965, Maui pie Regsessi ieh Sgdy at the Effects ef Aeiiviiies en nLhe Ira flie Wm, (Chicago, IL: National Safety Council: 1965). Rogness, R- 0. and S. H. Richards, 1981, WWW Iexas, Transportation Research Board Record 819: 17-19. Satterthwaite, S. P., 1981, A Saryey ef Researeh late Relaiiehships Between Iraflle W, Transportation and Road Research Laboratory Supplementary Report 692: 1-41. Savage, W. F., 1974, Directiehal Media Cross-meg, Traffic Engineering (August): 21-23. Schoppert, D. W., 1957, Pr dicti Tr ffi Ac id nt Fr m a w El ent Tw Lan Hi wa Wi r v 1 h ulders, Highway Research Board Bulletin 158: 4-26. Short, M. S., C. J. Chang, and G. A. Woelf, 1982, Effects 91 irafl1e Sighal installaiieh oh Aceident s, Accident Analysis and Prevention Volume 14, Oxford, England: 135-145. Simpson,D.,1988,A vi f e 11 De artment f ran ie B e Highways and Transportation (February): 26-29. Squires, C. A. and P. S. Parsonson, 1989, A ident 1 e w Way Lefl-Tm Laie Median Treatment, Transportation Research Record 1239: 30-40. Taylor, R. J ., 1979, Continaous Lefl-Tu_rh Lanes Reduce Accidents, American Association of State Highway Officials Quarterly, Volume 58, Number 1: 26. Telford, E. T. and R. J. Israel, 1953, Median Study (California), Highway Research Board Proceedings, Volume 32: 208-231. Thakkar, J. S., 1984, Shady of die Effeet ef Twe-Way Lefi- 1m Lanes eh Tiaffic Aeeidents, Transportation Research Record 960: 27—33. 204 Transportation Research Board, 1978, ost an af E tivenes f Hi hw Elements, National Cooperative Highway Research Program Report 197: 9, 88-93. Uckotter, A. D., 1974, Analysis ef Meter Vehicle Aecidents a; Cemmercial Driveway Leeations, Purdue University/Indiana State Highway Commission (June): 1-80. United States Government Printing Office, 1299 Census ef Beeniatien and Hensing (Washington, DC. 1990). Van Winkle, S. N., 1988, Raised Medias vs. Flush Medians, Institute of Transportation Engineers Journal (April): 25-26. Webb, G. M., 1958, Aceidentfimey, Traffic Engineering (October): 18-20, 62. Worsey, G. M., C. E. Dare, R. N. Schwab, and S. C. Tignor, 1986, Investigatien ei Wann'ng Signs fer Median Cress-Overs, Public Roads (March): 113-121. Worsey, G. M., C. E. Dare, and R. N. Schwab, 1987, i l ' er n in ef e e r Medi r s - ver ' n'n , Transportation Research Record 1114: 37-44. ”111111111111111111111115 312 3015703154 9