IHESB [ Unigersity J This is to certify that the dissertation entitled ‘ Developing a, Methodology to Predict High Accident ‘ Locations on Rural Highways in Saudi Arabia ‘ Using Speed Distribution Characteristics 3 presented by Muhammad S. Alisa has been accepted towards fulfillment _ V - 1 of the requirements for ‘ l Ph.D. degree in Civil Engineering N Date August 8, 1984 fl/éém c , %%@ 3 g Major professor ' M5u;.,.,.ur ..- A - 1 m" - , - . 042771 MSU LIBRARIES RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. DEVELOPING A METHODOLOGY TO PREDICT HIGH ACCIDENT LOCATIONS ON RURAL HIGHWAYS IN SAUDI ARABIA USING SPEED DISTRIBUTION CHARACTERISTICS By Muhammad S. Al—isa A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Civil and Sanitary Engineering 1984 i37€;‘/697Q¥ \ (3%} CIN?YRIGHT BY AlrISA, MUHAMMAD S. 1984 ABSTRACT DEVELOPING A METHODOLOGY TO PREDICT HIGH ACCIDENT LOCATIONS ON RURAL HIGHWAYS IN SAUDI ARABIA USING SPEED DISTRIBUTION CHARACTERISTICS By Muhammad S. Al—isa Traffic accidents and highway fatalities are among the major problems confronting Saudi Arabia. The design and implementation of highway—safety programs and practices in the country are in their infant stages. The experience and practices of developed countries, such as the U.S., in the field of highway safety provide a potentially trans— ferable highway safety technology to Saudi Arabia. Highway safety improvement programs, which proved to significantly reduce the number of accidents on U.S. highways, could be eXpected to accomplish similar results if implemented in Saudi Arabia. The absence of an adequate accident recording system in Saudi Arabia, which is an essential tool in highway Safety programs, necessitates the use of non—accident meas— ures in the identification of hazardous highway locations. This research has studied speed distribution characteristics, among other traffic performance measures, as a pOSSIble Muhammad S. Al—isa surrogate measure for accident potential. Variations in these characteristics observed on Saudi Arabian rural highways were compared with variations in these same parameters reported for locations with known levels of risk on U.S. rural highways. The statistical tests and compar— isons verified the suitability of using the speed distribu— tion skewness index as a surrogate measure for hazardous locations on two—lane rural highways in Saudi Arabia. ACKNOWLEDGMENTS The author would like to express his sincere appre— ciation to those who made the completion of this research possible. Special appreciation is extended to my disserta— tion advisor, Dr. William Taylor, for suggesting this topic and for his continuous advice and encouragement throughout the work. Thanks and gratitude are also due to other members of my guidance committee: Drs. Thomas Maleck, K. Rajendra, Keith Honey, and John B. Kreer. Also, thanks are due to King Faisal University for supporting this research. ii TABLE OF CONTENTS CHAPTER PAGE LIST OF TABLES . . . LIST OF FIGURES . . . . . . . . vii MAP . . . . . . . . . . . viii 1 INTRODUCTION AND BACKGROUND . . . . . 1 Highway Safety Practices in Saudi Arabia . 3 Highway Safety Improvement Program . . 7 Transferability of Safety Technology . . 11 The Use of Non-Accident Measures 0 O O O O 16 Objectives of this Research . . . . . 19 Research Problem . . 2 LITERATURE REVIEW AND CONCEPTUALIZATION Conceptualization . . . . . . . 22 Approach Framework . . . Determining suspected locations Spot—speed studies Statistical analysis Classification of safety level of the section Countermeasures Evaluation Application to the Case of Saudi Arabia . . 42 3 EXPERIMENTAL DESIGN . . . . . . . 47 Constraints and Limitations of the Experiment 48 Assumptions and Conditions . . . . . 49 Site Selection and Description . . . . 51 Deciding on the General Study Area . . 52 Choosing the Two-Lane Network . . . . 52 Selecting Study Locations . . Potential Hazard of Curves . . Driver Behavior on Curves . . . . . 55 iii CHAPTER Sample of Potentially Hazardous Locations . Site Description . . . . . . . Configuration of Monitoring Site . . . Spot Speed Studies . . . . . . . Conduct of the Studies . . . . . . Sample Requirements . . . . . . . Data Collection . . . . . . . .. Statistical Analysis . . . . . . 4 DATA ANALYSIS . . . . . . . . . Statistical Results . . . . . . . Normality Tests . . . . . . . . Speed and Speed Distribution Characteristics Normality of Spot Speed Distributions on Tangent Sections . . . . . . . Speed Behavior When Negotiating Curves . . Normality of the Speed Distribution on Curves Comparison with Other Tests of Normality . Trends and Variations Associated with the Use of Skewness Index . . . . . 5 CONCLUSIONS AND RECOMMENDATIONS . . . . Specific Findings . . . . . . . Transferability of Skewness Index . . . Recommendations . . . . . . . BIBLIOGRAPHY . . . . . . . . . APPENDIX A O O O O O O C O O O O B . O O O O O O O O O C 0 C ' 0 o o o o o o o o o o D O O 0 0 O O O O O O O E iv PAGE 56 57 59 63 63 67 71 71 73 74 74 75H 84 90 109 130 130 138 140 141 143 145 159 160 162 164 168 TABLE 10 11 12 13 LIST OF TABLES Number of Traffic Accidents, Injured, and Fatalities in Saudi Arabia at the End of Each Year from 1972 to 1982 . . . . . . Traffic Accidents in Saudi Arabia During 1982 by Region and Cause . . . . . . Summary of Geometric Feature Ratings for Average Conditions on Various Classes of Rural Non—Freeway Highways . . . . . . Roadway Geometric Expectancy Failure Mode Analysis Based on Literature . . . . Drivers Involved in Accidents, by Nationality Standard Deviations of Spot Speeds for Sample Size Determination . . . . . . Spot Speed Distribution Characteristics for Tangent Sections with Minimum Marginal Friction Speed Data for Tangent Sections on Two—Lane Rural Highways . . . . . . . . Percentage of Vehicles (Passenger Cars and Light Vehicles) Exceeding 50 mph and 60 mph Vehicle Age Distribution . . . . . Observed Speed Ranges on Tangent Sections on Two-Lane Roads in Saudi Arabia . . . Index of Driving Licenses Issued in Saudi Arabia During the Years 1972—1982 According to Type . . . . . . . . . Results of the Modified Spot Speed Distribution Characteristics for Tangent Sections . . PAGE 32 39 45 69 77 78 8O 82 83 85' 87 TABLE 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Spot Speed Distribution Characteristics for Tangent Sections with Marginal Friction . . Spot Speed Distribution Characteristics on Curves and Their Approaches . . . . . Comparison of Mean Speeds (mph) for the Curve and Its Approach . . . . . . Comparison of Standard Deviations for Curves and Their Approaches . . . . . . . Spot Speed Distribution Characteristics for Curves . . . . . . . . . . Empirical Speed-Curvature Relationships . . Workload Potential Ratings (RC) of Horizontal Curves . . . . . . . . Spot Speed Distribution Characteristics for Flat Curves . . . . . . . Modified Spot Speed Distribution Characteristics for Flat Curves Below 2° . . . . . . Spot Speed Distribution Characteristics for Sharp Curves . . . . . . . . Spot Speed Distribution Characteristics for Sharp Curves that are Rated Normal . . . . Speed Distribution Characteristics for Reverse Curves . . . . . . . . . Summary of Results . . . . . . . . Normality Test Results Using Kurtosis . . . Empirical Skewness Index-Curvature Relationships Variation in Mean SpeedvdifliDirectioncflETraffic Comparison of Skewness Index for Different Traffic Direction . . . . . vi PAGE 29 91 93 94 96 100 112 113 116 117 120. 124 127 131 132 136 137 FIGURE 1 10 11 12 13 14 15 LIST OF FIGURES PAGE Fatalities in Various Developing Countries, 1978 O O C O O C O D O O O 4 Highway Safety Improvement Program at the Process Level . . . . . . . . . 9 Highway Safety Improvement Program Subprocess . 10 Accident Involvement Rate by Deviation from Average Traffic Speed . . . . . . . 24 Driver Expectancy Problems Rating Form . . ., 31 Measurement of the Angle by Placing a Theodolite on Point PI . . . . . . . 58 Configuration of Measurement Apparatus for Horizontal Curve . . . . . . . . 60 Configuration of Measurement Apparatus for Winding Situation . . . . . . . . 61 Configurations of Measurement Apparatus for Tangent Situation . . . . . . . . 62 Roadside Development Conditions . . . . 76 Empirical Speed—Curvature Radius Relationship Obtained for the Data Collected in Saudi Arabia 101 Empirical Speed-Curvature Relationship Obtained for the Taragin (1954) Data . . . . . 103 Empirical Speed—Degree of Curvature Relationship Obtained for the Data Collected in Saudi Arabia 104 Empirical Speed—Curvature Degree Relationship . 105 Empirical Standard Deviation—Curvature Radius Relationship Obtained for the Data Collected in Saudi Arabia . . . . . . . . 107 vii FIGURE 16 17 18 19 20 21 22 23 24 MAP PAGE Empirical Standard Deviation-Degree of Curvature Relationship Obtained for the Data Collected in Saudi Arabia . . . . . . 108 Relative Number of Road Accidents Versus Radius of Horizontal CUrves in Meters . . . 110 Roadside Development Condition . . . . . 115 Sight Distance Situation at Curve Locations . 122 Delineation Treatment of Curve Sections . . 123 Positive Guidance Measure on Curve Location . 125 Recorded Number of Accidents/Number of Ramps . 129 Empirical Skewness Index—Curvature Radius Relationship Obtained from the Data Collected in This Research . . . . . . . . 133 Empirical Skewness Index—Curvature Degree Relationship Obtained from the Data Collected in This Research . . . . . . . . 134 MAP PAGE 0 0 0 O O 0 C O O O O O 53 viii CHAPTER 1 INTRODUCTION AND BACKGROUND In contrast to many developing countries, Saudi Arabia is a rich country. The sudden rise in the national income is producing challenges which are different from those of other developing countries. Saudi Arabia is changing rapidly from a pre-industrial country to a modern industrialized society. This rapid development exerts pressure on all public utilities and facilities, including the transportation system. The vehicular population reached 2,467,903 by the end of 1981. This is an increase of 1705 percent within the eleven years from 1970 to 1981 (Traffic Statistics, 1981). This expansion in vehicular population was paralleled by the construction of 20,238 km of main roads and 24,186 km of agricultural roads (116), since most of the initial main road networks consisted of single, undivided two-lane roads. The development and construction of the road network and the significant increase in automobile ownership are accompanied by a high incidence of traffic accidents. As shown in Table 1, in 1982 traffic accidents injured 15,872 TABLE 1 NUMBER OF TRAFFIC ACCIDENTS, INJURED AND FATALITIES IN SAUDI ARABIA AT THE END OF EACH YEAR FROM 1972 TO 1982 Number of Number of Number of Year Vehicle Accidents People Injured Deaths 1972 4,147 4,583 570 1973 7,197 6,530 834 1974 9,808 7,901 1,058 1975 10,897 8,771 1,154 1976 13,475 10,532 1,594 1977 15,709 11,606 1,975 1978 15,785 11,413 2,033 1979 18,051 14,824 2,378 1980 17,743 16,832 2,871 1981 18,748 16,218 2,731 1982 17,897 15,872 2,427 SOURCE: Traffic Statistics, General Department of Traffic, Ministry of Interior—Public Security, Kingdom of Saudi Arabia. 3 persons and killed 2,427 people. This is a relatively high death rate compared to developed countries, such as the U.K. and the U.S.A. (see Fig. 1) (51). As in any other developing country, traffic accidents are among the leading causes of death and crippling injuries. Highway Safety Practices in Saudi Arabia Safety experts agree only on the complexity of the problem of transportation safety. The current strategy of transportation safety has been classified into two basic approaches: 1. The preventive engineering approach, where accidents are reduced or minimized by two general methods: (a) countermeasuring highway malfunctions or failure; and (b) considering safe design measures for new highways (127). 2. The human engineering approach, which seeks to discover the limitations of man performing within the complexity of the automotive transportation system. The results of this type of research are then used to redesign the system so that the system fits the needs and capabilities of the road users (106, 114, 115). In Saudi Arabia, the highway-safety authorities still hold the traditional "violation-error" attitude toward accidents. This is reflected in the traffic statistics book, where accidents are classified by cause rather than by location (see Table 2). Therefore, safety programs are Deaths/10,000 vehicles :: q a a E i? E i 5 H r l I I f -" 1‘2! u 1‘. ‘I Q C "'I" u r I r r g 77/ . ?:¢l_“., 4 4 e’ ‘l \ O I 9 if ‘I’ /-/ O ’1 I / 4‘ a»??? ’/ / Country 4“ ‘, «f. waxy. x7 any ‘0 «sir-0 s! ./l x/l‘f. J‘ ‘4 / Fig. 1. Fatalities in various developing countries, 1978. SOURCE: G. D. Jacob and I. Sayer, "Road Accidents in lmveloping Countries," Accident Analysis and Prevention 15(5) (October 1983):337—69. N mumawrfi. .qmm .a .summmm uflaasmnnoenmucH co stomach: .xoom mueumenmnm assumes .momaom . . , an 8. an: 23 8:: . Lalo . x: 83 Sun 25» find I a . av 3 a: I «N a: (Each: Its...” _ a.» 9 8. as can 2 2. «8 25:. El»... . s an 3 8 2: a a 2... “50...? wk... 2 8 3 3 mu. 2 9 5.. 24:22 \ik I a a nu um I ON Br ¢<¢< ea. 0 m 3 a an 8 p 3 .2 2:230 .“J n «p 2 2. 3» o I «8 «:3 Al. I 2 a 8 8 p 2 2.. sh: wk: 3 3 9 2 So w 8p 3.. c503» :1. . . u a on 3. an. 3 z: .33 .2395 an? 2 3 S 3: So on o: 8: 2553 .31. 8 8 an new 2: a .3 88 “:3 3.1. o 3 8 3 2v 2 a: 82 «2.8.2 OJ 3 8 2: 8 on. 8 8n an: 38.2 in 3 a: S .3 22 2 8.. 83 553.. a}... «2 a: Sc 5.. 38 3 8o 88. :92: 22:: :04 Swan 9:3 mo .55 .850 .256 26.255 :9: 3:82 5:5 :29 .. b . . ..., brags. r a... .4 Iva I... mzoémz 3.... A. a. a. a... we a 5.3. a... .33 I» s. .. . w...“ 2&5: has) .3... 32.. 3.. ._........ aka . Us. 1) . fiJa mmD: IMPEMESTATION SCHEDULE AND IMPLEMENT I4. COMPONE T SAFETY IMPROVEMENT PROJECTS I r——-—------—-----—‘-_——-——-—-' I PROCESS 1 l EVALUATION DETERMINE THE EFFECT HI COMPONENT OP HIGHWAY SAFETY I IMPROVEMENTS I I I I.....____....__..___.._..___.._-.._..I Fig. 2. Highway safety improvement program at the Process level. . SOURCE: Goodell—Grivas, Inc., HighwaySafetyEngineer— i£g_§tudies: Procedural Guide (Washington, D.C.: U.S. Department of Transportation, Federal Highway Administration, June 1981), p. 3, 10 PROCESS 1. caIECT Am MAINTAIN om DPIKISI l N”! M OGGNIAT 11H”! IlFll-UL'I “ml 35 1 I muons: J I ”III“ I ucxnun um lzhrii‘z “if". “:5“: film" PL‘NMNG‘ CWNT -'I i V I PROCESS 2. IDENTIFY HAZARDOUS LocATTorS Aw ELEMENTS I PROCESS a. CONDUCT ENGINEERIPG STUDIES WEI-([55 l ”NIH! 3 DUI“! "MUS i I PROCESS 4. ESTABLISI PROJECT PRToRTTTEs I _ _ _ _ ._ ._ _ _ _ _ —1 L j Y PROCESS 1.‘ SCHEDULE AND IHPLE NT SAFETV IMPROVEMENT PROJECTS MIPLE‘ENTAT'ION “mwmwr “SCENE WEH’ PROCESS 1. OETERMTIE m: EFFECT or HIGHHAY SAFETY IMP "3:933 ‘ “NINE“ affaifiéfé; warms: "ERNEST... "W” I EWLUATIGW CWPOVENT will“ 4 "NW DUKISTIA'H'I (MINING. Fig. 3. Highway safety improvement program subprocess. SOURCE: Goodell—Grivas, Inc., Highway Safety Engin— EEEleiStudies: Procedural Guide (Washington, D.C.' U.S. Department of Transportation, Federal Highway Administra- tion, June 1981), p. 3- 11 3. the absence of highway records to identify the location of geometric features for use in coding accident locations. In addition to these deficiencies in accident and highway records, the use of accident records in the development of a highway safety program in Saudi Arabia may be constrained by the following factors: 1. the general official tendency not to release information; 2. the refusal by the people to allow public officials to keep their accident records or any records of their bad conduct; 3. the absence of manpower skilled in accident investigations; and 4. the rapid change in the highway system, which complicates the task of maintaining highway records. The above discussion points out some constraints that must be overcome in establishing a highway safety program. In addition, a better understanding of the transferability of safety technology developed in the industrial world to developing countries such as Saudi Arabia is essential. Transferability of Safety Technology The scientific and quantitative methods of highway safety studies imply the following reasoning: If the accident history of a site is found to deviate from the norm for its class, there surely is some reason for it. If so, a 12 responsible agency and its professionals should examine the cause for this deviation and if a cost—effective remedy can be found should remove the cause of substandard performance. (44) Application of this reasoning to developing countries may not be possible because of differences between selected locations in (41, 51): l. the road users involved; 2. data collection and analysis in various regions; 3. driver behavior; 4. road-user knowledge; 5. traffic—law enforcement practices; 6. vehicle safety measures; and 7. highway engineering practices. Since the nature of the road safety problem is different in developing countries, Haight suggests that "the lessons to be learned should not be regarded as 'technology transfer.' In fact, many of the 'truths' about traffic safety in developed countries may be untrue in developing countries." For example, Robinson suggests that the application of geometric standards developed for highly motorized countries to low—volume roads in developing countries leads to designs which are uneconomical and technically inappropriate. He also points out that "more research is needed to determine whether geometric design standards have a different influence on road safety in developing countries from that observed in industrial countries" (102). Jacobs and Sayer conclude that "although — fi 13 research findings from developed countries can provide some guidance, the inevitable uncertainties surrounding their transfer to developing countries emphasize the need for caution in their application" (51). These concerns apply to the situation in Saudi Arabia as well as to any developing country. But while other developing countries suffer from a lack of financial ability to spend on road-safety improvements, Saudi Arabia has the financial ability to adopt the latest tchnology. This is evident from the road network currently under development, where high standards are incorporated in the highways now being built to connect major cities. However, there are disadvantages to this approach: 1. The geometric standards adopted may have a different influence on road safety in Saudi Arabia from that observed in industrial countries. 2. The low-volume roads and highways connecting non-major cities will not receive sufficient safety consideration, as they are not likely to be upgraded to freeway standards. 3. Accidents which do occur on highways with these high standards will strengthen the traditional attitude that all accidents are due to driver failure. 4. Safety concerns become only part of the modernization projects, rather than being the primary objective. 14 5. Highway safety may not be institutionalized properly in the official government bodies, but considered only as an offset of the design function. 6. The local agencies will miss the learning experience which results only by 1233; in-depth studies of road layout, vehicle design and road-user behavior. Therefore, the creation of a highway-safety program in Saudi Arabia requires research to overcome these constraints and verify the suitability of alternative methods and tools. For example, the deficiency of accident records, which is the main tool in highway safety programs, may suggest the use of non—accident measures. The Use of Non—Accident Measures Accident history is widely accepted as a primary determinant of hazardous locations. Several methods (144) are now used for identifying hazardous highway spots based on accident histories: 1. number-of-accidents method; 2. accident-rate method; 3. number of equivalent property damage only accidents (EPDO method); 4. equivalent prOperty damage only accident rate (EPDO rate method); 5. rate-quality control method; and 6. various combinations of one or more of the above methods. 15 Ordinarily, the accident histories of highway locations for certain periods of time (usually three years) are used to screen high—hazard spots on the highway system and to devise cost-effective remedial safety projects. the traditional measurements of the various 70, 92, 96, 117, 122, However, methods have been criticized (44, 128) for the following shortcomings: 1. No agreement has been reached on measures of accident exposure. 2. There are imperfections in rcording accidents. 3. The time period and number of accidents required to obtain statistically significant results are undesirably large. 4. Past accident experience is invalidated with any major changes in the transportation system. 5. The accident histories of locations are subject to random fluctuations, which brings into question the assumption that the number of accidents on a system in the period before treatment is an unbiased estimate of what should be expected to occur on the system during an equivalent "after" period had treatment not been applied. These shortcomings have brought criticism to safety programs in developed countries (41). Haight stated: In developed countries the choice of countermeasures, the design of programs and projects, and administration and final evaluation of these programs are often thought to be on reliable statistical evidence. Unfortunately, the reliability of the evidence is often exaggerated, sometimes grossly (41). 16 Some researchers, such as Whitelegg (138), propose that researchers and practitioners put road traffic accidents into a wider context of societal development, and use long—term policy objectives as opposed to emphasizing traffic accident performance measures. Other researchers (92, 123) suggest the use of non—accident measures to circumvent the aforementioned shortcomings of accident measures. This research addresses the possibility of using non—accident measures for ranking hazardous locations and the evaluation of remedial safety projects in Saudi Arabia. A specific research program is proposed to explore the potential of using available non—accident measures. Research Problem The determination of reliable non-accident indicators of traffic safety on a given highway section is quite problematic, to date. Among the non—accident measures suggested are various traffic-performance measures. A traffic—performance measure is defined as any measurable parameter that describes the flow of traffic at a certain point or over a particular section of highway. This category contains all measures that are based on quantifying other traffic characteristics, such as quality of flow. The value of any of these measures depends on the additional knowledge of its quantitative correlation to one of the direct accident measures. Without this correlation, 17 it cannot be considered a measure of accident potential (42). Some of the literature (33, 34, 56, 75, 95, 113, 119, 123, 125) reported the following traffic performance measures as potential surrogate safety indicators: —speed —speed variance —speed distribution skewness —acceleration noise —headway distribution —traffic conflicts —erratic maneuvers —lateral placement -brake applications The basic advantage of these indirect measures over the direct ones is the frequency with which these indirect measures occur, which means a statistically reliable sample can be obtained in a relatively short time interval. But, since conditions prevailing during the short period of measurement are not necessarily representative of the variable conditions over a longer time period, there is no assurance that the measured value will be representative for the longer time period. Therefore, after each design Change in existing conditiOns or for a new highway site, the characteristics of the traffic performance measure have to be determined to check the safety level and evaluate any improvements. Among the above transportation performance measures, traffic speed has received the most attention in safety research. It is easily observed and measured with little expertise, and spot—speed distribution characteristics have been found to correlate with hazardous locations. In Saudi Arabia, speed is the only recognized traffic performance measure identified as a factor in traffic safety (see Table 2). Although speed characteristics are not as stringent a screening device as traffic accidents, the financial ability of Saudi Arabia allows the use of less tight screens than traffic accidents in the safety—improvement programs. As suggested before, the difference in traffic conditions in Saudi Arabia may alter the nature of safeyt problems from that in developed countries. This implies possible differences in traffic—performance measures, especially the ones pertaining to safety. For example, the differences in traffic enforcement, driver behavior and drivers' knowledge may affect the traffic speed behavior, and thus the traffic speed characteristics. In this research, a methodology is developed for identifying high accident risk locations on rural Saudi highways. A model is developed based on measured characteristics of traffic speeds and statistical parameters derived from these measurements. Variations in these characteristics observed on Saudi Arabian rural 19 highways are compared with variations in these same parameters reported for locations with known levels of risk on U.S. rural highways. A statistical comparison is then made to determine the potential of using these parameters to predict high accident locations in Saudi Arabia. ijgctives of this Research In an effort to develop a surrogate safety measure and demonstrate its potential in the identification of high-hazard locations, this research will consider the following objectives: 1. A review of the literature, to demonstrate the relationship between traffic speed characteristics and safety in the United States. This will include empirical evidence as well as theoretical verification. 2. Development of a general approach that utilizes spot speed characteristics and serves as a guideline for highway safety improvement programs. 3. A study of the suitability of this approach for application in Saudi Arabia. 4. The development of an experimental design for field measurements of traffic speed characteristics on Saudi Arabian highways. 5. An analysis of similarities and differences between the U.S. and Saudi Arabia in traffic speed patterns. 20 6. A discussion of the potential application of speed characteristics as a surrogate measure for highway safety in Saudi Arabia. CHAPTER 2 LITERATURE REVIEW AND CONCEPTUALIZATION The research on traffic flow indices for the detection of high accident potential highway sections and roadway locations has received extensive attention from professionals and academicians. Speed-distribution Characteristics are among the driving-performance measures which show the influence of both driver and road conditions (1, 14, 46, 48, 60, 70, 75, 80, 92, 115, 117, 119, 121, 122, 123, 125, 131, 141). The speed distribution skewness measure is of special importance because of its sensitivity and potential to be a surrogate indicator of hazardous locations. This chapter will review the literature, trace the theoretical evidence, and integrate the available facts to build a rational formulation of the suggested relationship between the skewness index and hazard in roadway locations. Finally, it will present an approach for a highway safety program, and discuss its application in Saudi Arabia. 21 22 Conceptualization In the United States, there has been a tendency to overestimate the extent to which speed contributes to traffic accidents. Steward stated that "considerable emphasis is directed today toward fast driving and speeding as a major cause of automobile accidents." He explained that the reasons behind this tendency are: a) a knowledge of certain statistics which seems to indicate that speed violations are frequently involved in automobile accidents; b) a desire by safety officials and other interested individuals to find a satisfactory and feasible solution for the nationwide problem of accidents. (118) In the search for a more definitive relationship, several researchers (32, 39, 52, 53, 97, 115) have conducted studies regarding the relation of different characteristics of speed to safety. The findings reported by Solomon indicate the following: 1. The accident involvement, injury, and property damage rates were highest at very low speeds, lowest at about the average speed of all traffic, and increased at very high speeds, particularly at night. Thus, the greater the variation in speed of any vehicle from the average speed of all traffic, the greater the chance of being involved. 2. The severity of accidents increased as speed increased, especially at speeds exceeding 60 mph. / 3. The fatality rate was highest at very high speeds and lowest at about the average speed. (115) Based on the findings of several studies, Warren (134) develOped a graphical relationship between accident invo (see dayt U-sh absoi strez conce speed speed achie 0f in concl have accid that 10 Co actio the o Proba' to m lower the II 11., w 23 involvement rate and deviation from average traffic speed (see Fig. 4). This figure shows the relationship for both daytime and nighttime accident rates in the form of a U—shaped curve. At that time, accidents appeared to depend less on absolute and more on the variation of speeds in the traffic stream. However, the cited research did not provide conceptual or theoretical explanations underlying the speed—safety relationships. Marsh and Carson (64) concluded that motor vehicle speed control is an extremely important element in achieving safe and efficient traffic movement. In a review of international speed regulation experiences, Smeed concluded "that the imposition of speed limits seemed to have had a favorable effect" (113). Wilson (139) also wrote that speed zoning reduces accidents, but he attributed accident reduction to the fact that drivers would be more alert and therefore better able to control their cars to avoid the necessity of emergency actions. Haur (43) explained the role of speed regulation in the context that on rural roads between intersections, the probability of an accident involvement was closely related to the rate at which overtaking took place. When upper and lower speed limits were imposed, slow drivers traveled at the lower speed limit and fast drivers at the upper speed limit. Consequently, the uniformity of traffic speed would NIP-.11 (rlwll: TIN-Pu: Fi averagegtr: 24 100.000 50.000 1 10.000 x} / 5000 — -. ‘ - \ 2 I 4-Lane Main Rural Highways Involvement Rate per 100 Million Vehicle-Miles 1000‘ / 500 / . '7 o I 100 50'- ‘ FreewayS’/’ . o 10 I l I _L L I J ~40 -30 -20 ~10 0 10 20 30 40 Deviation from Average Speed, mph Fig. 4. Accident involvement rate by deviation from average traffic speed. limi repo redu of h them acci traf acci cont spee 125. ICC The Con the and and 25 limit the desire of a driver to overtake. Therefore, as reported by Lam, the greater uniformity of speeds leads to reduced accidents. (58) The research above presented some logical explanations of how speed zoning reduced accidents, with the general- theme that traffic—speed variation was correlated with accidents. However, these studies did not explain how traffic—speed variation took place, nor how it caused accidents. One of the significant works in speed zoning that contributed to the explanation and verification of a speed-safety relationship was undertaken by Taylor (124, 125, 126). He developed a new theory using the concept that a relationship exists between the rate of accident occurrence and the distribution of speed on rural highways. By studying many Ohio speed zone sites (124), he found that just lowering the higher speeds and raising the lower ones is not sufficient to cause accident reductions, and that the normality of the speed distribution must be taken into account. His theory is based on the fact that in situations where all drivers are able to determine and evaluate the conditions that exist at that time and at that location, the resulting speed distribution is normal with no skewness and with normal kurtosis. The normality of the distribution is explained by the fact that the driver is a human being and he has to evaluate nontangible associated costs with travel variat sectio inabil Accord parame conclu work 0: 0f haz exPlan aPProa 108, 1| exPlan: 0f the and th. 10st 11 drivin strong This it 26 traveling, such as comfort, convenience, or service. The variation in the distribution which occurs on certain sections of highway is attributed to various drivers' inability to properly evaluate the same situation. Accordingly, Taylor proposed utilizing speed—distribution parameters to determine unsafe highway sections. The conclusion drawn from this research is as follows: —There is a strong relationship between the ‘ rate of occurrence of accidents and speed . distribution on rural state highways. -Changing the speed distribution from non—normal to normal results in an accident rate reduction which is twice that found under any other set of "before and after" conditions. —The best parameter to use in determining non-normality of speed distribution is the skewness of the distribution. While this research sets the stage for further work on traffic—speed characteristics as surrogate measures of hazardous locations, it does not offer theoretical explanations. The literature on human engineering approaches (7, 19, 24, 27, 28, 29, 30, 36, 57, 63, 73, 100, 108, 109, 132) provides deeper and more comprehensive explanations. This approach looks upon the three basic components of the transportation system —— the driver, the vehicle, and the environment —- and considers the driver to be the most important part of the highway—traffic system. In the driving situation, the three components are coupled strongly through informational and mechanical links (57). This informational and mechanical flow between the driver, the ve attrib b I may be the at inters enviro indivi percep drivin 132). D‘mmr—hmP—a demand that: resPOn making c“Dre inalit 27 the vehicle, and the environment requires the following attributes of the driving tasks (90). 1. Perception (sensing the information) 2. Comprehension (understanding or recognizing the information) 3. Decision (making a decision based upon the information) 4. Action (performing some physical action based on the decision) In a traffic situation, motor—vehicle accidents may be attributed to a variety of failure modes affecting the aforementioned informational and mechanical intersections linking the driver, the vehicle, and the environment. The three components act collectively or individually to degrade the speed and accuracy of perception, comprehension, decision, and action by placing driving demands beyond the limit the driver can handle (28, 132). Wallen stated that: The human error identified in accident causation studies is frequently related to information failure (including recognition error) strongly suggest[ing] that the demands of the driving situation may be more than the driver can handle. (132) The direct effect of increasing the driving demand is to increase the reaction time. Forbes stated that: "As the complexity of the task increases the required response time also increases. In addition, the chance of making an error increases." (28) Thus, we see that the time required for perception, comprehension, decision, and action is a very critical quality in highway driving and highway accidents. Forbes expi hov cri T01 rel 1101' 28 explained that in high driving and high accidents it is not how fast the driver is, but how the driver stops that is critical. In addition, he pointed out that the increased volume and pace of traffic and the more efficient and reliable performance of vehicles make driver reaction time more important and more critical (28). Therefore, it is possible to analyze the various components of the driving task and to measure experimentally the effect of different factors (environment, drivers, and vehicle) on the human reaction. In the case of the roadway environment, the driving demand and thus the reaction time increase with increasing geometric complexity of those highway features perceived as potentially hazardous situations in the driving environment. Versace, recognizing the complex factors causing accidents, used factor—analysis technique in studying roadway and accident data (130). He wrote: Only a single factor emerged from a vast amount of data in this analysis which explained where accidents occurred. Although only highway variables were included in the analysis this one factor conveys the psychological. There are more accidents at those places where the situation places greater demands on the momentary perception—decision motor capacities of the driver. For further conceptualization, Messer, Mounce, and Brackett, in a recent study, offered a thorough explanation of how a highway geometric failure may influence an accident event. They attribute a failure sit inc 29 situation to the existence of geometeric design inconsistency, which is defined as a change of such magnitude and unexpected nature in the physical dimensions and appearance in the roadway geometry between adjacent sections that unfamiliar or moderately inattentive drivers may be surprised, confused, and misled into making potentially unsafe driving decisions. 6) The likelihood of inconsistency occurring in relation to the driver's expectancy is explained by the following assumptions: -Drivers require accurate information about the road ahead to safely control their vehicles. —Drivers generally "see what they expect to see" and "expect to see" what they have been seeing along their recent driving experiences. In a study of decision making in a hazardous activity, Cownip found that: The subjects tend to adjust their behavior so that their level of risk taking is more or less independent of the severity of hazards they encounter, but their decisions are influenced by their experience of the results of previous decisions, since no other information of the stochastic properties of hazards is accessible to them. (19) Therefore, when the road ahead disagrees with what is expected, the driver experiences some level of surprise, conflict, and associated uncertainty. The greater the uncertainty, the greater the information required to resolve the uncertainty and the longer the response to vh llll 30 whatever situation lies ahead of the driver. Thus, the unexpected geometric design feature may lead to any or all of the following driving-control responses: 1. a design—correct response 2. an incorrect response 3. no response The last two of these response outcomes would increase the probability of an unsafe driving situation. Messer and his colleagues Mounce, Brackett, and Wenton (76) operationalized the previous notion and introduced the term of driver workload, which is the time rate at which drivers must perform a given amount of work or driving tasks, based on driver~expectancy considerations and using a subjective rating scale developed for identification of hazardous locations (see Fig. 5). A group of 21 experienced highway—design engineers and research engineers rated the workload associated with several geometric features as shown in Table 3. The above discussion explains how road geometry may influence the occurrence of accidents. Kontaratos (57) pointed out that their effects come primarily through the sensory pathway reaching the driver, such as visual, auditory, and tactile perceptions. However, the way the driver sees and analyzes the situation ahead of him also can determine the likelihood of an accident occurring. The process of speed perception is defined by Krzeminski as "the awareness of elements of the environment through 31 Ratings: 0 -- DRIVER EXPECTANCY PROBLEMS RATING FORM Nothing unexpected or unusual at this location. Actions required (if any) entirely consistent with criving strategy on approach. Standard geometry. with pathway(s) for intended movement(s) clearly evident. No interferences by other traffic likely. Situation somewhat unexpected. Driver must be alert, but should be able to respond adequately at "last minute" to most combinations of adverse circumstances. Some initial confusion on intended path(s) or movement(s). Interference from other traffic may create some degree of confu- sion or uncertainty for average driver. Very unusual situation; will "surprise“ many unfamiliar drivers. Driver required to make major change in driving tactics from those employed over past few miles. At least a "near accident" almost expgcted if driver is even mod- erately inattentive; evasive actions ikely to be required. Intended pathway(s) confusing under fairly normal traffic or lighting conditions. Other traffic, or lack of it. aggravates situation and misleads driver or deprives him of important cues. AEEI‘OBCh Rating 0 v- N -w p U" 0‘ Fi.g. 5. Driver expectancy problems rating form. SUMl tre Ian lan ime 32 TABLE 3 SUMMARY OF GEOMETRIC FEATURE RATINGS FOR AVERAGE CONDITIONS ON VARIOUS CLASSES OF RURAL NON—FREEWAY HIGHWAYS Geometric 2~lane 4-lane Feature High Mediocre Divided Undivided 0 Bridge Narrow Width. No Shoulder 5.4 5.4 5.4 5.4 Full Hidth. No ShOulder 2.5 2.5 2.5 2.5 Full Hidth, Hith Shoulders‘ 1.0 1.0 1.0 1.0 o Divided Highway Transition 4-lane to 2-lane ___ __’ 4.0 ___ 4-lane to 4-lane ___ ___ 1.8 ___ 0 Lane Drop (4-2 lanes) ___ __ ___ 3.9 o Intersection Unchannelized 3.7 2.8 2.4 2.1 Channelized 3.3 2.5 . 2.1 2.4 0 Railroad Grade Crossing 3.7 3.7 3.7 ‘ 3.7 o Shoulder Width Change Full Drop 3.2 2.4 2.1 2.1 Shoulder Hidth Reduction 1.6 1.2 1,0 1.0 0 Alignment Reverse Horizontal Curve 3.1 2.3 2.0 2.0 Horizontal Curve .3 1.7 1.5 1.5 Crest Vertical Curve 1.9 1.4 1.2 1.2 0 Lane Width Reduction 3.1 2.3 2.0 2.0 0 Cross Road Overpass 1.3 1.0 0.8 0.8 0 Level Tangent Section* 0.0 0.0 ‘ 0'0, 0.0 *Assumed NOTE: Ratings of two—lane mediocre road (i.e., surface treatment pavement without paved shoulders) and all four— lane highways usually assumed to equal 0.75 and 0.65 of two— 1ane high—type highway ratings based on results of Exper— iment II described in Chapter VI of Volume II. Value system of ratings described in Fig. 2. r——————--—----------..lli 33 physical sensation" (56). If the driver analyzes the situation properly and makes necessary adjustments in direction and speed, he will properly avoid accidents; if not, he may be "overcommitted" and an accident will result. The human-engineeering explanation relates the hazardous locations on the highway to the occurrence of accidents. It implies that when drivers traverse or pass through a hazardous location, they will encounter increased driving demands and experience slower reaction times, which affect their driving performance and may result in accidents or hazardous maneuvers. Therefore, hazardous driving situations, which are partially affected by the roadway environment, may be investigated to uncover hazardous locations. MacDonald (63) reported several experimental measures which can be used to investigate the task of driving demands experienced by drivers. These measures include: 1. subjective estimate of driving-task difficulty; 2. calculation of the complexity of attentional demands of the road—traffic environment, based on observation of "events"; 3. indices of drivers' physiological arousal; 4. performance by the drivers of various tasks; and 5. performance of the driving task itself. In a paper explaining how drivers responded to highway hazards, DeWitt stated that: 34 Driver performance measurement rates the combination of decisions the driver makes in determining how and when to make use of manual driving skills as they relate to basic driving knowledge needed to prevent accidents, in spite of adverse conditions and the errors of other drivers. (24) In a study involving the field evaluation of selected delineation treatments on two—lane rural highways, Stimson and Kittelson explored the relationship between traffic performance and accident probability on two—lane rural highways, based on the general hypothesis that: Each of several traffic performance measures and geometric variables could be used to independently predict a portion of the accident potential. The traffic performance measures would indicate the manner in which drivers traverse a given section of roadway, and the geometric variables would in effect define the available factor of safety inherent in the roadway design. Extreme values of traffic performance measures in combination with a limited factor of safety would be expected to result in above average accidents. (119) This hypothesis summed up the previous explanations in a practical manner which explains the interaction between the driver and the roadway. It suggested that the researcher observe and analyze driver behavior as a means of evaluating adverse roadway conditions. It also adopted an engineering approach to the accident problem by considering failure in the highway design as the starting point of an unsafe situation. As far as driver behavior in traversing a hazardous section of a roadway, DeWitt quoted Vansodall, a traffic sal vel 35 safety specialist, "that you can do two things with a motor vehicle: you can change its speed or change its direction" (24). Usually, drivers judge speed in part by vision and in part by tactile sensations, and select a reasonable and safe speed for prevailing conditions. Actually, drivers have the ability to make unique and sometimes surprising adjustments (27,36). MacDonald reported that in a detailed field analysis, Curry found that velocity and an "attentional" demand rating of traffic complexity were negatively correlated because drivers typically decrease velocity as traffic complexity increases in an attempt to maintain task demand. Or, as explained by Bakove, the reduction in vehicle speed enables the driver to adjust to the volume of increasing visual information relative to what can be mentally processed effectively (75). The above behavior is expected from most of the driving population who traverse a hazardous location. However, subjective variations in perceptual abilities, motor skills, and mental disposition prevent all drivers from driving equally well at any one time, or the same individual at different times. Thickry found personality differences associated with either the tendency to "freeze " when suddenly confronted with a high—stress situation “P 0r rise up to an emergency with superior performance. But as a whole, it can be assumed that a driver probably does about as well as can be expected "given the circumstances" 36 (57, 132). On the other hand, misjudgments of speed are expected from unfamiliar, risk—taking, and inexperienced drivers who fail to evaluate the conditions of the road for one reason or another. The collective speed perception of all the traffic traveling a certain section of a road can be measured by observing the distribution of individual vehicular speeds. When drivers are able to determine and evaluate the local situation under freevflow conditions, the resulting speed distribution will be normal and show no skewness. The distribution will skew to the left when drivers perceive the area to be more dangerous than it actually is, and it will skew to the right when drivers perceive the area to be safer than it actually is. This phenomenon of traffic speed distribution skewness, as explained previously, may be used to identify high-hazard locations. This notion was mentioned in a report published by the Minnesota Mining and Manufacturing Company, saying that: "Monitoring the distribution of vehicle speeds to predict accidents is like taking a patient's temperature or blood pressure" (79). With this background review of empirical verification and theoretical explanation, we can suggest the following approach for a highway safety improvement program, beginning with the identification of hazardous locations. 37 Approach Framework The approach presented here is based on findings and practices of previous research, and suits the conditions of Saudi Arabia. The basic hypothesis of this approach is that speed-distribution skewness can be used to independently predict a portion of the accident potential. The speed distribution would indicate the manner in which drivers travel a given section of a roadway, and the geometric variables would, in effect, define the available factor of safety inherent in the road design. Extreme values of skewness index, in combination with a limited factor of safety, would be expected to result in an above-average accident rate. A highway safety program based on this concept would consist of the following steps: Determining suspected locations. Any route chosen for safety inspection would be divided into subsections. From the original roadway design drawings and field observations, suspected high hazard highway locations can be identified using the following information: 1. Any previous accident history taken from police reports, reports from local people, or personal observation or experiences. 2. The reported research findings which correlate accidents with individual design elements (15, 53, 94, 105) (See Table 4). 38 3. The reported research findings regarding the design elements and combination of elements (geometric features) that might lead to violations of driver expectancy and create a geometric design inconsistency (75) (See Table 4). 4. The rating of driver—expectancy tests which measure the readiness of the driver to respond to events, situations, or the presentation of information. The rating will be carried on by a diagnostic expert team who will visit the site. Using a seven—point scale, the diagnostic group (highway design engineers having expertise in highway design, traffic engineering, and human factors) will rate features of the roadway. Spot—speed studies. At each selected suspected location, a spot—speed study is conducted to obtain a speed distribution for each site. Statistical analysis. The resultant spot-speed distribution would be checked for skewness. Classification of safety level of the section. Hazardous or safe sections are designated by the skewness index. Parameters can be compared to the same parameters at locations with known levels of risk to determine the skewness value at which a location ishazardous. Countermeasures. When extreme values are found to be critical, indicating a safety problem, research is conducted to arrive at the most suitable countermeasures. Evaluation. Safety improvement of highway geometrics may be evaluated by measuring changes in the speed distribution skewness parameter. 39 TABLE 4 ROADWAY GEOMETRIC EXPECTANCY FAILURE MODE ANALYSIS BASED ON LITERATURE K‘IDHYS Klimt . “'1 '5 3 '3' :mmt m tutu comm-m WW“ mucus. "ammo—s “HRH ans 0' [Min uswu mint Hut-At rpm-n noun/m smncv msnons Soverelevnion («I no not vs: a.- bill vary poo-m tag: “at an“ on land pro-mun av “one". "UM on rm tumt mud Iluury Icndanu 70'. (90) II amulet-tion - .M It/ltl'. curve Il't apnea dc not Lung Uqu u Ind-nu" and can 0 too It. L D to wind- Lawn aid" H e Om-Iuer And can (msIIuncy with unilar appearing (and; and mlyn 9N6 Dues perception of d Inflmnu entering wad? L-uI-m am Snow grade lines. andqu changes Avoid sudden changes in grade (Q) nude no wool a- lane ar 1 lane unduud- md-Iy (21 I m fillllulsmg bat-pen "rttnl grade and “um MI (24) finned p:dlaudmhlp he- rnt gr Id: and _I linden“ I?! speed changes do ml corral-u with grade (fl) Shortened nah: nuance nun-g lands an wood the gum In: "at" H "In (a_y) can too high 1.10 to Inna: Lou" bound? Id! on [but bound - (E) I an application 0.10 and or u l. Suva tion MI little Izflen In 09"- ling MR (7_7I Curvature [wound curves or La- 'm. or "net sea-tr Ict have "Ill! II tight dunner Irony! mud: “VI! an it Ind-n diHeI-ent nun should anon-ted with In- Lode uith dr Iv 400 It . touring - brne/ dItuMd lib-6m with rennet! In I» avoided outed nmcnt nu «tunic .l§_7I quor '1 wood? (L: __m (m not contu- cu m Direct» verse curves WI do not change mt with rut of In: curvatures III: men an dil- should be "aided Longer dun tur ve nu: Runny cum . roam; PI. nun-e Incorrect cwnnent unr. (a a) AM or anon Il:d uIth IMP (7_7) VII-nu an over- th Wu! d duunze vary and Avail Ill twrulur! “(169M run (ya) (um dnconnn- new u tun (Inn union “III pnser ve con- Dnnn POHIIM u- out tutehq'! 5h- harp (u not ll end 0 Shorter radios (urvn Males on "at" 9' “an" :90 1‘ long linger“ up long [-seo'I lHocIltcd both ? - and 4 - lane (uni menu urine: and ma- :Iao'. m rve vIth higher cutout mus. (fl) I." of cums! on a! canal oped "[21 (fl) lu- lu accelera- In“ mun“ - l mu duunu non once-oi l' ' Short Imam curve) t 17‘! when) would be avoided (Q) vnnuuon, ouvgn lo pzyvo: prop. MJIAL-WK Ior one (Two need a". SI. u con-Ian wide tight an drum or MIIVIK (r nude on nor-ouch (L7) rva I wore- Iunte utlute tumu over-uh; speed so :eM speed In woman and on one Minn super» Iwmflt" ll 5.- u curvatu'd «Human (uuhm Imu) III" II” non! Iuflic control devices 2. vunut Elm?“ MUG“ MM!) Design 'Mavy Ian at change of grlflf Speed 0..an earn an don-cur SIM dun - a ham: mtrol Irvn‘ (E) N: change) not to Imp luau (I61) (31) "In per-nub probin- m mm; mrv nudenu (g) un‘w S- ” upgrade in Ada not I procla- 2 lens 4 4 lanes un— ( S-nloavt sud: steeper w gm. Dun pentat- ion IuiId nan-a, lanes “than of mun-ed weed void NH" and M“!!! dip “Linc“ (L9) guy he: dent run than all ugh-a .9. u considered 0- pther (g) (noun-e “tilt!!! ram (fl) tangent - 5 I (rut - I0.7 - IL! unturned“?- IammdLI(l) Invn: canlrvI twins (naming. no puslng) in.“ tutu" mac-ht div Iced Z_) Iuo CIDII roll... a“, 9",, Inc lnflic control appearance? Curved dawned My». nt “(Ideal mum - Lhtev cum at m w "mung Straight Made In I ecu-n! loves! accident loom (3) Sa- u upgrade $19M dilunu nurlc- her no. drive him up (last lrdng oIao-nt [II-irate by n- Yunnan“ .1 Ion» In taxi ted lIth :ntldn vehicIun on a! sight .mnr) mttwrtian “an,“ a, gradient (Mt-pt? 40 TABLE 4——continued l W' GITIIIA m H519 (“Hm "IL“ mm mm an!“ U CHM um mm m man “VIC ”In.” w"- m wit ' J. mmmmsJ Willi"!!- lmnoctI-n Inmtius o-la ken-It m- ! 1 sun am on m Mont mun: has hum "III-m M” 5"! mm have mu ll'l run II Ind m1 correlate with In". Inu-nuou - I. mun: Dy hr um mun-u (I!) (y um— (g) antenna Chap I! paw—t non/Ill- nan-V pun- groan emu-p 7m! b one I.» a! Inc-nonl- Win. a nest Slaw mm: .1th untd pm- d M mm: point than (during or Iwau dual - 1“"- '07-. mam tram: cm a launl "ulna-n emu-g 111—) I. «(Ida-t r“... us. ”(gr 15) (Ian I! not-y lam: natal lieu"!- comm Inuructl- 0r dale-I points 1 points he! wily Halal-d 90 I (2) predictor I": $I~t ‘ cm! a. a rain I! a I (3) [WI 'LI a! Ihunxtlu Is 7“ III-h valued. In n tram: sin-73' kzI‘ont nu heron-I with hem“ of all!“ I I”. (15 _) kmt control In .II ”PM accl mung ”an hrs Cat-try! In I MI le- S- I"! had "ll 1! tin: mule sou-l Omin- leh about: in II or." NIIHI' anpn . r.- vIaI-u "ant.- Imu ~ III 0"! «at-a up" dI—I- I. urea a! canv- cy but nor! a usxutlun Oran unau- “lapel ’0“ I) Ill-r rum l I‘d-ht run at Orv- Inn rum not 5“ Ian any can ”I I), n J n that o! la— addl avian: at unw- mIl-n- m- l-d out lac-tum uIth (L0) Ino mm or tau a! II- - «mu. mnov null-M (if! may «on mod n "- Influu In «giant mar run run undated um hum of not“ and shorter l h of both Vnfllc call“ am: located In Iron hauler-nu and mice: an and IIpht mlnauu Inn (2) diam M— yin- [li-ildu Im tlght - nitric"; II. I" r.- Iuwm (L0) (rut. In. Ohm u hut-v 1|. 9' run-y MM final and manual lu- dm "an: adequate ore"! Hun Inn (20) norm-OI would JdI- 110' not ulth I- inmllm (Ll) I. ans; m Iy h- nun-q sud-I IW-u Mud- n’t lulu- .L I IalnIan II: Ir ‘ din.“ .- vitihlllqi all no IutI- visible u- mm ha than”! II La- ulm not W “Mt "mu“. m,"- Structure: In. |. clean 01 “mun u «flu—e brim “h“ h_ ”‘"” I!) (5) mm :I'IIIDI (upt- may I— mm- emu um um “clam "a n. “ WI distance «a cron— correlation. “.3" re- :7. h In slut-u can't be on» ”coil ”(E. II“ when “CUM! of differ-It emu wen-In Only La- um: I- 02) munch wlu calm hula: :II n Ian- met: (15) MM I:h- and Cruel- hI—It I ma ho! tram: (__) comland with as“. an ram (2_) um vim .fl critical an arid— In In: new (A!) Accld-It run an uni-r 1 land roads Ina an I- 0- 1’) In.” In”! a- unt-r ol how 2 U‘ a In. not (H) n .lae HtD 41 TABLE 4-—continued mu: mm amen me “318 GISIM FM LII! m “LEE F U “(m MET” I”! I“ MGI‘ mu mm " figs can Inn-e) nectar emu Shoulder Sims Very “III; III wee of MM (_) lens- shan‘er rode m Ianr etc-Idem rem mes Inch ewe Insular: (g) 9001" I100 not gone Wm» a! «(Menu (g) S.” con-eleven Shula? fidu Dew and echmts (E) PosItII-e "tum-III. heme! Imlur III!“ and «an! mu. (mg) Io renew-um. or IochmIIe relaun- slnp bet-ecu «(Ice-In IN SW1“? IIOUI (A; knee»: nu decreases as Mulder eIcth Inv «meet on neuronal curm (L) The amen-M rue In. creases es the "newer own Incl-uses a! uncut: for selected value" ) "once: exact- um) emana- uq It: III! III Ieee huh-e a lane hie Deuce“. flm MIC! ”It Del- diastole (than) re- heflon of “IO! um: Threshold of want-7 loeonh Hones should bedulorlesfigg) Steep (lrl) Hopes cause am" over- reecuon end veMcIe menu IQ) J Inn-dude {MINI-ea! Shudder Type CMI-SIM MI ens MI!!! ”1* Genre] III-Nun mafia structures or .30 the- bmieuy ”out queer“ I: when warranted "out! new“ rebut-nu and tenure (2) (made: edjacent to My; Increue feuIItIes (12. 2) (bjecu adjacent to roam, cease IuereI «niece-nu finch occur es nun u "0' vrIor to the object (2. !) Holem emct- Luml DIsoIec—M. any by causing: Pew-(med lane mm urn-mm. Irene WlIutIoe Loss of 11¢". ant-Ice CIeer N' - 50' "CUM! emu" ne- 0' My mush“ o! yer- eereI muse? "mun of mm: o! Ilaee of mum? Fem "tuners Laser kcIda-t late (9 flatten flue l-slope ”win" Acclanl late )5 an "out “wars GI HIM Deniers Mien mun be 2' to "11' nice on My». Ive” nub l or are lanes (g) “In; tend to decrease the MP of «(Menu (33) IIIgI-Ieys 11th truer!- IbIe Indiana-9°04 width “_7 losulve comleuon hen-eon Injury ech- dents end Mr of why: - 25! 0! III accidents on rweI. 000-th1104 ecceu. four-lone :11va (g) finer of echmu Increese Inn Inner of Shanon: preseMIn I change In mum‘s menu; New decIsIen by the III-Iver (32) SOURCE: C. J. Messer, J. M. Mounce, and R. Q. Brackett, fi_ghway Geometric Design ConsistencxfiRelated to Driver Expec- "Procedures for Determining Geometric Design Consistency" (Report No. FHWA— RD- 81— 037). tancy, Vol. III, 42 Application to the Case of Saudi Arabia As explained previously, it will be necessary to depend on indirect measures to predict highway—safety problems in Saudi Arabia or to evaluate safety improvements, until there are complete records and adequate capabilities to utilize them. The methodology presented provides a simple and fast technique to identify high-hazard locations. The following discussion will offer more reasons for using this methodology in Saudi Arabia. In reviewing road-network development in Saudi Arabia using a study prepared by the Central Planning Organization, it is discovered that the present road network is built with various design standards and with various construction practices. At the beginning of petroleum production, roads were built by Aramco to follow the terrain and to serve a light volume of traffic. Horizontal and vertical alignments for high-speed traffic were of no major concern. At the same time, the government built a number of roads with various standards which are not in accordance with modern road design and construction procedures. Road geometries were sub-standard, and pavement strength varied. However, some of the roads were rebuilt or rehabilitated when the government launched its first 5—year Program to develop a major network of modern two—lane highways connecting all the centers of population. Because Saudi Arabia lacked skilled manpower, foreign co: co ac Mi de he re a: CC 43 consultants were engaged to design and supervise the construction of a 4,000—km program. The program actually accomplished 8,027 km by mid-1970! By the end of 1970, the Ministry had standardized its methods for selecting and dealing with foreign consultants. Technical experience that has been gained since the early days of the program resulted in the 1971 revision of the general specification and design standards. Since most of the main road network consists of two-lane roads, the need to launch a program of 4,000 km of divided highways and expressways with higher standards of geometric design and safety was apparent (78, 104). The above discussion reveals that the rural road network in Saudi Arabia has the potential of having problems of inconsistency, because: 1. Even though sections of the old, low—standard roads have been improved, the remaining sections are still in Operation and are connected to higher-standard roads. 2. The road network in Saudi Arabia is built with different design standards. The variation in design standards comes from the variation of foreign consultants Who practice different design standards than those of other foreign contractors. 3. Some roads are built with inexperienced contractors Who rush the job and do not comply with any design standards. Thus, compound geometries, poor visibility, and lack of transitions are common geometric features on the tw ___—___1 44 two—lane rural roads in Saudi Arabia. 4. Finally, the lack of adequate positive guidance, such as warning signs and advisory speeds or the improper use of such signs, increases the problem of inconsistency on the two—lane roads in Saudi Arabia. Drivers in Saudi Arabia, therefore, face problems of inconsistency in the rural roadway network. They are confronted with two types of inconsistencies: from abrupt changes in design and from various changes in roadway quality. This situation is made worse because of the absence of roadway information, which adds to the uncertainty and brings about unsafe driving situations. The problem of roadway inconsistency may be substantiated when it is pointed out that non—Saudi drivers constitute approximately one—third the number of drivers involved in traffic accidents (see Table 5). Apparently, they have less experience with the roadway system and the variations in roadway design standards, which makes them more prone to accidents resulting from these inconsistencies. If further analysis could be done on the available accident data, we may find that a major portion of the Saudi drivers who get involved in accidents are traveling outside their familiar residential region. A solution to the accident prone roadway system of Saudi Arabia needs to be undertaken immediately. In the United States, a very strict screening process is used to identify high-hazard locations, because the U.S. faces 45 TABLE 5 DRIVERS INVOLVED IN ACCIDENTS, BY NATIONALITY Y Saudi Non-Saudi ear . . Dr1vers Dr1vers 1979 20,765 8,783 1980 17,740 10,531 1981 18,451 11,905 1982 18,096 10,922 SOURCE: Traffic Statistics Book, Ministry of Interior— Public Safety, General Department of Traffic, pp. 251—4. acc yea hi1 46 severe financial limitations. Detailed records are kept of accidents over a period of at least three consecutive years. Then, only those sections of highways with the highest incidence of accidents get money allocated for the implimentation of further safety features. Saudi Arabia has less severe financial limitations; therefore, indirect methodology provides an acceptable alternative to predicting highway safety problems in that country, especially when the paucity of data and lack of skilled capabilities to carry on accident analyses is noted. Speed studies are easy to conduct and analyze, and will facilitate and expedite any highway—safety program. These conditions encourage the utilization of the proposed methodology. Because of time limitations and absence of adequate local capabilities, this research will be concerned only with the potential application of traffic speed distribution characteristics as a surrogate measure for high—hazard locations in Saudi Arabia. The next step is to design an experiment to explore this potential. CHAPTER 3 EXPERIMENTAL DESIGN To explore the potential of utilizing the proposed approach and to assure the transferability of the U.S. experience of speed distribution and safety relations to Saudi Arabia, it is necessary to discover similarities or differences in the traffic speed behavior between U.S. and Saudi Arabia rural highways. The hypothesis used in the experimental design is that this can be accomplished by conducting spot speed studies on suspected high—hazard locations and classifying them as hazardous or as safe based on the speed distribution. If the speed distribution skewness agrees with the findings in similar situations in the U.S., then the degree of risk reflected by the Speed—distribution skewness index may be attributed to the tested section in Saudi Arabia. Similarities in speed behavior and degree of hazard will encourage the utilization of the technique for the evaluation of Saudi rural highways. This experimental design will be based on the available literature and past experiences. Field investigations and pilot spot speed studies were utilized to identify constraints, limitations, and particular 47 at Se in Ff) 48 attributes in the study area or the study procedures in Saudi Arabia. This experiment, in addition to the field investigations and pilot studies, consists of the following two major tasks: 1. site selection and description; and 2. spot speed studies. It is important to note that this experiment is not a mere spot speed study; rather, it is conducting a speed study in particular settings and special traffic conditions to develop a sensitive surrogate measure of a hazardous situation. Spot speed studies are a new experience in Saudi Arabia. Therefore, it is necessary to be familiar with constraints, limitations, assumptions, and conditions of the experiment to come up with results that are accurate and reliable. Constraints and Limitations of the Experiment The factors that constitute limitations to the experiment are the following constraints: 1. Lack of accident data (especially accident records by location) made it compulsory to use indirect means of identifying sites with potential hazards, such as using the literature on accidents related to roadway geometries and physical characteristics of the highway. 2. Absence of any spot speed studies and data from Saudi Arabia made it necessary to depend on past eXperiences from the United States in choosing the sample dat sys Stl 49 data. 3. Paucity of engineering information about the road system made it necessary to make extensive field measurements. 4. Time limitations dictated the need for limiting the study area and sample of locations. ‘ 5. Shortage of skilled manpower made the researcher work by himself, which increased the time of the study. Assumptions and Conditions As explained in the second chapter, the concept of speed-distribution skewness as an indicator of hazardous location is founded on an implicitly rational assumption that some correlation exists between the skewness index and accident experience at a particular location. It also implies that as individual drivers interact with the physical characteristics of the roadway, they make an individual evaluation of the adverse roadway conditions. Hence, the observed spot-speed distribution reflects the drivers' behavior when traversing a hazardous location. Since this experiment is dealing with a problem of observing and measuring driver behavior when facing a hazardous situation, it is necessary to avoid, eliminate, or control unknown or undesired factors that may influence driver behavior. These factors are many and diverse, and include volume as a major factor influencing speed choice. Vehicular speeds are controlled to various degrees by th to on vc ax tv 50 the traffic stream and by traffic—control devices designed to regulate traffic flows. Volume has a pronounced effect on speed, particularly on two—lane rural highways where volumes as low as 700 vph (high for rural roads) reduce the average spot speed (90). It has also been found that on two-lane rural highways, the average spot speed decreases linearly with the increase in volume (18, 66, 90). To avoid this effect of volume, the spot speed studies in this study were carried out in free—flow conditions, where drivers are free or virtually free from interference by other vehicles (140). In this way, the speed measured is the speed which reflects the drivers' interpretation of the conditions over the observed locations. The rural road network in the study area, as well as most of the rural highways in Saudi Arabia, carry very light traffic and have free—flow traffic most of the time. The free speed will then be a function of the drivers, their vehicle characteristics, the highway characteristics, and the environment. To further isolate the highway characteristics from other influences, the factors controlled by the environment, such as light, pavement wetness, visibility, etc. were kept constant for all vehicles. The absence of any speed control (speed limits, advisory speeds) or speed enforcement gives drivers more freedom to assume a speed which they think suits the prevailing conditions. However, this situation may increase pre rea cha phy the C011 1'01] whj Spe SEC f0] 51 the variation of individual drivers' speeds as compared to a situation with strict speed control and enforcement, such as in the U.S. Site Selection and Description The study sites were chosen using the procedures presented in the prescribed approach. For statistical reasons, at least two highway sections with similar characteristics were selected for each test. Sites whose physical characteristics or previous accident history met the criteria for suspected hazardous locations were considered for the test. As the last step in selecting a field study site, each prospective site was inspected personally. This field inspection included the following items: -checking roadway geometries for consistency over the route -locating geometrically appropriate subsections at which to observe traffic speed -marking the locations where the spot speed-distribution studies were to be conducted -obtaining all pertinent information about the roadway section using the information sheet shown in Fig. 1, Appendix A The process of locating the study consisted of the following steps. 1. deciding on the general study area; 52 2. choosing the rural network; and 3. selecting the study locations. Deciding on the General Study Area Given limited resources and time, this research had to be conducted in one area of the country. The Eastern Province was chosen to be the general area of the study because it is one of the first areas in the country to enjoy a road system (see Map 1). It also features old and new highway facilities. Two-lane roads serve most of the villages and cities in the area. This area attracts many people from other parts of the country and also includes a big foreign labor community, which makes it a good representative of the population of drivers in the whole country. Choosing the Two—Lane Network After a preliminary field investigation and consulting with the roadway department, it was decided to concentrate on the two—lane network in Alhassa area (see Map 1), because: 1. it carries sufficient traffic to obtain the required sample in a reasonable time; 2. it contains roads with both good and poor design; 3. it is accessible and familiar to the researcher. I . , , ‘ $0.!» J F . .71». -_ , 16”.. up. thaw? rain!!! mun 9W5 Add}... w..- - "p ‘- V. ' \A—A“"“" .' a . \OAIMQQIH‘ ' _. o. 5" (Asa 3.: .' . .g, ’ {Ki-9 I“ I g " A .1‘ flaw} $16 e’ o loc: obs: con‘ int: netI 10c: of 1 cho: p031 and her: dom Cup ”81 res: rat: shat Cur. if; 54 Selecting Study Locations Rather than identifying potentially hazardous locations using local police information or citizen observations, this research was left to choose among conventional hazardous highway locations, such as curves or intersections (17, 75). Curves on the two—lane rural network were selected as the study population of hazardous locations. Limiting the study to curves has many advantages, one of which is elimination of any bias coming from subjective choices of potentially hazardous locations. Also, it is possible to compare results in Saudi Arabia with findings and trends in the U.S. This is because the hazard on horizontal curves is empirically and theoretically documented (5, 53, 54, 98, 99). Also, speed behavior on curves is well reported (49, 72. 100)- Potential Hazard of Curves Curve hazardousness is well documented in the research, both empirically and theoretically, Many researchers (5, 53, 54, 98, 99) have found that accident rates vary with the sharpness of the curvature and that Sharp curves have a higher incidence of accidents than flat :urves. Moreover, the hazardousness of the curve increases if it is unexpected or associated with other hazardous ele wii 55 ants. In general, the hazard of the curve increases the increase in curve sharpness (49, 72, 100). Driver Behavior on Curves In relation to driver behavior on curves, 3 study by )epartment of Main Roads, N.S.W. (22) observed that :les generally decelerated through the approach half of :urve, reaching their minimum speed on the departure of the curve center. And while passenger cars tended :celerate through the remainder of the curve, :rcial vehicles maintained minimum speed. As an explanation for this behavior, Ritchie found a 1g inverse relationship between Speed and lateral .eration for speeds above 20 mph. This means that drivers negotiate a curve and select their speeds according to a velocity lateral acceleration relationship. Therefore, if speed is considered a manifestation of drivers' desire or utility for expedience, lateral acceleration may be viewed as a comfort index or safety criterion. (100) This behavior was also operationally explained by in, who concluded that: a) operating speeds on curves are linearly related b) sight distance also affects operating speeds on a curve, but to a lesser extent than :urvature c) curve super—elevation has no effect an vehicle speeds. (121) The above discussion on vehicle speed behavior fves conforms with the expected behavior of drivers ~sing a hazardous situation. It also shows that speed 56 stribution characteristics may be a suitable check for zardous situations. Sample of Potentially Hazardous Locations To explore the general speed behavior of Saudi drivers 1 to study the trend of skewness index with various grees of hazard, it is necessary to choose a wide ectrum of hazardous situations. Since the hazard :reases with the sharpness of the curve, tangent :tions, plus a sample of curves with a wide range of ii and several reverse curves, were chosen to provide desired spectrum. Previous studies of driver behavior on curves used ferent sample sizes of locations. Taragin (121) measured eds of passenger cars at the point of minimum sight tance for the inside and outside lanes of 35 curves on -lane rural roads in the U.S. Emmerson (25) measured ads for 12 curves on two—lane rural roads in England, 1 curve radii ranging from 21 to 460 m. In a study of measurement of vehicle free speeds, Both (12) selected rural locations. Finally, in a study of using skewed ad distributions to locate points of high accident ential on low-volume, two—lane highways, Krzeminiski acted 12 sites to conduct his study. In general, if the locations are considered vidually, then the emphasis on the sample of vehicles an re v2 st t6 e) as v2 lc 57 l the sample of locations would depend on the available :ources of the experiment. But if the effect of location 'iation is to be considered, then the sample of locations uld be 30 and above, depending on the statistical ting to be done and the available resources of the eriment. Since this research is considering the curves separate treatments, but is also interested in the iation of their effects, a sample of 30 or more ations was required. The sample of locations selected included 61 curves h radii ranging from 120 m to 5000 m. The study sample 3 included 12 reverse curves. For comparison with s—complex situations, speed studies were also attained 12 tangent sections. Site Description The absence of engineering information on local road Letrics made it necessary to make extensive field urements to obtain information on the curves. Some of information —— like curve length, cross—section, and side characteristics —— were obtained at the time of study. To obtain the radius or the degree of curvature, as necessary to conduct some surveying using a dolite. Knowing the curve length, it is possible to determine degree of curvature and thus the curve radius. As a in Fig. 6, the angle may be measured by placing a 58 Fig. 6. Measurement of the angle by placing a dolite on point PI. 3p, Cu: EV: 0b: 101 59 eodolite on point PI, which is the point of tangent :ersection. The intersection angle a may be found using the lowing equation: a: (a1 + a2) — 180 = 180 — 6. Using the arc definition of degree of curve, which has ditionally been defined as the angle subtended at the ter of the curve by an arc lOOft. 10ng,the degree of cur- ure and the radius may be computed using these equations: a: 1‘ Da ‘ 100L 2. R = 5723.578 3 In Table 14, all geometric information is tabulated each curve. Configuration of Monitoring Site Each study section must contain a particular type of ection (i.e., test site) a few hundred feet long over h traffic speed is to be monitored. In this experiment, d monitoring was conducted at about the midpoint of the e, which is the prescribed practice in a study of field uation used in selecting delineation treatments on two— rural highways (see Fig. 7). For reverse curves, speed rvations were made at the midpoint between two curves, rescribed in Fig. 8. In the case of tangents, the ‘vation was taken on a flat section in the middle of a tangent, as prescribed in Fig. 9. 60 0mg. .a .a: .o>u:o Hancowfluo: pom msumumamm ucosoasmmwe mo coeummswflmcoo "mozzom ..\. a; I}?! g I 73.3“» gr. *1. 9.33.68?! .1) c.2250 .o .2 e: I‘M 51E!) 23250.92: roar-zit: ..... 5 Us. .98 232.692"! tau 9.: Etta»; .‘tluld 61 .3 .a .0: .coaumsufiw wcflwsfls pom msumnmmam usosousmmoe wo coeumuswwwsou "momzom .w ices—.5355 82:93 E... 2:558 3335. 25:5 6 5.35 9.23 .28 nc< 9 .EoESoE .533 “8on N. 02:0 .5822 G 25682.. .22... :29...» .o .5092} 23m @ .EoEou-E .223 Ocean .§ 0.630 7:009... e .gstsnnIoz :0:de £02..“ 6 OCJU .0 G . .500 Q r a of .h 9:395 7:009} C 25.0 J .wE 62 .coaumzuflm ucomcmu sow .Hq .Q .OHH "MUMDOW . .mam msumpmmmm acosmusmmos mo macsumnswflwsoo o .n-CoEfiOeOCO .0230“ V:- OC:‘—¢3 0003—027 E33> 68.36 2...... .58 ti 6 L235 .353 v9.3. 055 :83: ES. ..: 8»... E 2v e :9: 83.0 o... :0:an .805. co :50; e a 30:33.3! ¢b=cooa c3...” 63 Spot Speed Studies The second major task in this research is conducting vd studies at the selected sites. This part isted of: -Pilot speed studies to observe the nature of traffic, liarize the researcher with the speed—measurement ce, and identify any peculiar criteria for speed urement in Saudi Arabia. —Verifying speed studies. A few sites were chosen to ire the approach speed along with the speed in the Le of the curve, to verify the assumption that vehicles .erate when they enter the curve. -Actual speed studies conducted on the selected .tially hazardous curves, along with the reverse curves angent sections. Conduct of the Studies The actual collection of data at the selected ions was made during a period of four months between ry 10, 1983 and May 10, 1983. To assure a sufficient a of vehicles in each location, the speed studies were :ted in the daytime for a period of four hours —— an 9:00 a.m. and 1:00 p.m. iecause drivers are influenced in their choice of ,1ar speed by many variables (9, 31), it is necessary e sure that the effects of variables other than 64 iway characteristics are eliminated or decreased. The ;t controlled variables are the environmental variables, Luse they are random in nature. However, in this study, .hose not to obtain data when it was raining or dusty, use adverse weather conditions tend to lower spot speed . Also, the speed studies were only conducted in the ime of weekdays, to avoid any possible differences een weekdays and weekends (90) and to eliminate the cts of nighttime driving, because low luminance levels e the driver to travel at slower speeds and to be more ious (65). Even though the influence of the hour of the on speed patterns is subject to disagreement, the :e of morning hours, 9:00 a.m. to 1:00 p.m., eased any : of the influence of this effect, especially when it aported that speeds are slightly higher around 3:00 to p.m. than during other hours (18). The distribution of road-user characteristics is a ,ficant variable in this experiment, because the ved parameter (speed) is a result of the drivers acting with the roadway characteristics. The variation eed distribution has been correlated with variations road—user abilities and characteristics such as age, residence, trip distance, and experience (40, 90, In this study, the assumption is made that driver cteristics are uniform across all samples obtained. that women do not drive in Saudi Arabia, one of the variables is eliminated; and there is no reason to 65 Lieve the remaining characteristics would vary ;nificantly among the sites chosen for data collection. ‘ Another factor which influences speed is the roadway racteristics. In this experiment, the effect of facility e on speed is eliminated, because all sites are located the two-lane rural network (90). In addition, a check " made to be sure that the pavement surface of the ected sections was in good condition, because rough ament surfaces require a reduction in speed and cause ad to be influenced by comfort criteria rather than aty criteria (81). The influence of the motor vehicle on spot—speed 'acteristics is related to the performance capabilities he vehicle. Generally, motor vehicles can be grouped passenger cars and light trucks, and heavy vehicles 18, 31, 90). Because heavy vehicles' free—flow speed ffected more by the vehicle capabilities (which could be controlled) than the road characteristics (12), it decided to consider speed measurements for passenger and light trucks only. Because traffic volume is a major factor influencing speed, it received considerable attention in this 'iment. Free flow traffic speed is the data needed to the hypotheses in this study. Consequently, only 1e free speeds —— i.e., the speeds adopted by drivers uninfluenced by the presence of other traffic —— were red during the study. In this way, the speed measured 66 the speed which reflects the driver's interpretation of roadway characteristics. The assessment of whether a cle was traveling at its free speed was a matter of ment by the observer. Consequently, speeds were not rded if there was any doubt whether a vehicle was eling at its free speed. As for the effects of interaction with other vehicles, as found that moving in platoons caused speed ctions, especially when heavy vehicles move as platoon ers or constitute a high proportion of the traffic am (141). Therefore, to obtain a more representative e of free-flowing vehicles, speed measurement was ed when there were platoons of vehicles or heavy :les in the observed direction or in the opposite lane. Although speed enforcement is lacking in Saudi Arabia, [rivers are not familiar with radar speedometers, it ecided that the observer and the radar meter should be aled in a small passenger car throughout the study. The accuracy of speed measurements also depends on the r use and operation of the radar meter. For accurate rement, the observer must associate a specific vehicle .ing at a Specific speed, make a positive fication of the vehicle, and assure minimum effect of vehicles (142, 143). The fact that speed measurement 3 experiment was conducted by one person eliminates ions due to human factors. More important to the y of speed measurement is the proper location and 67 ection of the radar. Usually, radar meters measure the eds of both approaching and receding vehicles. In this eriment, the radar meter was portable and set up inside mall passenger car, so that it was possible to align the trument correctly with respect to the line of travel of vehicles. Finally, decreasing the variation due to d-measurement errors was accomplished by selecting an uate size and representative sample. This is discussed he following section. Sample Requirements The choice of the sample size to be used in conducting data—collection effort is one of the more important sions to be made in the planning phase of any research. The exact sample size required in any statistical ysis is dependent upon the size of the interval and the 1 of confidence which is desired. It is also dependent the particular parameter being estimated. Walpole and Meyers (133) stated if the sample mean used as an estimate of the population mean, we e (1-a0 100% confident that the error will be less a specified amount e, when the sample size is given 2/25 68 e Z is a normal random variable with mean zero and ance one and d is the standard deviation of the lation. Walpole and Meyers (133) stated that the sample dard deviation "s" can be used as an estimate of the lation standard deviation if the sample size is equal greater than 30. This sample standard deviation can tained by taking a small set of observations, or it e estimated from past studies. Oppenlander (87) found a standard deviation of 4—5 mph can be used as a "rule umb" in minimum sample size determination, as shown in 6. Webster and Gruen (136) offered the following ssion for determining the minimum number of vations to predict the properties of a normal ibution by a sampling procedure. 2d = minimum sample size normal deviate to the desired confidence level - standard deviation of the sample - normal deviate corresponding to the percentile stimated = permitted error in the estimate series of graphic solutions to the above theoretical IS were developed by Oppenlander, Bunte, and Kadakia 69 TABLE 6 STANDARD DEVIATIONS OF SPOT SPEEDS FOR SAMPLE SIZE DETERMINATION . Are Highway Average Standard Standard Error 1C a Type Deviation of Estimate Two—Lane 5.31 0.41 4.16 0.38 Four—Lane SOURCE: Oppenlander, J. C., "Sample Size Determination pot Speed Studies at Rural, Intermediate, and Urban ions," Highway Research Board Record No. 35, 1963, 8—80. —‘_ 70 1) (see Appendix A). Using the "rule of thumb" value for the sample andard deviation, the minimum size can be determined from g. A-l in Appendix A. for s = 5 and confidence level = 95% N = 150 vehicles Past studies on vehicular speeds have shown that a lple size of 150 vehicles would be an adequate size for s determination of mean spot Speed at each study site , 87). This sample size was based on a desired accuracy less than 2 mph. Another approach was used by Shumate and Crowther O), in which they sampled all vehicles that passed in a en two—hour period. Both (12) reported that on the most hly traveled roads, observations for at least two-hour iods were required to obtain a satisfactory sample, but Less heavily traveled roads a considerably longer period required. Given the conditions of low traffic on Saudi rural .ways and the fact that the speed variance on Saudi ways may be greater than on U.S. highways, it was ded to increase the sample size to 200 vehicles. A —hour period was sufficient to obtain the required le size at all locations, and the sample size exceeded vehicles at some locations. ___— 71 Data'Collection Recognizing that measurement techniques may produce a Lificant bias in data, the author took all measurements 'ehicular speeds with the radar meter. When the data ection was completed for several sites, the raw data each site was input into the computer using a —sharing terminal. (The computer facility at King a1 University was used for this purpose.) By the end of data—collection phase, the raw data for all sites were rded on magnetic tape. Statistical Analysis After recording the data on magnetic tape, the data processed using the statistical methods in the SPSS to the following statistics: x = mean value of the spot speed distribution S = the standard deviation of the distribution #3 = skewness index #4 = kurtosis The statistical tests are based on the general ption that spot speed distribution characteristics in Arabia are similar to those in the U.S., and more cularly that the spot speeds on a particular roadway ion from which a sample is drawn are normally ibuted. The statistical tests include the following. Normality of the spot speed distributions are checked Tables F and G of Appendix B (2), which give 0.05 and 44__‘________________—______________7:::]I 72 1 points of the sampling distributions of/J3 and If a given sample has a value of ,u3 beyond 0.05 point for that statistic, the population may be ned to be non—normal. A more strict test would use 0.01 its. Those tables are used for any normal distribution ‘ lation, but they provide a very strict test of lificance for spot speed distribution normality at Ldent-prone locations. Krzeminiski, in his study, :rmined that the boundary point separating normal from -normal or skewed distributions is a value of Y1 11 to 0.1. Consequently, if the skewness index found for artain spot speed distribution is greater than 0.1, then distribution is deemed to be non—normal. Similarities and differences in spot speed ,ribution characteristics between Saudi Arabia and the are explored by comparing the statistics found in i Arabia with the ones in the U.S. Trends in speed behavior on Curves, and also trends in ness index with curvature degree, are explored using ession analysis. For comparison purposes, the Gosset "student—t" ribution was used to check any significant differences. CHAPTER 4 DATA ANALYSIS The review of the literature has established >nclusively that curves on two—lane rural roads have a .gher accident potential than tangent sections. Based on reeds observed in Saudi Arabia, there is no reason to :pect a different pattern in this study. The main purpose this analysis is to find out whether the skewness index the speed distributions on curved sections reflects the ticipated hazard. Since the ultimate goal of this study is to determine the skewness index can be used as a predictive tool thin a general highway safety program for Saudi Arabia, is necessary not only to verify the transferability of a hazard surrogate measure, but also to establish its uitations, strengths, and weaknesses. Spot speed studies were conducted on a wide spectrum situations, ranging from least hazardous (tangent :tions) to most hazardous (reverse curves), to obtain ,a for the following analyses: -ana1ysis of the speed—distribution characteristics to 73 rrr-—————_—————————______—______::::lIIIIIIIII 74 dentify similarities and differences between driving ehavior in Saudi Arabia and driving behavior in the U.S. —verification of the skewness index as a surrogate easure for hazardous locations in Saudi Arabia. -exploration of any variational or trend haracteristics of the skewness index to establish imitations and criteria for its use. Statistical Results The speed data were processed in the computer using me statistical methods in the SPSS to obtain the following peed—distribution parameters. 2 = mean value of speed distribution 3 = standard deviation of the speed distribution AFB = skewness index M4 = kurtosis Normality Tests The normality of a distribution is used not as a ntinuous variable, but as a classification variable. The 0 classes are normal and non—normal (skewed) as termined at the 0.05 percent level of analysis of the efficient of skewness. The skewness—index value for each eed distribution was checked against the 0.05 percent gnificance limits as shown in AppendixB (Table F). For comparison purposes, normality of the speed ;tribution was also determined at the 0.05 percent level —fi_ 75 by the analysis of the coefficient of kurtosis, as prescribed in Appendix B (Table C). Speed and Speed Distribution Characteristics One hypothesis in this research is that drivers in §audi Arabia are similar to drivers in the U.S. However, :he absence of speed control (speed limits, speed zoning, ,dvisory speeds, and speed enforcement), and the lack of ositive guidance in the traffic operation on two-lane ural roads in Saudi Arabia, suggest that traffic—stream haracteristics including the spot speed distribution may ossess some differences and variations from similar spot peed distributions in the U.S. Therefore, it is necessary 0 identify these differences (if any) and study their ossible effects on the use of the skewness index. Tangent sections with minimum marginal friction were 10sen to observe the Saudi driver's behavior under roadway >nditions where the physical characteristics (road rometrics, roadside development) as displayed in Fig. lO—A Lpose no perceptual difficulties. The parameters of the ot speed distribution for five straight and level tangent ctions with paved shoulders, minimum off-road objects, d adequate lateral clearance are tabulated in Table 7. For the mean speeds, Saudi drivers are shown to proach an overall average speed of 62 mph on two—lane agent sections, as compared to an overall average of 53 1 reported in Table 8 for U.S. drivers. A. 1g. 76 vi? Tangent section with marginal friction. 10. Roadside development conditions. SPOT SPEED DISTRIBUTION CHARACTERISTICS 77 TABLE 7 FOR TANGENT SECTIONS WITH MINIMUM MARGINAL FRICTION Mean Standard Sk wnes N l't mph Deviation Kurtosis Fidex s 02%;“: y E s 60 9 0.210 —O.179 Normal 64 11 —0.004 0.496 Skewed 64 12 -O.517 0.517 Skewed 61 10 —0.154 0.685 Skewed 60 9 —0.277 0.305 Skewed Overall Average Mean Speed = 62 mph = 10.20 Overall Average Standard Deviation 78 TABLE 8 SPEED DATA FOR TANGENT SECTIONS ON TWO-LANE RURAL HIGHWAYS new Speed (mph) 1 . -3 I Nun: 1 Speed Upstreaa Sovnst:eaz Ups::eas acvnstxcaa 31:0 L£:1t Statima Static: Szaczon Station I saber (:ph) 1 2 1 2 M .6 55 49.2 49.3 50.6 51.2 .A 3.‘ 55 $5.5 55.2 $6.3 $6.6 LA 7 5'3 57.5 57.4 $5.3 56.9 '3. So 40 47.6 - 45.3 51.6 51.: A 29 55 64.9 54.4 $6.7 55.5 5 so so 51.: 5:.o', 52.4- 1.5 A 13 55 56.7 455.8 57.2 57.3 A 16 55 56.9 55.6 57.3 57.1 106 SC (7.9 #3.: (3.1 (5.5 A 25 55 55.9 55.0 56.1 55.1. 1‘ 19 55 £9.83 49.0 ‘9.8 49.7 1‘ 30 SS 52. 52.6 53.5 53.5- Speed Variance (97m)‘ .55 55 72.3 72. 72 S 53.1 .3'.’ 55 65.? 61.3 150.; 92.5 A 7 55 77.8 31.1 "3.5 103.7 50 40 61.5 69.5 $7.0 56.3 29 55 éO.I 55-.) 0.8.1 7.5 60 S 98.9 90.0 112.. 96.6 13 55 59.5 62.9 32. 51.5 16 SS ‘75: $7.1 65.4 65.6 .C6 50 77.5 74.8 35.8 39.7 25 55 {1.5 $1.3 65.0 59.5 19 55 59.6 57.4 33.2 57.8 30 55 $7.3 39.2 62 6L7 iOURCE: W. H. Stimpson, H. W. McGee, and W. K. SOD , "Field Evaluation of Selected Delineation ents on Two—Lane Rural Highways: Final Report" ngton, D.C.: U.S. Department of Transportation, 1 Highway Administration, October 1977), p. 206. 79 For speed variation around the mean, the speed idistribution registered an overall average standard deviation of 10.20, as compared to an overall average ‘standard deviation of 8.09 for tangent sections on two-lane rural highways in the U.S., obtained from Table 8 (119). This relatively high average speed and standard deviation underline an excessive speeding habit for many Saudi drivers. This excessive speeding habit is better displayed in Table 9, which indicates that on the average about 87 iercent of Saudi drivers exceed 50 mph, and about 50 lercent of them exceed 60 mph, as compared to 75 percent xceeding 50 mph and 36 percent exceeding 60 mph reported or drivers in the U.S. (135). Speeding on two-lane rural roads in Saudi Arabia was sported in 67 percent of all the accidents which took lace in the country during 1980 (see Table 2). Although 1e report probably overemphasized speed as an accident Luse, it still reflects a long—time observation of gressive driving and excessive speeding of many drivers Saudi Arabia. In explanation of the excessive speeding problem on a two—lane rural roads, Knell (55) reported that on >—lane carriageways or motorways in the United Kingdom, ,vers are inclined to take greater risks, and usually eed speed limits despite the risk of fines. The absence speed control, which is known to have a marked effect in 80 TABLE 9 PERCENTAGE OF VEHICLES (PASSENGER CARS AND LIGHT VEHICLES) EXCEEDING 50 MPH AND 60 MPH Percentageof Vehicles Percentage of Vehicles Exceeding 601nph Site No. Exceeding 50 mph 50 90.00% 59.00% 52 85.50% 56.00% 46 88.70% 46.90% 2 82.70% 44.40% 45.30% 3 87.80% 81 reducing extreme and average speeds and consequently decreasing the variation around the mean (50, 64, 112), nakes the situation on the two—lane rural roads of Saudi Arabia worse. Without enforcement, aggressive drivers are left free to decide their own speed, which is most often higher than the safe speed. The fact that men constitute the total driving population in Saudi Arabia, and young drivers form the iighest proportion of that population, may also partially Explain the problem of excessive speeding. Higher speeds ind higher—risk driving have been associated with men and Loung drivers (90). Moreover, the survey published by The Ministry of 'ransportation (see Table 10) indicates that new cars which are increasingly imported to the country) constitute he highest portion of the vehicle population. These cars re driven faster than older ones because they ride more Dmfortably, travel more smoothly and quietly, handle atter, and are generally in better mechanical condition '0) than the older cars. At the same time, the driver population in Saudi abia also contains some excessively slow drivers, as idenced by the observed wide range of speeds shown in )le 11. The average minimum speed at the five tangent Ltions was 38 mph, as compared to an average maximum ed of 90 mph at the same locations. This problem may be attributed to the fact that each 82 TABLE 10 VEHICLE AGE DISTRIBUTION Percentage of vehicles Average age of . . "r _ vehicle Private ' . Pick- 2-axle 3-ax1e Truck- 1.actor ( ear) passen- Tax; up Bus truck truck trailer semi- y ger car trailer 0-2 38.6 13.6 57.6 21.7 19.6 48.0 17.2 36.8 3-4 15.3 23.0 14.3 3.8 13.4 26.6 10.3 17.9 5-.7 24.8 45.2 11.0 12.8 35.1 21.5 25.1 22.8 8-10 14.2 11.6 11.6 40.8 21.7 2.1 17.2 9.5 11- 15 4.7 5.8 3.5 18.3 6.2 0.9 17.2 4.3 >15 2.4 0.8 2.0 2.6 4.0 0.9 13.0 8.7 Total 100.0 100.0 10010 100.0 100.0 100.0 100.0 100.0 Average 4.9 5.7 3.8 7.5 6 4 3.4 8.6 5.5 age SOURCE: National Transportation Survey, Volume 11L: 'Roads and Road Transport" (Kingdom of Saudi Arabia, Central Dlanning Organization, 1975), pp. 5—14. TABLE 11 OBSERVED SPEED RANGES ON TANGENT SECTIONS ON TWO—LANE ROADS IN SAUDI ARABIA §ite No. Speed Range 46 34 8O 50 38 97 52 39 99 82 43 90 83 4O 85 84 year many new drivers, with little training and short driving experience, are licensed (see Table 12). Those inexperienced drivers often drive cautiously and maintain their cars at very low speeds (90). The lack of speed limits, which are reported to bring lower speeds closer to the mean (56), contributes to this wide variation in speed. In summary, the absence of speed control, combined with certain driver and vehicle characteristics in Saudi I Arabia, partially explain the high incidence of excessive i speeding. However, the same combination of factors on U.S. , highways would probably not result in the same speed distribution, as the higher speeds stem also from the aggressive driving attitudes of many Saudi Arabian drivers. The lack of adequate driver education and the absence of stringent licensing tests allow many inexperienced drivers to drive in the roadway system. These drivers often travel at excessively low speeds. Therefore, the spot speed distributions in Saudi Arabia would feature relatively higher means and standard deviations than similar speed distributions in the U.S. The distribution often contains extreme high and low speeds on one or both ends, which increase the variance, the skewness, and the kurtosis. Normality of Spot Speed Distributions on Tangent Sections The distribution of individual speeds for all drivers on tangent highway sections was found to be normal in the J.S. (55). However, the normality test in Table 7 shows 85 .mm .m .Amflnmn< Hvsmm we Eovwcflx .uemmmuH mo ucmsusmqwa ngwcmo .muflusomm oflansmluOflsmucH mo suumflcflzv wpmw» HH mdfluzn moflumflumum oflmmmnH "mombom mm. xx" 5.: - .L co..- .2: 02 XwOZ_ _ r n . . 1] New—r. oo— mom—um _ec— aONNQ .9: p356 w vm— wmsmm npn :9~::.nn wannh—m pmm whpth m; nmwpvn men. mvmwo~.mmc whromr 35¢ adrOh :uu.il|... .I H. -Myll: lawn... 2 I _ l l I I. _ l I _ l I 1. l l umn . I new I imflwflpw ...... I _ 1 _ . ,2 a! own—.8. a8. _.2 SE .3. 8: . S 3: v2 28 2; 5: an. 82 8 82 8 E. "2 83 of 1 Lflwfi $59 on: h—mu _. Nm wnav M 5 Omnv “ NR wows wN— ammo— as... mwhNN chm newnn 3‘0 chew” New mmwum.—wnm cNMMN can UZEDQ emacwmccp moNcN mo— mmmOvmchr NNMNm A; mnwmc Kn ammwop cnc ww—mc— nun nnnuum mNm pNMQON we: mpwwn Lops. NooMN— chm wh<>_:& I“: .15....— c2 :5... 02 Tau:— 02 not... 02 5...... oz .52: c2 .23.. 02 new... 92 5.2.. oz .22.. 02 not... wa>h :3. ”1...... 1... .1.: 3.. _.1... 3.. =1... 1: .1.... =4. .2... _.2 1.... ....2 _.1... .3: :3... :4. 3... “TM. .311 thH mnma «Ema mhaa whmH brad mhma mhaa omma meH «mad wc mmwe OH UZHQmOUU< NmeINmoH mmH~E mo meZH NH mam/Z. 86 that all but one of the speed distributions are skewed. In the U.S., this would indicate that drivers faced some perceptual difficulties, and thus the site was potentially hazardous. In theory, the higher average speed and standard deviation observed in Saudi Arabia should not change the shape of the speed distribution from normal to non—normal. However, since the coefficient of skewness measures the second moment about the mean, a few drivers who drive with excessive speeds would lead to a distortion of this index 2 m - 3 [43 = 3 = ELEJL£15 = skewness index In2 E(x — i) To reduce the effect of those few drivers with extreme speeds, the speeds at both ends of the speed iistributions shown in Table 7 were eliminated by only :onsidering speeds within two standard deviations of the mean (E i 20). With this data modification, the 'esults were as shown in Table 13, and the skewness .ndexes registered the expected normal speed distribution ’or all the tangent sections except one. This modification ad little effect on the shape of the distribution, as hown in Appendix C. In addition to the spot speed observations on tangent ections of highway with no roadside interference, another roup of tangent sections with more marginal friction on 0th roadsides (unpaved shoulders, slopes, ditches, and 87 TABLE 13 RESULTS OF THE MODIFIED SPOT SPEED DISTRIBUTION CHARACTERISTICS FOR TANGENT SECTIONS Site Mean Standard Kurtosis Skewness Normality No. mph Dev1ation Test ' Test 46 60 7.435 0.311 —0.154 Normal 50 63 9.779 0.635 0.227 Normal 52 63 10.482 —O.620 0.280 Normal 82 59 8.744 0.418 0.658 Skewed 83 59 7.993 0.671 0.129 Normal Overall average speed = 61 mph Overall average standard deviation = 8.89 88 Less lateral clearance, as displayed in Fig. lO-B) were selected and spot speed data obtained. As shown in Table 14, the presence of roadside friction resulted in a speed distribution with both a lower mean speed and less speed variation. Consequently, the skewness indexes indicate a normal speed distribution for all these tangent sections with no modification of the raw data. Since this level of roadside development is commonly found in the United States, this result tends to verify the relationship between sites with relatively low hazard and a non-skewed speed distribution. The skewness index is very sensitive to traffic aperational characteristics as well as to the physical :onditions of the roadway (125, 126). This sensitivity may Lead to unrealistic results because of the complexity of :he factors influencing spot speed characteristics. In the .bsence of speed control measures in Saudi Arabia, the ariables influencing speed choices increase and are even ore complex than in the U.S. Human factors, which are more omplex and random in nature than the physical factors, ssume a more significant role in speed choices. Even with the Saudi attitudes of aggressive driving, 1e lack of speed control, and different driver and vehicle aracteristics, the drivers' speed behavior on tangent ctions with some side friction proved to be normal, which similar to the drivers' behavior in the U.S. Where there no side friction to restrict very fast drivers, when 89 TABLE 14 SPOT SPEED DISTRIBUTION CHARACTERISTICS FOR TANGENT SECTIONS WITH MARGINAL FRICTION ite Mean Standard Kurtosis Skewness Normality No. mph Dev1at10n Test Test 5 51 8.364 —0.312 0.133 Normal 3 54 9.447 -O.187 0.193 Normal 6 46 7.565 —0.435 0.227 Normal 8 53 6.29 —O.230 0.062 Normal 5 47 7.619 —0.297 0.130 Normal 6 47 7.233 —0.823 0.204 Normal Overall average speed = 50 m.p.h. Overall average standard deviation = 7.75 90 eeds which exceeded two standard deviations beyond the an were eliminated, the skewness indexes registered rmal. This conclusion sets the stage for testing the pothesis that speed-distribution characteristics can be :ed to identify operational deficiencies and thus .zardous situations. Speed Behavior When Negotiatinngurves Since spot speeds are to be observed only at the nter of the horizontal curve, it is necessary to termine whether driver behavior when traversing a curve similar in the United States and Saudi Arabia. In the 8., when traversing a horizontal curve, drivers celerate to a constant speed, then accelerate when erging from the curve. This speed behavior is a measure the interaction of the individual drivers with the inging roadway conditions (46). This behavior was checked by observing speeds on eight ves and their respective approaches, the results of ch are shown in Table 15. On tangent open sections of a d, a driver's choice of free speed is determined by his ired speed and vehicle capability. However, on curves, speed is limited by the road geometry, a restriction h applies to all drivers. Therfore, the mean and dard deviation of the speed distribution on the curve d be expected to be less than on tangent sections. 91 TABLE 15 SPOT SPEED DISTRIBUTION CHARACTERISTICS ON CURVES AND THEIR APPROACHES Approach Curve Mean Standard Mean Standard mph Deviation mph Deviation 44 6.719 43 5.697 46 8.835 45 6.261 I 41 7.698 37 7.0741 45 6.447 39 5.730 3 46 8.165 41 6.511 ' 51 8.364 33 5.608 53 6.290 34 4.318 47 7.619 41 6.511 92 When comparing the parameters of mean speeds and tandard deviations of the curves and their approaches sing the Gosset "student" distribution, as shown in Tables 6 and 17, the differences proved to be significant. This same that drivers, when confronted with a curve, changed heir speed to suit the new conditions, and they also drove loser to the mean. The fact that the speed changes when moving from a angent section to a curve section verifies the assumption lat the speed the driver assumes on the road is related to Ls expectancy. Drivers select a reasonable and safe speed )I prevailing conditions, and change this speed when the :pectancy is altered. The observation that there is a lower standard aviation means that the geometric complexity of the curve, ,ich is more restrictive than the tangent section, has fluenced all drivers to drive slower and closer to the an. This tends to eliminate extreme speeds, and speeds ken at the center of the curve may not require dification before testing the skewness index. The ysical conditions of the curve rather than the driver aracteristics determine the speed of traffic. It is a common observation that drivers slow down when ey encounter a horizontal curve on a rural highway (121). a extent to which they slow down appears to be related imarily to the degree of curvature. Taragin (121) >orted an average speed change of 0.7 mph for each 93 TABLE 16 COMPARISON OF MEAN SPEEDS (MPH) FOR THE CURVE AND ITS APPROACH APHZSHCh fii;xf Difference 44 43 1 46 45 1 41 37 4 45 39 6 46 41 5 53 34 19 p 0 74% o . —x (x. —i)2 1 —6.5 42.25 S = 330 = 7.07 —6.5 42.25 _ 12.25 = 7.50 — O = 3.5 t -———7.07 76 3.00 —1.5 2.25 - o 6.25 = 2 5 tO'OS 2.365 F0 10.5 110.25 11.5 132.25 D.F = 7 —1.5 2.25 t = 3.0 > t0.05 = 2.365 350 reject the null hypothesis 94 TABLE 17 COMPARISON OF STANDARD DEVIATIONS FOR CURVES AND THEIR APPROACHES Curve Approach Standard Deviation Standard Deviation Difference 6.719 5.697 1.02 8.835 6.261 2.57 7.698 7.074 0.62 6.447 6.730 0.75 8.165 6.511 1.67 8.364 5.608 2.76 6.290 4.318 1.97 7.619 6.511 1.11 §=0 3940 xi — i — i) —0.54 0.29 S = 4.?7 = 0.82 1.01 1.02 —0.94 0.88 t = 1.868; 0 V6 = 5.38 —O.81 0.66 ' 0.11 0.01 t0.05 = 2.365 FmD F = 7 1.20 1.44 0.41 0.17 t = 5.38 > t0.05 = 2.365 —0.45 0.20 4.67 reject the null hypothesis 95 ree change in curvature. This implies that the tighter urve, the greater will be the influence of road try on drivers' choice of speed. That is obvious when ean speed and standard deviations of the individual distributions for the curve sections in Table 18 are ssed with the radius of the curves and their degrees rvature. (See Appendix E.) For the mean speed, the curvature radius produced stically significant effects at a significance level < 0.05. This equation shows the speed increases as adius increases. v = 36.67 + 0.0023R r2 = 0.57650 4| ll mean speed (mph) R = Radius (feet) F = proportion of variance of the independent variable explained by the regression The same trend found in this study is supported by the Lcal speed curvature relationship reported by McLean Ln Table 19, which shows the 90th percentile speed, )ercentile speed, and median speed as curve radius ses. sing the radius as the measure of the curve, Fig. 11 ys graphically the relationship between the mean and curvature radius, which is comparable with the 96 000.0 000.0- 300.3 000.00 .000 .00.00 .000 00 030.0 300.0- 000.0 000.00 .000 000.03 .000 00 030.0 000.0 000.0 000.00 .000 000.03 .000 00 000.0 000.0 300.0 000.00 .000 003.00 .003 00 000.0 000.0 000.0 000.00 .000 003.00 .003 00 030.0 000.0 000.3 000.00 .000 600.30 .003 00 303.0 000.0 000.0 000.00 .000 000.30 .003 00 000.0 300.0- 000.0 000.03 .000 033.0 .000 00 000.0 000.0 030.0 000.03 .000 033.0 .000 00 003.0 000.0 000.0 000.03 .003 600.0 .000 0 300.0 030.0 003.0 303.03 .003 000.0 .000 0 000.0 000.0 000.0 000.03 .000 000.0 .0300 0 003.0 000.01 300.0 000.03 .000 000.0 .0300 3 003.0 000.0 000.0 000.00 .000 600.00 .003 0 000.0 300.0 000.0 000.00 .000 .00.00 .003 0 300.0 000.0 003.0 000.00 .000 000.0 .000 0 mwmmwmwm 00060000 060000>60 00000600 flammw mowmwmwma 00 000000 600000 0000 wm>m30 mom moHHmHmmBoom unmwcmum cummm mLMWU opsum>uso 0w .oz mmwcsmxm 0mm: mo summed we mwhwmm mzwcmm muflm iiil 98 0...0 000.0- 000.0 000.0- 000H0 000.00 .000 000.0 .0.0 00 000.0 0.0.0 MWM.M 000H00 .000. 000H0 .0000 0M 00... 000.0 000.0 000 00 .000. 000 0 .0000 0 000.0 000.0 . 000 00 ..00 000 0. .000 .0 . 000 0 .00..0 ..00 000.0. .000 00 0.. 0 000.0 000.0 000.00 .000 000.0 .000. 00 000.0 00..0 000.0 000.00 .000 000.0 .000. 00 .00.0 00..0 000.0 000.00 .000 000.0 .000 00 .00.0 .00.. 000.0 0.0.00 .000 000.0 .000 00 000.0 000.0 000.0 000..0 .000 000.0 ..00 00 000.0 000.0 000.0 000.00 .000 000.0 ..00 00 0.0.0 .00.0 000.0 .00.00 .000 000... .000 00 000.0 000.0 000.0 000.00 .000 000... .000 00 000.0 00... 0.0.0 000..0 .000 000... .000 00 000.0 000.0 ..0.0 000.00 .000 000... .000 00 000.0 000.0: 000.0. 000.00 .000 000.0 .0000 00 00..0 0.0.0- 000... 000.00 .000 000.0 .0000 00 00..0 000.0- 000... 000.00 .000 000.. .0000 00 X®UCH C ”M” UM. ®HDUN>HDU um” .02 00003000 0.000000 00.000000 0000 000 w00Lw 0o0mwmw0. 00 000000 00.000 00.0 |\I 99 000.0 000.0 000.0 000.00 .000 000.00 .033 00 000.0 000.0 000.0 030.30 .000 000.00 .033 00 030.0 000.0 030.0 000.00 .000 000.00 .003 00 030.0 000.0 000.0 030.00 .000 000.00 .003 00 003.0 300.0: 000.0 000.03 .000 000.0 .0000 00 003.0 030.0| 000.0 000.03 .000 000.0 .0000 00 000.0 300.0 000.3 000.00 .000 000.00 .003 00 300.0 030.0 000.0 000.00 .000 000.00 .003 00 000.0 000.0: 000.0 000.00 .000 000.0 .0.0 00 0E u . mmmflwmwm m0mounsx :o0u00>00 00002000 ”WMMW mommmwwmq wmswwwmmm mswwmm mhflw 100 TABLE 19 EMPIRICAL SPEED—CURVATURE RELATIONSHIPS Dccendcm wnablo 10503130000! Tuacln. Dug. Enmoncn. ’- “‘ -°-° 00m Pctccnmo speed can: oat-:cnmo 30000 new” ’00“ (km/h) (hm/M [km/h) _W Curve radius 59.1 + 0.0058 52.3 + 0.0038 40.8 + 0.0970 R (m) t2 = 0.59 t2 = 0.91 t2 =2 0.77 \/R 432+ 2.10 VF! 31.7 + 2.05 VR 25.9 + 2.52 \/R r2 = 0.67 r2 = 0.90 [1' = 0.68 Curvalme 89.4 — 0.450 03.1 -. 0.550 "73.7 — 0.190 C (deg! 100m) :3 = 0.74 ' r2 = 0.73 t3 = 0.87 Exponenual' 53 (1 .. a 4.0145) 39 (1 .. e —0.01R) 74 (1 .. o -c.oum) :2 = 0.73 ' :2 = 0.71 r2 = 095 V0 (1 -—e-Bn) ' An Mun-.9 crcceCmo was used [or the cxponcnzul model to that Ibo pmamolou ler um model may be alt-trims! SOURCE: J. R. McLean, H . . Dr1ver Behav1or on Curves: View," ARRB Proceeding§ 7 (Pt. 5) (1974):135. 101 0 (I (M Speed (mph) ; 3 32 E” b». s 3 :jb lfigiNS ‘ . : mm ‘3 40:: 2:3: 1:00: '40:: 23:5: Radius (feet) ig. ll. Empirical speed—curvature radius relationship d for the data collected in Saudi Arabia. 0—————————————--—————————————————:::::IIIIIII- 102 reed—curvature relationship obtained for Taragin's (121) ‘54 data, as shown in Fig. 12. The regression equation is: ; = 49.731 — 0.949Da r2 = 0.60410 v = mean speed D = degree of curvature This equation says that as the degree of curvature creases, the mean speed decreases. This trend is also pported by the speed—curvature relationships reported in ble 19, which shows the 90th percentile speed, the 85th rcentile speed, and the median speed decreasing with the :rease of the degree of curvature. More specifically, it similar to the trend displayed by the Taragin equation 5): V = 46.26 - 0.746D a a Fig. 13 displays graphically the relationship between r speed and degree of curvature. This is also comparable h the empirical speed—degree of curvature relationship nd by Taragin (121), as shown in Fig. 14. For the standard deviation, the effects of the radius found to be statistically significant at p‘< 0.05. The ression equation is: Sd = 5.953 + 0.0005R r2 =0.57061 Sd = standard deviation 103 Sun nan/m ”mm 901“! SPEED DAM .00 — — VIMB'R mm! “_-— yum/[model . '. ’10 -. _ .- v-A-fl/l mud ¢’ /' — {non-tut new 0 9 ‘ fi '0 .. -— - —' " ‘ O 9 . fi § § § 3 Fig. 12. Empirical speed—curvature relationship ained for the Taragin (1954) data. SOURCE: J. R. McLean, "Driver Behavior on Curves——A iew," AARB Proceedings 7 (Pt. 5) (l974):l35. 104 .0 :9 fi‘ 5; E 5019 C o 'H In E "U a) 3 w a c - m Va, A 33 c . 5% (6 Lu” v ‘ OJ €$3 z 8 “SE 3 O. i c :9" 3 m; .. J 4 p 3 7‘59! 3 y: :3 23 40 ' 3: Degree of curvature ig. 13. Empirical speed—degree of curvature onship obtained for the data collected in Saudi 105 s‘~ __ sun: spun aaszo on \ ~~~ cunvuune mo 000:0:szva 5‘ “~- - . ‘~~ L I I K\ ‘ Min-u'mcmmz 2:th \. ‘F ~~ so-mcsrm I; 39:10 7\ \ \<~ . ~) I V wane: sn££o\'~ \ 4' S N‘s \-"s \ \s‘ I X \‘N ‘r\ Vs o 4 I '12 II 20 u an 02 Curvature, degrees ig. l4. Empirical speed—curvature degree onship. DURCE: A. Taragin, "Driver Performance on Horizontal H , Proc. 4RB 33 (l954):452. 106 R = curve radius This equation indicates that the standard deviation eases as the radius increases. This rising trend is displayed graphically in Fig. 15. Using the degree of curvature, this regression tion is S = 8.967 — 0.2112Da d 2 r =O.48602 Sd = standard deviation Da = degree of curvature This equation indicates that the standard deviation :ases as the degree of curvature increases. This ,ning relationship is displayed graphically in Fig. 16. These models confirm the assumption that as the curve es tighter, the drivers decrease their speed and drive r to the mean, and thus the mean speed and the ard deviation decrease. A better fit of the data may tained with a non—linear equation, but the ionship is significant even with a linear ximation. Since these equations are being used not to ct behavior but only to demonstrate a phenomenon, inear equations are not tested. The findings confirm the notion that the influence of sive speeds on the skewness index will decrease as the tric complexity increases. This should increase ience in the skewness index results, as data Lcations will not be necessary. 107 12.000 3.303' snrpea 5.839 J-JC? (J A 0:9-“ «at an at -ovvv hvflvs 400- 20“ 22:: L: Standard deviation Fig. 15. Empirical standard deviation—curvature .us relationship obtained for the data collected in Ii Arabia. 108 5.332 “’5 a (9 O 5.353- 3 .332} u! Cl a 10 2: 10 ‘0 Degree of Curvature Fig. 16. Empirical standard deviation—degree of ture relationship obtained for the data collected in Arabia. 109 The fact that Saudi drivers perceive, evaluate, and espond to conditions on tangent setions in the same manner 5 do U.S. drivers allows the testing of the hypothesis hat they perceive and evaluate the adverse roadway onditions imposed by curve sections similarly to U.S. rivers. The general speed behavior found for Saudi drivers 3 similar to that of U.S. drivers. However, this general peed behavior has not yet been related to the perceptual ifficulties associated with traversing each individual urve, which is the main theme of this research. Normality of the Speed Distribution on Curves The speed distribution skewness index is the tatistical parameter used to register those perceptual ifficulties associated with traversing a curve. While the Jeed distribution of Saudi drivers on tangent sections avealed a normal spot speed distribution, the speed Lstribution on individual curve sections is hypothesized > become skewed depending on the severity of the curve. According to AASHTO practice, curves may be classified : flat or sharp. "Flat" horizontal alignment is considered » be 3 degrees or less, which is the maximum curvature :gree recommended by AASHTO and the limit adopted by many ates O45).A."sharp" curve is any curve with a degree of .rvature greater than 3. As shown in the relationship in g. 17, which was developed by Babkov (5), alignments with ad curvatures flatter than about 3 degrees or with an R 110 8 gaunt I I 4: ¢-_--.-. C05uzn. 1962 (Glen! Datum) s- ___. Bttal. 1957 (F126) M . ' -..-- Covutsh.1955 (Hungary) :2 MM [00:01:20,105 (Japan) 5 5 mo Nowny.|959 (213.01.302.01 —-1«--'-. '0 on 090») .H 9 Pa“. 1953 (USA) U . ---- Tan 1c Enquneeztn f; . “Handbook. 1930 (05A) .. ' Vuudlu’v. 1903 (USSR) '3 0 3.51m. 1955 (0550) 0 H ’v o n 0.; o :-“ L4 in 0’ : .D I’ ' E :MNWW :3 : 0 fl 0 n 2 a, .. ——u g §}~40- (D ‘ u‘wm > x a -r-4 ‘~Tfi>L.- o m l ‘ H \~ a) 1 m I 0 man 2000 3650 ' 50410 Radius of horizontal curves, m Fig. 17. Relative number of road accidents versus 'adius of horizontal curves in meters. SOURCE: V. F. Babkov, "Road Design and Traffic SgetY:" Traffic Engineering and Control (September 1968): 111 of 2000 ft. produce a relatively small decrease in accidents, while alignments with road curvature sharper than 3 degrees produce a rapid increase in accidents. This is in agreement with Messer's rating of the potential workloads associated with traversing different curves, as indicated in Table 20. He found that curves of 3 degrees or less produce less geometric inconsistency than curves of more than 3 degrees. However, excessively long curves were rated proportionatey higher. Using 3 degrees as the dividing line, this research dealt with 13 flat curves and 48 sharp curves. A separate analysis for each of these two curve groups is presented in the following discussion. Most of the flat curves would not be expected to represent a particularly hazardous location, and according to the hypothesis being tested, the spot speeds in those curves would be expected to remain normally distributed. However, as shown in Table 18 and reproduced in Table 21, there is no consistent trend in their rating of normal or non-normal. There are three curves rated normal and 10 curves rated non—normal. Curves with similar or equal degrees of curvature varied in their rating from normal to non-normal. These results might be explained by the fact that drivers in Saudi Arabia approach curves at a higher rate of Speed than do U.S. drivers, and thus, while a flat curve is not hazardous in the U.S., it may be a hazardous location 112 TABLE 20 WORKLOAD POTENTIAL RATINGS (RC) OF HORIZONTAL CURVES Degree of Deflection Angle, [5° Cueraoture 10° 20° 40° 80° 120° 1° 0.5 1.0 2.1 4.1 6.2 2° 1.2 1.5 2.0 3.0 4.1 3° 2.1 2.3 2.6 3.3 4.0 4° 3.1 3.2 3.5 4.0 4.5 5° 4.0 4.1 4.3 4.7 5.2 6° 5.0 5.1 5.3 5.6 6.0 7° 6.0 6.1 6.2 6.5 6.8 8° 7.0 7.0 7.1 7.4 7.7 NOTE: All ratings are for two-lane, high—type highways. Ratings for two-lane mediocre roads (i.e., surface treatment pavement without paved shoulders) equal Ratings for all four—lane highways equal 0.65 rating shown. 0.75 of rating shown. 113 umzmxm omm.o woq.o- Aqmlwmv Nem.HH Ho am.o Hm ewswxm esm.o wNH.o- Amo-omv qsm.mfi om em.o mm emzmxm moq.o mmm.o- AmAlomV ooo.NH om mm.o we emsoxm Hfim.o moH.o Aflolflmv Afiw.afl mm 00.0 sq emsmxm ofiq.o mom.ox Amm‘emv mofi.mfi Ho ow.o mm emsmxm ofim.o mmH.ou Amo-mmv omw.mfi om om.o Am emzoxm mflq.o www.01 Asolmsv amq.mfi so om.o om emgwxm omm.o Afin.o- AmaquV mHN.NH so ow.o mm Hmeuoz Hom.o Awq.o- Aomqmmv NqH.mH so NN.H am Hmeuoz NON.o som.o- Aao-mmv ofin.w ow oo.N mo emzmxm mmm.o Afiq.o flowiomV oNN.a we oo.N No Hmenoz mHH.o mmo.m Aon-omv wNN.w ca mm.m mm emsoxm hom.o qu.o Amnlomv owo.n ow mm.m me x umme xwvcH meoupsx owcmm COflumH>wQ LQE opsum>n=o .oz uflawsyoz mmwcsmxm vmmam upmvcmum :moz wo mwymwm ouflm mm>mao Hwm nae mpsum>pso .oz sufiamsuoz mmmcsoxm vmmam vumvcmum 2mm: mo mmpmmm wuflm om 304mm mm>m=o Hp=o umaw vcm mCOfluoom ucmwcmu HmEuo: mo mwmu: .Nom u NW mamsvm afi momma Hampm>o . msuoz N Nae Mes Nee Nmm NooH sow H Nmm NHw Nmm NNN IND... Now HQEHOZIGOZ ”N N ©m m OH 1.01 q HMEHOZICOZ m o o m o H Hmeuoz NH we a ms 0 m wdeem mo .02 memmam m>fimwmoxm coauofium :Ofluofium LUH3 saw: usonuflz mm>pso mm>uso mm>unu “mam How m>nso msowuomm mCOHuomm mwum>wm aumnm mumm umamavoz umfim ucmmcmh emfiMHcoz ucmwcme > asouw >H asouo p=u mmmm>my wcm apmcw Hmsgo:1:o: wo wwmucmowma HHmhm>o .o H mm m andm mwm wH H :OHumoHHHwOE usosqu mo>p=o umHH vcm m:0Huomm acowcmu HmEpo: Ho mwmucoowma HHmwm>o NNo NHN NNH Nmm wa Now Now HmEmoz N Nw NmN wa Nmm NNH Noq NON HmEHOZIcoz N H «H OH q H m H HmEHOZICOZ HH «m N o m N s Hmsmoz NH wq NH MNH o m m mmuHm mo .02 :OHHUHHm :OHuonm cUHs unocqu mm>ssu mm>pso mm>n=0 uMHm mow m>uso w:0Huomm mumm mCOHuomm ompm>mm apmcm muma vawHwoz umHm ucmmcmH vachoz ucmmcme > anonw >H ADOHU I ’2‘ v 400 "o w / E (D U a’ 306 z < (I) 2 :> / g 200 E 2 / 100 20 / / MIN. so ,/ ___/’33,.“ ’7:::::::=:::::::——————* «0" -"_'-‘ 5 . 1 £5 ( ' B S w STANDARD DEVIATION (S)-M|LES PER HOUR Fig. A—l. Minimum sample size vs. standard deviation (percentile = 15% and 85%, desired confidence level = 95%). APPENDIX B TABLE I" 0.05 uml 0.0l Points of the Distribution of 7.. Nurtuul Universe“ tupproxlumte values) W Probability That 7. Will Exceed Listed Value in Positive Direction Is u ".05 (MN 25 0.7 I4 I .073 30 0.664 0.985 35 0.624 0.932 40 0.587 0.869 45 0.558 0.825 50 0.533 0.787 60 0.492 0.723 70 0.459 0.673 80 0.432 0.63 I 90 0.409 0.596 100 0.389 0.567 I25 0.350 0.508 I50 0.32l 0.464 I75 0.298 0.430 200 0.280 0.403 250 0.25 I 0.360 300 0.230 0.329 400 0.200 0.285 500 0. I79 0.255 750 0. I46 0.208 1.000 0. I 27 0. ISO ‘ Table F is abridged with permission from It. C. Gary and E. 8. Pearson. Tests of Normalin (London: Biomett'iku Ollice. University Collette. I938) and from Ralph B. Asostino and Gary L. Tietien. "Approaches to the Null Distri- bution of V57." Bianwtriku. Vol. 60 (I973). pp. l69-73. The points listed are on the positive tail of the distribution. With a minus sign attached. they are equally valid for the negative toil. Also see Ralph D'Asostino and E 8. Pearson. ”Tests for Departure from Normality. Empirical Results for the Distribution of la, and Vb..." Bionic-(film. Vol. 60 (I973). pp. 6l3-22. Since the publication of this paper. Prof. F. J. Anscombe has pointed out that b, and Vb: are not independent in samples from a normal universe. Hence the joint tests suggested in Section 6 of the paper are net correct. though they might be nearly so. [Letters from E. S. Pearson. April 4. t974.| For a discussion of the operating characteristics of a 7. test for normality. sec K. 0. Bowman and L. R. Shenton. “Notes on the Distribution of VII-,- in Sampling from Pearson Distributions." mum-mm. Vol. 60 (I973). pp. I55-67. 161 TABLE G Percentage Points of the Distribution of 7:. Normal Universe* (approximate values) Probability That y, Probability Thar y, F ails below Listed Falls above Listed Value [5: Value Is: n 0.01 0.05 0.05 0.01 25 --l.28 —1.09 1.16 2.30 30 —1.21 —1.02 1.11 2.21 40 -1.11 -0.93 1.06 2.04 50 —1.05 -0.85 0.99 1.88 75 -0.92 —0.73 0.87 1.59 100 —0.82 -—0.65 0.77 1.39 125 —0.76 —0.60 0.71 1.24 150 —0.71 -0.55 0.65 1.13 200 —0.63 —0.49 0.57 0.98 250 —0.58 -0.45 0.52 0.87 300 -0.54 —0.41 0.47 0.79 500 —0.43 -0.33 0.37 0.60 1.000 -0.32 —0.24 0.26 0.41 2.000 —0.23 —0. 17 0.18 0.28 5.000 —0. 15 -0.11 0.12 0.17 ‘ Table G is adapted with permission from R. C. Gcary and E. S. Pearson. Tests of Normality (London: Biomeiriku Office. University College. 1938). E. 5. Pearson. “Tables of Percentage Points of Vb. and b: in Normal Samples; a Rounding Otl'." Binmeln’ka. Vol. 52 (1965). pp. 282-85 and Ralph B. D'Agostino and Gary L. Tietjen. “Simulation Proba- bility Points of b: for Small Samples," Biomem'ka Vol. 58 (1971), pp, 669-72. The last reference contains data for values of n from 7 to 50 additional to those listed in Table G. Also see Ralph D'Agostino and E. S. Pearson. "Tests for Departure from Normality. Empirical Results for the Distributions of b, and V7)...” Biometrika. Vol. 60 (1973). pp. 613-22. for charts of the percentage points of the distribution of b, = y, + 3. Since the pub- lication of this paper. Prof. F. J. Anscombe has pointed out that b, and VS. are no! inde- pendent in samples from a normal universe. Hence the joint tests suggested in Section 6 of the paper are not correct. though they might be nearly so. [Letters from E. 5. Pearson. April 4. 1974.] For a discussion of the operating characters ofa 7, test for normality. see K. O. Bowman. “Power of the Kurtosis Statistic. b, in Tests of Departure from Normality." Binmetrika. Vol. 60 (1973). pp. 623—28. APPENDIX C 162 RLISA SaiLU DATA :ILE NUNAM? (CREATION ULlE 3 '7/Et/c3 ) SJBFILL obi 335 CC); _ 7n, ’"“*““ “M“**‘ 1 5e ** 1 1) E Sbnat MPH 9. tttxtt ( 5) i “1'45 MFH 10. *tttttiatttittir ( 15) i 4b~bu MPH 11. utxutrttxtuattakttxraeatxrasxattaaaat ( 36) i 51~ss MPH 12. tittltiatttrttllttttxtiritxt ( 77) i Sb-bi MPH 13. tttitistttttttxttttttatria!txtttttrtturtt’tt ( 43) § blnoi MPH 19. attrittttttrtttettttttrt ( 23) § bb’7fi MPH 15. utwtuattttutatanaeexxtxrt ( 24) { 7;M75 hPh 15. artaxtaatttxt ( 12) % 76‘50 "PH 17. ttxrxxittt ( 3) l Bluoi MlH 13. :ttttttt ( 7) E 56'92 MPH 190 ‘** ( 2’ Z 91~95 MPH 1 20¢ ** ( 1) § 96-lu6 ”Pd 7 1 - IOOOODOOOOLOOOOCOOOC‘Itdtbds0.0200000000610060000001 4 1; LL 3; 9. r. ihLUULNCY ”‘ JALIL casts zus wrsszws cases b 163 ALISA SPEED DATA FILE NONAME (CREAYION UAIE 3 01/19/64 1 SUBFILE GBL‘ COBS CODE 9. 10. 11. 12. 13. 1h. 15. 16. 17. 18. VALID #4#4J§ I 41-95 MPH 4d-a‘l49' (nether-p304 b ( 15) I 46-56 NPH ot4¢o4bb¥bo om ‘h‘409#~w#~00¥#96¥45§bb¢ ( 55) I 51-55 MPH 44444444¥4oodeea~¥4oo44§§§oot ( 27) I Eé-bt MPH ¥¥GOGO¥¥¥“¥Ufii¥§‘¢‘"¥¥¥¥l¥¥¥§¥I6¥“§¥4¥¥'§ ‘ 93) I ol-ofi HPH 4m4ce-‘I-I- woe-r5. ”94.-.5qu 06-F- v“ ( 23) % 66-70 MPH 44¢¥¥444¥4¢4ee4~4candidature ( 21,) 1 71-75 MPH :e-A-o4-r-q-‘ammvm ( 1.)) I 76-8L HPN 444¢JJ¢+AHM ( g) I 81-85 MPH 444- ‘ 2) § 86-9U MPH IIIDOODOIOIIOOOOI0.010.000.0001OIIOOOOOOIOOOIOIOO 30 94 -L FREQUENCY CAS£S19b "ISSIHG CASES 0 APPENDIX D : cocoa. . z captu—u—Luucu zo-nmuxwui auu~.¢rzafza u:» an v_eh«: uu:¢~:4>ou\uu. .m4¢>¢u —z_ nuruaqmeou nee w—zu-u—mmucu M 532:3 4.: .: a: 3.513.. a: 1 . . mgmnocnm— ”momnpnm. aquamaoun .—:<_mzou. coco“. e nmnnonno emo¢_noao mange—~ooooms Naiamwatnonm. : uuz¢u_u_29_m puuu—pu44u uu=¢u_u—=u_n 000.00.00-00 0.....0000 000000000000 5 uuzczu4¢_ 4<_—¢ 00201000100000.000000 pus can" ph-a_a¢_¢«> do umucu m-m~.a :.c- >3: m manna: manna a. .hgm.“ “on... . Wasp... S . I 3 «2.3.25; u. 2.33 53.. 2:33 .3 .5... a: 321...... a; 23.3.“... 2.5». .. axe—32.“ 3:2: ._ .2. 5.3:... 4...; .5 3.3.3.. 3.. 3.32:; 3.3:... £5 .3.» 3.3.3... 3.3.5.... :3... 24“: ..wa=¢_x<: bzuazdmda do . nae-co do omth-u~nbucu ro—mwbruac aux—agxrezra wt» we v—xp<= uu:¢_cc>ou\uuz¢—¢¢> ”Q n.u~=..~m . ._.on . ~m.-mo._c .¢.~mc.. onoana.er azapwzou .namum.. . . ~.o.oo. an .mnnm...a- anomoao. waca.n.u.n «a ..s...z. dozen—axon .us a.m. a a gee.“ can a ca=._¢.. .nagsuu.p. unruo.tzeu oz. mazc.u.aauou .zo~..:oo map a. cg. nuauaagas ad. «www.9waa .w.~n..._ one_n..mo .cr..»zou. ..¢.~.. .a :J . ...-...a ..¢o.~°.ea aeoooooa. m.o~.r.o.- .o no uuzao...ze_m caau..».au uozao.a_=u_n .....'.-.'.. '...-..-.. ...-. .‘0'..- 1; a uag...ae. a...c.a Edna...» can. a e .o¢.u a.“ a uamaaca» ........u.o ze.a.:ea exp 2. ac: nude....» .u--.-.... .--u--iin- n- . zeapcaou oz. 2. neac..¢.. ...--.- . ,. . an a. .p..a_c...., to annoy spree.» zo.aa.»eo can $3“... ”.32ch . . "a 5.3% ”we... . “mama.“ a . o wuz.u_a.:e_m a m«.=oa 2.»: w. :am no can so aux..¢.. ac u mpaaza . cam. . oaeacaaz ugaaasgsu ac uuxcua .e .._ «Ecz:2 aucm a: cacu.zu .u.u4:._... moan... ..u: .9.” mo~w.._. ouzuamux zaur 24w, ooUJQc-k.» azuozuawo 166 sauce. x on—zu—u—uuwou zo—mwucwus :5~—4qtme:2= u:» kc r_u—<= uu:¢—z<>¢u\uuz<_¢cuwz~ uutUD—bzou but coho » ua=<—¢4> .md<>¢U~2_ uuzua—uaou 92‘ m»:u~0—uhuou .ro—uraeu u2> z— U¢< muam<~¢¢> JJ< o n—nwe.noo ~w.~ocn~r 1n~m~wmom -»z<»m:ou. Opswoc-mh celu—w—nneflo. watumwo~uwsmo z wu: L a cert. can u u4= I ZO—>¢=°U U1» 2— ho: nuani-IC’ lollilllll IIIIIIOIICIIIIIOIIIUIO xO—>I=CU Urn 2— nw4¢¢_x<> II hum ro—u p»—d—m<—g<> to uuucu numono— zo—n<—>un ohm envenow ochww- 0” can a :D—mwx Mnnmm' u¢4<2< Mansu z uaa—~4=r m=_ac¢ u so— avatar Lu—m zc awxmbzu aanJmu_¢‘: o—Bnnum chum oahm mmrm=uh untommuz 215: :c—>¢_>u= ascoztum >uc~m .uu4c¢_x<> sauo:uaua uuz¢u.z. ocean—caau sea m-zu.u_aauou . .za__¢:au u:_ z. aca auao..¢¢> 44. o~..~. m~a.a.~.. «noononn. apanoaogo .pzapmzoo. .uo..oo.- o~ooa_.am actuaoao.~¢~. ~.oo...~.- an bbuu—hmdau uuz.u-u—zo—a 0000000000 0000llkelll0 uaa¢-¢«> «bus 5 a zaazu ab» = uJ=4—¢<> OOOUOIOOOII ZO—bdaau Utb 2. be: Oustdudd) OIOIIIIIOI h llllllll it t 167 u08 no amuau enama.a 23.—«.aua a.» nmpnp..o own Jana—nu: anppo. aza: =;«=:<_m .us.“ .ouaaa_z<: ~=uczaaaa APPENDIX E J ___..- hh... 168 on cacao. x omwzuuo~muucu zo-uwucadc au-4¢t¢ozza u!» an v—xpaz uu:¢—¢<)09sau:¢~¢¢> nonoowooo aoooon. . oofihou.p. .oncennnenan. uownmnwn. .za.m:=u coaueoouoon..c .o°.uouonooooou on occa.no nacaeun~a.n». .n.a~w~m.-n.- z . aa>¢9pz_ hogan—azou pus a.m. . a «ozxa up“ a u4=<.¢.. .m4.>xu_.. uu.ao_azou a=< m.zc.u.aaueu .=o__.:aa u:— a. 4“ auaa..¢.. 44¢ «cwe_.-n actuammmanm~. .mumam~m. ..:..azou. gnaw... a: . panama... ~.o.a.... n=cue~nwnann. .aiuwpwa.-n.o a cu:.u.a.2u.m . »p_u_.madu muzauma.=a_m a cuzazuaep a...¢¢c uaaa.¢¢> «.uu . a .c .e¢¢a a.» m ua=._¢¢, ous.......t :o..¢:ea .2. :— _ez ”mama—c.» illicit-.. . i .u. ze...:ou a:. z. muaa._~a, c.----.uo.-.iaaaiuuiu . boa n.~n p..a_m._¢., .o acuau en.~_. zo__q.,ao a.» . . ..ano. mace... .w. a.g:.mum nawww. axaasm c copmasaa coo .oo..a ,— moeno. .;=n.. .. z¢_mma¢uuz o n.~. Hacaam x uu..«u—:za_w . 2.125 :cux "33.5w no 5.5 m9 u0=<:.§> a... w—m»..4:« c—Tc. a uJ.._-J:z »:.oa¢ : ... 23:92. aw_a 29 aged—:4 .m.u4:¢.¢<» .oaa_. .>.= .zaw an=m,. .mzuemug zaa: mm :33; ~33 332...: .2253; 169 O no vacu—u—za—s uuzcu—u—zc-m cuzqguaop .o...ouauui zo._¢:au u:_ 2. .oz magma—«c; ....o.ou-. qncmn.~ u on. own o pno pb—uubncau «van was soon u¢:»1>¢:u a: muzuao anne uuz<0uu—zu~w a: o,gs no.2)uzm oub8u~u-uuueu zo—mmucuea :u-a¢t¢o=?: u:— no w—cpgs wu:<-xc»auxuu2¢~c¢, 5:3 uzuaumzou .pus eons .esum as. ou.. aauumumn.~.c. ~aoua~one.~n. a zocxw cum 0000000000000000000000 ZC<>CQOU NIP Z- WMJQ‘—2(> uu3w9_u‘ou oz. a—zg-u—uuuao .rouptaam m:— z— wag ”aqua—24> 44¢ untuuroc—smeo «wataz aubm z? Guam-24 .w.uJ:4~¢<’ uazoanur I‘m: ooHJQa—x<> pzaazuuua .cbw 1'7() ~.ca~ . . «a. . «one - .w a.m-nm .mm.. p - n c. —:._mzeu .e..m¢.nnn.m. .uo.u-~nnmmw.o ewpww~m.. ma.um~.~ne.n. ~n--w~w~m».. ‘a naaupmmunoam.. .oouuooooeao .- o‘eocun.~u maauomomooo,. .ao4«-m¢_~u.o ¢ a<»¢~—z. buzuo.uzou .9; con. . p a zoczu c.n a ua=..¢¢. . .a4¢»¢u.z_ uurua.*‘ou az« m.z».o..‘uau _~ .za_.¢_:axoo cut—3:3 co. .:._u_m5:m2_ 4u>ud.uuz¢¢.aop c: 4.’uasu e nnmwnn..a «ouuwnmudmoo. ”wan—n... ..=<.mzau. oo.oe. mm . oosnpguo ”moonn~.. aqua—wonno.n. ~e-u....~eco. 4a ,nm. , «aflo..u coo. «onuan~_o.~. ~mnm.. ~o.oe.a u; on one... oooo.n~.= mouucuooaooa. oanu~.~»¢«~«.: x uuz.u.m_=a.m pp.u..u¢au uuz¢u_u_zw_m . . uuzcuuae. 4..»ccs uag..¢¢> «pug u a «gas. a.” a .4=<_¢¢> .--.-'-.--' ze—h‘qdflu U?’ I. DO: ”Uda‘-‘.’ .‘I-II'..I . I |l lib “I'I 30~P(J°U fit“ 2‘ MUJM‘-M‘, I.--O.I..-'I-'-'--." .us o.~ »~.4_=<_¢‘> .o auuou ~..~.. =o_—« >.= o_m . paano. -e.n._ .n 4¢=c_mux pennm’ Ja‘nsa x ”upwaaac _ee. _~w.n;; ~a~r~. oouon. .u ==_mmu¢uux o-.~. u¢.:cm x yucca—“.zu.m . 43.32“ z.ux mm¢.:cm .o can u: uuz<_¢.> &= a.m.4.=¢ arena. ; gag—paax ~¢:_<>a:u no “Jayme ‘= ..~ zua=22 awhm g: swam—‘6 .m.44:¢.¢‘: OOOOOOOICOOOOCOC.0.0.000...0.00.00.00.00...COCOCCCOICOCOCOOCIOI... 1'721 onvnzcngg ..a_.ap... nun: .xn¢zmuu an»: anacegnn . .omauu ..o. nag. ._.... ¢.-.»orn.«. «squane 0; «.us xcapn. ... an.» ..a.c. nocpuna. acrdutr. D uqnu >2)P9 > an on “:8 Q 06>:nw 8a.! Devan m anal—m—flbznn “an”. 2.: €me "mafia . :. M. .. raw» u an .... nae ona .n.wa u o . nomqm a. «.u..o.r... won. son .. . . . 14 1 <’,-DC'-H“ .8 “I“ HIP—"~08 114 1 4| lllt 000......I ‘..-’DPH“ ‘9‘ H: ‘3” MDC'-°: -‘I.I-CUI.. ¢.-..orn a . «.9 nuaoa u a can» «pa->ov- w»...»r gorna>znn m -IJ.O|.-."I .0..’.-... o .|.--.----.- a..z_q.a.aon nr..g.a_.. u.»z.~.n.xnn Duos I — oeucoo.0° we NOMOOG u on. a. I O“. o o d J a no wvd oo‘ ”9. onooomu .eozuq.z.. ........o .u.....o.nne. «9......» >PP cpa—pornw tun n: qzn nattq—czo newum.n.~gqu .ze “egg—enaon .zqnu‘oru. «.u.»ern a «.o «gun: a on.. so. aeyq.eu¢nn _z.«a¢.p ‘ any... 452%-: umszmfixgmaz Juana...“ Jung"-.. c»a.»znn\no¢.1_»znn :._a.. om dam czzoaa.r.~ne gnu-nunao: nonnaununaqo. anon .eeeoa unc- 1'722 soa4o cacao. sauce 0 Gupta-Unusuou zo—naucwuc au-aou\uuz¢-¢¢> 3.53325" r..&m§n. ":3”: “Emma" 23.2232” 3......“ a¢»¢upn~ ocean-s3.» .9; o... » o.¢o¢¢u on» a u4u¢.¢¢» .a4¢.¢u—2. uuzua.m¢oo 33¢ np¢u.u.m;uou oto-¢aou us» c. ucs muao<~¢<, Jd< . a o . nmmmn...~ ~o.uo.a~e~.n. agape—n.. ._z<.azoo. ~moun-ao mmuuow... oeaUQ~opa~o.. nacump~aono.s coo¢a uuacu.5.ao.n .p.u.».¢mu uuaau.g.zm.o -na--.u-.--- uua4guae. 4...¢«. uaa..... . «au- 5 “- a .occu a.» . uaa«..«, nuanuouuuua ae.-c:au mg. 2. be: ouao¢~ug> annotoau-. . p- . ao_-<:ou u:» :. .udac.¢<> -’ u y u... p..~... n»uwu~ «muuau. .pu‘ .mmu .—.4.a‘..¢“=“«=mw m“ “mmmum “3.:um_u.mmmmummm aux.u.u.zu.a u u..:a. a¢ux mu¢«:a. mo 2:» we uuz¢_¢‘. ‘o mwmwawm« coma“. . “4;.paas =ua«=oo u¢:..>¢=o go uuceua cow.a ... gusts: .upm so oagu~zu ...uaag.g«. wooa.. .3ua .a—n ansomu wazomauz z.w: amazau;m :ugm ..u4a..¢¢, pxaazusua 31293000852487 .*