THE WABEUTY OF THE TRAFHC CONFLICTS TEMMTOTHEURBM OFF- WSITUAW Thesis for the Degree of M. & MICHIGAN STATE UNWERSITY DONALD fiAMES MERCER, P. E. 1974 A B S T R A C T THE ADAPTABILITY OF THE TRAFFIC CONFLICTS TECHNIQUE To THE URBAN OFF-RAMP SITUATION by Donald James Mercer, P.E. The Traffic Conflict Technique, developed to study the operation of intersections, is expanded and tested at six off ramps, three of parallel design and three of taper design. Nine types of conflicts are tabulated over a total of ninety minutes at each ramp, using video tape. The mean numbers of total conflicts per minute is compared to the accident rates at those ramps. The Traffic Conflict Technique is found to be 9a percent reliable in measuring the quality of operation of urban off ramps. The conflict rate is.found to increase as the flow increases. No differences in operation are found between parallel and taper ramps. A procedure for conducting future conflict measurements at urban off ramps is presented. THE ADAPTABILITY OF THE TRAFFIC CONFLICTS TECHNIQUE TO THE URBAN OFF-RAMP SITUATION By Donald James Mercer, P.E. A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Civil Engineering 1974 a? ‘ A C K N 0 W L E D G E M E N T L Thanks are due to the Michigan Department of State Highways and Transportation for providing the data used in this project and for the use of its data processing equipment to help reduce the data. Many employees of the Department were consulted in the progression of this work. Each willingly contributed all that was asked of him and more, the net result being a far better paper than otherwise would have been written; if one could have been written at all. Mr. Gail Blomquist has been most helpful to me on this project as well as all my work at Michigan State University. Special thanks go to the one who has been my prime encourager, my editor, my typist, my proofreader, and, temporarily, my widow on this project: my beloved wife Mary JoAnn. ii Chapter 1 Chapter 2 Chapter 3 Chapter h Chapter 5 Chapter 6 Chapter 7 Appendix 1 T A B E N T E N T S LIST OF FIGURES o o o o o o o o o o o INTRODUCTION 0 o a o o o o o o . . o o Th9818 o o o o o o o o o o o o o 0 o 0 Background 0 o o o o o o o o o o o o 0 LITERATURE REVIEW 0 o o o o o o o o o PROCE DURE USED 0 O O O O O 0 Traffic Conflicts Technique . . . . . Data Collection Technique . . . . . . StatLStical AHEIYSIS o o o o o o o o 0 STUDY SITES 0 o o o o o o o o o o o a 83818 for Selection. 0 o o o o o o o a Parallel Ramps o o o o o o o o o o o o Taper Ramps o o o c o o o o o o o o o RESUI’TS . O O C O . C O C C O C C O . Conflict/Accident Correlation . . . . Conflict Type/Accident Type Comparison Taper Ramps/Parallel Ramps Comparison Conflicts as a Function of Flow . . . CONCIJUSIONS O O O O C O O O O O O O O RECOMMENDED PROCEDURE 0 o o . o o o o CONFLICT AND FLOW DATA . o o o o o . . Appendix 2 ACCIDENT DATA 0 o o o o o o o o o o 0 Appendix 3 DATA COLLECTION FORM . . . . o o o . . REFERENCES 0 O 0 O O O O O O O O O O 0 iii 52 53 Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 4: oowmm 9 L I_S T O F F I G U R Diagrams of Conflict Types . . Alignments of Parallel Ramps . Alignments of Taper Ramps Linear Regression Plots for Acc Rate = f(Conflict Data Points for Site P.1 . Data Peints Data Points Data Points Data Points 10 Data Points for Site P.2 for Site P.3 for Site T.l for Site T.2 for Site T03 iv 19 20 23 26 26 27 27 28 28 CHAPTER 1 I N T R O D U C T I O N THESIS The following thesis is proposed: The Traffic Conflicts Technique developed by Perkins and Harris to measure accident potential at intersections can be adapted to measure the quality of operation of urban off ramps. If so, the technique can be used to detect any signif- icant differences between the operations of off ramps of the parallel design and those of taper design. This thesis is tested in this work by tabulating the conflicts by number and by type, that occurred at six different urban off ramps. Three of these ramps were of parallel design and three were of taper design. The two-year accident experiences of the ramps are calculated, in Accidents per lOO-Million Vehicles, on the premise that the accident rate is symptomatic of the quality of operation. The goal of this work is to find a meaningful relationship, if there is any, between the conflict rates and the accident rates at these six ramps. In addition, the data are analyzed to determine any relationships that exist between the various traffic flow rates and the rate of conflicts. BACKQBOUND The question of which type of off-ramp design best serves the driver has been debated among highway engineers for a number'of years: it is currently being posed by the Michigan Department of State Highways and Transportation. The parallel design, which has a short added lane before the gore, provides abundant deceleration length off the thru lanes, but the added pavement area can induce erratic movements. The taper design, which leaves the thru lanes directly, forces the driver into a stereo- type path: but it also provides him with a small target and may cause him to slow excessively on the thru lanes. Numerous studies of the two types of ramps have generally favored the taper design, based on the path that is followed by the driver. Those studies (see Chapter 2) found that at parallel ramps the driver tended to follow a long flat taper rather than drive the path presented by the pavement. So, it is usually concluded, it is better to provide the driver with the path he wishes to drive. Accident data has not shown a significant difference between the two types. But accident data may be too insensitive to detect subtle changes in driver performance that may result from the difference in the ramps. The pathmdriven argument is discounted by those who favor the parallel design. They argue that the greater target presented to the driver by the parallel ramp is of more importance. They feel that the small target and the short distance to maneuver provided by a taper ramp leads to erratic movements in the thru lane by the exiting driver. To help resolve these differences, the Michigan Department of State Highways and Transportation is conducting a study of the operations of the two types of urban off ramps. The thesis presented in this work was developed during the conduct of the Department's study. It was determined that the thesis would be tested by this work; the results of this work and the data used will be used by the Department as one aspect of its study. CHAPTER 2 L I T E R A T U R E R E V I E W Being that off ramps are a major feature of a limited-access high- way and are apt to create friction in the traffic stream, numerous studies have been conducted to find the optimum design of some aspect of off ramps. Seven have attacked the taper versus parallel design question. The first of these was Conklin's study of two rural off ramps in Oregon (1). One of these was a taper ramp with a deflection of 4010', providing a 530 foot opening, reduced to 280 feet by paint lines. The other was a 470 foot parallel ramp that was followed by a 138 foot radius curve. Conklin measured speed and placement of exiting vehicles. He found a 22 mph reduction (45.5 to 23.5 mph) at the parallel ramp and a 3 mph reduction (49 to 46 mph) at the taper ramp. On lateral place- ment, he found that nearly all exit vehicles were off the thru lane at the midpoint of the taper ramp, compared to less than half for the parallel ramp. Only 20 percent of the exiting vehicles at the parallel ramp left the thru lanes in the first 200 feet. From this, Conklin concluded that the taper ramp was "definitely superior" (l, p. 16) to the parallel ramp, both in speed of operation and placement of vehicles. It should be noted that he actually compared a normal taper ramp to a substandard parallel ramp, and his work compared not only the ramps up to the gore but also the curvature beyond the gore. Pinnell and Keese (l3) studied ten ramps, both on and off ramps, in Texas. Their work concentrated on the on ramps, but they pointed h out that at a parallel ramp 5 percent of the vehicles used the ramp as designed, 35 percent followed a direct taper path, and the remaining 60 percent made a delayed move onto the ramp. They concluded that "This lack of usage [of the added lane] is related to the exit ramp driver's desire to follow a natural and easy path." (l), p. 57) Fisher (1) evaluated the accident experience and operation of New Jersey freeways. He found that "nearly all drivers will use a deceleration lane of the parallel type if it is 1200 feet long. When the length is decreased to less than 800 feet, some drivers will not use them and the accident rate is increased" (2, p. 130). The three taper ramps he studied were extremely short, not typical of current design practices. Jouzy and Michael (5) studied speed and placement of vehicles on several designs of on and off ramps in Indiana. All off ramps were tapers. They found that drivers began to decelerate on the thru lanes more than 1000 feet in advance of the ramp. They theorize that drivers ”desired to follow a natural straight path of exiting with a minimum of maneuvering" (5, p. 51). They also observed that ramps with almost identical geometries had different patterns of vehicle behavior. They favored a lZOO-foot taper. Lind and Hong (é) analyzed the accident experience on Milwaukee's expressways. Concerning off ramps, all of which were tapers, they found no correlation between the design and any type of accident (é, p. #4). They also noted "Drivers appeared to decelerate slightly on thru lanes and will not move over to the deceleration lane until they have a good view of the ramp nose or exit ramp, or both"(§, p. an). Davis and Williams measured headways, speeds, lateral placement, and deceleration rates for six parallel ramps in Toronto (g). They found that the drivers entered the ramp early and followed a long taper path and that drivers were not clearing the thru lane before decelerating. Therefore, ”A direct-taper type of exit would seem to be indicated since it would appear to fit the vehicle paths better than the taper plus added parallel deceleration lane" (g, p. 73). Mercer completed a study of rural off ramps in Michigan (Q), comparing driver behavior at parallel and taper ramps on the basis of speed reduction, path driven, and accident rates. He found speed reductions of about 7 mph at four parallel ramps and between 9 and 13 mph at three taper ramps; the difference at the taper ramps was significantly greater. That study produced more evidence that the parallel ramps are driven with a taper path. There was no difference found between the accident rates at the two types of ramps. Other studies on ramp design included Fukutcme and Moskewitz (fl) work in California, intended to determine the optimum length of tangent. They found that exiting vehicles began to decelerate 135 to 220 feet ahead of the beginning of the ramp. If there is a surplus deceleration distance, drivers maintained their speed for the first part of the ramp, then decelerated. Taylor concluded that "The direct-taper deceleration lane is opera- tionally superior to the parallel-lane type" (l&, p. 22); based on a review of the same literature as discussed here, principally Conklin's and Jouzy and Michael's works. Taylor, in another aspect of his work, defined eight "erratic movements" at ramp gores (15, p. 3). In general, these ‘were not as sensitive as the canflicts developed for the Traffic Conflicts Technique. Pahl (2) found that thru vehicles approaching an off ramp tend to move to the left, then return to the right at the gore. Tipton, Carrell, and Pinnell (1.5) argue that "The fact that parallel deceleration lanes are not driven as constructed may not necessarily be a bad feature” (15, p. 12). The parallel lane, they feel, has an "advantage under high density conditions . . . ." because it can "offset undesirable geometric features . . . ." (l5, p. 12). Martin, Newman, and Johnson (2) add the comment that "Congestion, if present, will usually occur upstream of the off ramp due to lane changing and overloading of Lane 1 by vehicles desiring to use the off ramp." (2, p. 29) It is arguments such as presented in the last two references that are the basis of the feeling of some engineers that parallel ramps will provide better performance on urban freeways. The criteria generally used to compare the operations of the two types of ramps have been vehicle placement, speed, and accident data. But a satisfactory answer to the parallel versus taper question is yet to be found. The Traffic Conflicts Technique has been tested and proven worthy for several applications (lg) at intersections, thus it has been proposed to use that method to again study the urban off-ramp question. That criterion may be more sensitive to differences in operation than are the others and can be easily measured with the equipment on hand at most highway agencies. Before the technique can be used, it is necessary to first determine if it will produce meaningful results. It is for that purpose that this work has been undertaken. CHAPTER 3 P R O C E D U R E U S E D TRAFFIC COEELICTS TECHNIQUE The procedure used to obtain data for this work is an expansion of the Perkins-Harris technique for detecting conflicts at inter- sections. In the abstract of the Procedures Manual, Perkins wrote: The Traffic Conflicts Technique was developed . . . to be a measure of traffic accident potential. A Traffic Conflict occurs when a driver takes evasive action, brakes or weaves, to avoid a collision. The evasive action is evidenced by a brakelight indication or a lane change by the offended driver. (L§_p ii) Perkins and Harris found that ". . . a high level of association exists between the traffic conflict and reported accident frequencies. In particular, high accident frequencies are always associated with high conflict frequencies." (}_1_ p 22) Most of their work was done on intersections. One study was conducted at a freeway curve and exit area, for which they defined different conflicts than were used at intersections. (l2) These conflicts, with others added, were used in this work. (Figure 1) Slow Vehicle. A slow vehicle is one that appreciably slows by braking for no apparent reason as it approaches the off ramp. Such a braking indicates that the driver has lost confidence in his ability Slow Vehicle wrl::> Rear-End Conflict Wrong-Lane Exit TDD __ __\_ = __ <5 Wrong-Lane Congestion Wrong-Lane Weave Late-Exit Conflict _ (a Weave -———*" === Weave Conflict ._————-"" -= Dr 1“" — \_ D! Figure 1. Diagrams of Conflict Types 10 to negotiate the roadway immediately ahead at his current speed. Since all ramps studied were designed to accommodate an in-gear deceleration rate, this braking action is an erratic move. A Slow-Vehicle Conflict occurs when there is a brakelight indication and there is no external stimulus to warrant the braking. Whether the vehicle exited is also noted. Rear-End Conflict. A rear-end conflict is the situation of one vehicle appreciably slowing, resulting in a following vehicle braking to avoid collision. A Rear-End Conflict occurs when there is a brakelight indication on a following vehicle. There is only one Rear-End Conflict per incident, even if more than one following vehicle brakes. The lead vehicle is also recorded as a Slow Vehicle if the criteria listed above apply. Wrong-Lane Exit. A wrong-lane exit is an exit movement by a vehicle that begins in some lane other than the outside lane. The driver may have gotten into an inside lane for several reasons: he may have been confused, not realizing that his exit was so near: he may have gotten trapped, unable to make a safe weave into the outside lane: or he may have been attempting to pass the slower moving outside lane and exited from the inside lane deliberately. A Wrong-Lane Exit occurs when a vehicle makes a direct move from an inside lane to the ramp. Such a move may precipitate two other types of conflicts: Hrong-Lane Congestion. A Hrong-Lane Congestion conflict is the situation of a wrong-lane exit vehicle being unable to make his move 11 smoothly due to a vehicle in the outside lane and being forced to slow to allow the other vehicle to cross his intended path. A Wrong-Lane Congestion Conflict occurs when a Wrong-Lane Exit vehicle applies its brakes and allows another vehicle in the lane to its right to go ahead of it. Wrong-Lane Weave. A wrong-lane weave conflict is the situation of a wrong-lane exit vehicle crossing directly in front of another vehicle, causing that vehicle to apply its brakes to avoid collision. A Wrong-Lane Weave Conflict occurs when a Wrong-Lane Exit vehicle crosses a lane to his right and the first vehicle directly behind him applies its brakes. Late-Erit Conflict. A late-exit conflict is the situation of one vehicle entering the off-ramp upstream from another vehicle that is on the ramp, resulting in the second vehicle applying its brakes. A Late Exit Conflict occurs whenever one vehicle passes and then enters the exit ramp ahead of a second vehicle, resulting in a brake- light indication from the second vehicle. Weave. A Weave is a complete lane change to the left, either from the outside lane to an inside lane or from the ramp to the outside lane. Such moves are considered to be erratic moves, the result of a rear—end or slow vehicle conflict or confusion by the drivers. Lane changes to the right were not considered weaves. Such moves are commonplace near and upstream from the gore as the traffic redistrib- utes itself as a result of the vehicles lost at the ramp. A Weave occurs when a vehicle makes a complete lane change to the left, either from the outside lane to the inside lane or from the ramp to the outside lane. Such a move can precipitate another conflict: 12 Weave Conflict. A weave conflict is the situation of a Weave vehicle making its move so close in front of another vehicle that the second vehicle must brake to avoid a collision. A Weave Conflict occurs when a Weave, as defined above, occurs and the first following vehicle in the lane entered has a brakelight indication. Drift. A Drift is a partial lane change, in which the vehicle crosses partly into the adjacent lane or the ramp and then returns to its original lane. This is an erratic movement, that is expected primarily at those ramps at which the thru lanes are on a curve. A Drift occurs when a vehicle encroaches onto an adjacent lane or the ramp and then returns to its original lane. DATA COLQECTION TECHNIQQE Data for this work were obtained by use of video tape. A camera was set up on a structure, if possible, or on the slope and set to view the gore and at least 500 ft upstream from it. Due to equipment limitations, only 30 minutes could be taped at one time. At least three such tapes were taken at each of the six study sites: two during near peak periods and one during an off-peak period. The data were taken from the tapes in the office. Conflicts were recorded by type and by time of occurrence to the nearest 0.1 minute as measured by the meter on the tape deck. Volumes were also recorded for each minute in three different listings: exit volume, outside lane volume, which included the vehicles that exited, and the sum of all remaining lanes. In taking data from the first tapes at most sites, volume data were taken for only a ”typical" 10-minute 13 period. Thus, for most sites there were 90 minutes of conflict data and 70 minutes of volume data available for analysis. Accident data were obtained from the Department of State Highways and Transportation files for 1971 and 1972 for each ramp. It is difficult to ascertain from an accident report exactly what factors triggered the incident and where the incident actually began. So only general limits were used: all accidents that occurred along the ramp-thru lane interface or about 200 ft on either side of the interface were used, with no attempt to determine whether an exit maneuver was involved. Volume data were obtained from the Department of State Highways and Transportation.records for 1971: one-half of the two-way average daily traffic was used in calculating accident rates. STATISTICAL ANALYSIS The data from all three 30-minute observations were combined to form a data file for each study site. These files contained the following information for each individual minute of observation: 1. Traffic Flow a) for the inside lane(s) b) for the outside lane, including exiting vehicles c) for exiting vehicles d) total flow 2. Number of Occurrences of each type of conflict a) Slow Vehicle b) Rear-End Conflict c) Wrong-Lane Exit d) Late-Exit Conflict e) Weave f) Total Conflicts 14 Four types of conflicts occurred a total of no more than twice in all observations at all six sites, and were not included in the statistical analysis: Wrong-Lane Congestion, Wrong-Lane Weave, Weave Conflict, and Drift. The basic statistics (mean value per minute and standard deviation) were found for each of the ten items listed above. To test the first point of the thesis, various conflict rates and flow rates were calculated and compared to the accident rates at the six ramps (in Accidents per loo-Million Vehicles): 1. Mean number of Conflicts per minute 2. Number of Conflicts per exiting vehicle 3. Number of Conflicts per total flow 4. Number of Conflicts per average flow per thru lane 5. Mean number of conflicts per minute per proportion exiting 6. Mean number of conflicts per minute times the proportion exiting 7. Total flow in vehicles per minute 8. Exit flow in vehicles per minute 9. Outside Lane flow in vehicles per minute 10. Inside Lane flow in vehicles per minute The reliability of the first point of the thesis was determined by the significance level of the hypothesis: The slope of any of the above relationships is not zero. The slope of the relationship having the greatest potential was so tested, using the gestatistic. The slope of the regression equation for the taper rampstas similarly tested against the slope of the equation for the parallel ramps. The frequencies of occurrence of the different types of conflicts were compared to the frequencies of occurrence of the different types of accidents. 15 To determine if the rate of conflict per minute is_dependent on the various flow rates and to determine if the number of exiting vehicles per minute could be predicted by counting the thru flow, six potential linear relationships were investigated: Tbtal Conflict = f(thru flow, exit flow) Total Conflict a f(outside lane thru flow, exit flow) Total Conflict = f(exit flow) Total Conflict = f(total flow) Exit Flow = f(total flow) Exit Flow = f(outside lane flow) To compensate for the high frequency of minutes with 0 conflicts, the regression equations were calculated three times for each ramp at one-minute, two-minute, and three-minute increments. The multiple linear equations were tested against the null. hypothesis: either independent variable is not significant. The linear equations were tested against the null hypothesis: H is not different than zero. The minimum values for R to cause rejection of that null hypothesis are a function of the sample size: for the sample sizes used in this work, they are: Sample Size(min.) R _ R2 90 0.203 0.0u1 70 0.235 0.055 68 0.238 0.057 45 0.294 0.086 35 0.333 0.111 34 0.338 0.11u 30 0.360 0.130 23 0.1+12 0.170 22 0.422 0.178 16 Because the number of occurrences of each type of conflict was small, (most commonly 0, to a maximum of 6) there would be no meaning- ful correlation between individual types of conflicts and traffic flows. Because the total conflicts per minute was also low (no more than 12) and the ratio of conflicts to flow was also small (generally less than 1:10), it was expected that there would be low correlation between conflicts and the flow. Because the increase in the flow rate may, through increased congestion, breed additional conflicts, the relationship between conflicts and flow may be curvilinear. It cannot be exponential (Conflicts = B x (Flow)A) due to the high frequency of 0 conflicts per minute. So it was determined to compute a second-degree polynomial equation: Conflicts = Bo +-81 x (Flow) + B2 x (Flow) 2 This equation was viewed subjectively to determine if it appeared to be a better predictive equation than the linear equation. The computations for these analyses were performed by the Michigan Department of State Highways and Transportation's Burroughs 35700 computer. The basic statistics and linear regression were computed using the BASIS (Burroughs Advanced Statistical Inquiry System) package. A separate program was written by the author (in FORTRAN IV) to solve for the second-degree polynomial equations. CHAPTER 4 STUDY SITES BASIS FOR SELECTION Six different urban off ramps, three of parallel design and three of taper design, were studied in this work. The criteria for selection we re 8 l. 2. 3. There should be no unusual alignment on either the ramp or the thru lanes that could induce erratic behavior by the driver. There should be sufficient deceleration distance on the ramp to allow drivers to decelerate in gear to the ramp speed after they completely clear the thru lanes. There should be a point about 1000 ft upstream from the gore at which the video camera could be placed to give adequate coverage of the ramp-thru lane interface. During the observation period, there should be no congestion on either the ramp or the thru lanes that would result in conflicts other than those to be analyzed. As the design practices of the Department have evolved, all the ramps of the Detroit area freeway system are of the taper design and all suitable ramps in other Michigan metropolitan areas are of the parallel design. This introduces a factor into the data: the sites used 17 18 to represent the different types of ramps also represent different driving populations. PARALLEL RAMPS (Figure 2) P.1 I #96 EB to us 127 NB, Lansing P.2 I #96 EB to Walnut Street, Lansing P.3 I 496 WB to US 27 SB, Lansing TAPER RAMPS (Figure 3) T.l I 75 NB to 7 Mile Road, Detroit T.2 US 10 NB to Meyers Street, Detroit T.3 I 75 SB to M 39, Lincoln Park l9 pm getaways: . am 935 vm own. I one: o... _ _ .ooom I n _ L Home» .«0 flash no.3 \\\U\\\l \l“ I .3“ pa u a wing 35. RN mm 2 m: 3 ms. was H md 33 um smug pm owe u omo: on _ —.Hmnw.3. MO fwd“; sham >l r a. J am End: 3 mm mm: H md 33 .800 u a i one u 82 2 /0/ Honda we vhmym pma ennui .w Ergo SHE. gown; .03...H .35... m2 mm“ m: 3 mm 09 H «a «Sm Figure 2 Alignments of Parallel Ramps A; on: n mmo: 0» s83 mo Pusan pawn O om.z op mm vs M may maam we owe a owe: ow swam... mo 9.58% .53 .003 n n 1., 20 2520 FEB m Figure 3 ._.—OH I Q :6 Elm. am mamas? as mz o“ m: ~.e spam “(p Lvumx " mun/n-“ 9+ .505 mo .33.. WEB. .mdoo a Q 9E5. IBLH mcomm.afl boom u 0 misfit, mice Um $3; 5... on m2 we H fish wcwm Alignments of Taper Ramps CHAPTER 5 R E S U L T S The distribution of each type of conflict by frequency per minute, the various flow rates, and the predictive equations for conflicts and exits as a function of flow are given in Appendix 1. The two-year accident experience and collision diagram for each ramp is given in Appendix 2. ‘QQNFLICTAACCIDENT CORRELATION The accident rates are tabulated on page 50. For the ten methods of measuring the conflict rate, the following linear relationships were found: Acc/lOO MV = AX + B where X = Conflict rate Conflict Rate Mean Conflicts per min. 36.0435 Conflicts per exit 303.8470 Conflicts per Total Flow 656.9650 Confl/min/mean flow 367.3205 Confl/min/prop exit 0.7919 Confl/min x prop exit 51.1202 Total Flow per min. - 0.2899 Exit Flow per min. 3.1947 Outside Lane Flow per min. 1.0476 Inside Lane Flow per min. - 0.3945 21 8 A 1 {Slope} Lponstant) - 5.5680 10.3413 19.4583 12.2242 40.5401 19.9824 57.7708 6.4614 23.9197 55.0074 Sta-Dd 0 Error 19.875 29.937 27.838 21.746 33.335 22.082 26.326 23.911 26.756 25.793 0.803 OQMZ 0.552 0.603. 0.050 0.750 0.258 0.480 0.190 0.323 0.65 0.20 0.30 0.36 0.002 0.56 0.07 0.23 0.04 0.10 22 The level of significance for the slope of the first equation was calculated using the trtest; that test assumes that the sample comes from a normal population. The calculated value was t - 2.70, which corresponds to a level of significance of more than 94 percent (Appendix 2, p. 51). Because of the high correlation found for the sixth equation, the possibility that the proportion exiting is a significant factor was investigated by calculating a multiple linear regresSion line for the equation: Accident rate =f(Conflicts per minute, Proportion Exiting) From that calculation, it was found that the proportion exiting is not a significant factor. The easiest variable to measure are the various flow rates. For that reason, the regression lines for accident rates as a function of the per-minute flow rates were calculated; but those relationships were not sigmificant. The linear regression found above is based on the assumption that the independent variable (Conflicts per minute) is an absolute value and the dependent variable (Accidents per 100 MV) is an estimate, with a mean and variance and was calculated by minimizing the vertical distance between the data points and the regression line. In truth both sets of data are estimates; to account for this an orthogonal regression line was calculated. This method finds the line that minimizes the perpendicular distance from the line to each data point. The orthogonal linear regression line found was: Ace/100 MV = 55.8388 x (Confl per min) - 32.9073 23 Both regression lines are estimates of the true regression line between all six data points. The data points, both regression lines, and confidence interval for the simple linear regression are plotted in Figure 4. Accidents per ioo-Million Vehicles 120‘ 100- (n 9 on c? r: ‘P A) cP H Simple Linear Regression ..————.. w/ 95% Confidence Interval LEQEND Orthogonal Linear Regression if? 0.5 120 1.5 2.0 215 Conflicts per Minute FIGURE 4. Linear Regression Plots for Acc Rate = f(Conflict per minute) CONFLICT TYPEZACCippNT TYPE COMPARISON The data were further analyzed to determine if the types of conflicts that were observed were indicative of the types of accidents that occurred. For this portion of the study, the sum of the rates of Slow~Vehicle Conflicts, Bear-End Conflicts, and Late Exits were compared to the proportion of Rear-End accidents: and the sum of the rates of Wrong-Lane 24 and Weaves were compared to the proportion of Angle Accidents. None of the conflict types were considered to be indicative of run-off-roadway accidents. The data are tabulated below: Prop. of Conflicts Proportion of Accidents BEER. SV+BE+LE WL +-w Rear-End ggglg. Off-Road P.1 0.69 0.31 0.42 0.08 0.50 P.2 0.79 0.21 0.50 0.25 0.25 P.3 0.49 0.51 0.50 0.00 0.50 T.l 0.69 0.31 0.42 0.33 0.25 T.2 0.54 0.46 0.40 0.30 0.30 T.3 0.50 0.50 0.57 0.29 0.14 There are no correlations evident from that data. TAPER RAMPSZBARALLEL RAMPS COMPARISON The linear regression lines for the three parallel ramp data points and for the three taper ramp data points are: Standard Parallel: Error _§E_ Ace/100 MV e 52.73 x Conf/min. - 39.16 _ 6.91 0.96 Taper: Acc/lOO MV = 12.48 x Conf/min - 26.15 14.35 0.11 Because of the small number of data points, the difference in those two slopes cannot be shown to be significantly different (t - 1.13, while t.95 w/l DF = 12.7). One goal of this work is to determine if there is any difference between the quality of operation of Taper and Parallel design of urban off ramps. Since the basis for measuring that quality of operation is the accident rate at each ramp, it would be incorrect to attempt to use that same relationship to test those same six ramps. The operational 25 difference sought-after would be evident only if one type of ramp had significantly higher accident rates than did the other. That phenomenon did not occur at the six test ramps. The taper design ramps had a higher proportion of angle accidents: this may be due to the higher volumes at the taper ramps and so it is not at this time considered significant. CONELICTS AS A FUNCTION OF FLOW At four of the six study ramps, there was a significant relation-' ship between the rate of conflicts per minute and the flow rates. Using two- and three-minute increments generally produced slightly higher correlation coefficients. For each set of ramp data, the various slopes tended to remain constant in the three calculations. Although there are significant linear relationships between conflicts and flows, the correlation coefficients are low, meaning that the linear equation should not be used to predict the number of conflicts. The second-degree polynomial equations calculated generally produced a U-shape curve. This is the result of attempting curve-fitting on poorly related data: it does not indicate that there is an optimum.flow rate to achieve a minimum number of conflicts. The data points recorded and the linear and second-degree regression equations calculated for one-minute increments are shown in Figures 5 through 10. 26 N.m MBHm mvme mo Honszz ”Mb . . p a“. p . . “—NW b _ .O_N>1i4r_ - ma.“ _ ”OPH- W . W M w 0 D s 0‘0 coanmonwom .3234 . o o o o o o o o o o . o I m» M . . . o o o moefia mA\: movaOCwoo coammonmcm axw 1 m.1 o o o o 00.0 _- «84.0 .0 N»..— .I You . K «.m MHHm mwwxm mo honssz mm on mm on as as m 0 Fe + a L _ . . Ft‘ _ _ . . 4? w A». w 0 a W . . _ w w _ _ f rL _ . . O O O O T . n scammonwom “nomad o I m 1 . . r m \\\. o o la. 0 \\\ o o I Is \ O O .I m“. \\. I m. \ codenamed. my. U... . u o a:.. team. . ”P C I. NE W meosaoaa soc Ama.esv endear Figure 6 Data Points for Site P.2 5 Data Peints for Site P.1 c‘ 9 ”111" F -‘. 27 «.9 HRHm neaxm so nonesz mm on we on m o . I‘ll I-Illl I '0 scammonom “noes; . . . o . . o . mew scammwnmom Nx . m. J a. o . s I. o [m «c6 n im m.a meHm mpfixm mo Hopssz ma om ma 0H m o rofimmchmm “noggin . o . . . . _ m _ . 1 m... tn rcwmmmimom Nx.\\m.wv\n.\ . . . :I.’ . u. w. 0 0 O O I Ti“. at. 0 fi m.ho . rm sandwmwcnflu you wocHH coammoammm OOoC fl Nm Figure 8 Data Points for Site T.1 Figure 7 Data Points for Site P.3 28 m.a HHHm npaxm Mo Honsdz mm 0N ma 0a m 0 Fa L a - . . - P. - b b r P P b - LU . W V b p p O - - I J - - mum scammonuom Hammad . . . IhI Id . mom 0 O C O O I 1 TL m. I.1 o o o O I My... 0 cowmwonmem mx . o r s I. In pcmofiawcmfim pom mocha conmOHmOm 00.0 H Nm N.H MHHm mpaxm mo “635:2 mm om D 2 a o r b a — n W P W - 1r W “L." ” W k p “I“ 0 F P p . L C I m on Sandro Moog .PIoIIbItI J m. w a m .4}, o o o o o o r ImsLm \ \ 1L. 1 K O O . O r MW. 0 cofimmfimom Nx I s I. O O [m N~.o u Im Figure 10 Data Points for Site T.3 Figure 9 Data Points for Site T.2 CHAPTER 6 C O N C L U S I O N S Based on the results obtained in this work, it is concluded that the Traffic Conflicts Technicue is 94 percent reliable in measuring the quality of operation of urban off-ramps. The technique can therefore be used whenever a level of significance of 94 percent or less is acceptable. The linear equation for accidents as a function of Conflicts per Minute determined from this work is a poor predictor. Thus, while the results show that (at 94 percent significance) a ramp with a conflict rate greater than that of another ramp will have a higher accident rate, the equation will give only a poor estimate of the numerical values of those accident rates. This work found that generally the number of conflicts per minute increases as the various flow rates increase. The linear models for these relationships are poor predictors of the conflict rates, however. This work was unable to detect any difference in operation between the three parallel design ramps and the three taper design ramps that can be attributed to the design type. This work was also unable to detect any correlation between the frequencies of occurrence of the different types of conflicts and the types of accidents that occurred at the ramps. 29 CHAPTER 7 R E C O M M E N D E D P R O C E D U R E When the Traffic Conflicts Technique is used in future observations at urban off ramps, the following points should be considered: 1. If a significance level of 94 percent or less is acceptable, a significant difference in mean conflicts per minute can be regarded as evidence of a significant difference in the quality of operation. To establish a significant difference in means, longer testing periods should be used. The length needed depends on the mathematical difference in means, the variances of the samples, and the level of confidence desired. For this work, a typical variance found was 1.5. Using that value, the following sample sizes, in minutes, would be needed: Difference Confidence Level in Means 91$ 94% 95%g 99% 0.1 800 1060 1150 2200 0.2 200 270 300 500 0.3 90 120 130 220 0.4 50 7O 70 120 0.5 30 50 50 80 Once a significant difference is found, the investigator must determine what causative factors are involved. While such a difference might well be due to the drivers' ability to negotiate the two types of ramps, other factors must be 30 2.‘ 3. 31 considered, including horizontal and vertical alignments, signing, and volume/capacity ratio. A large number of sites may be needed to adequately compensate for the factors not being analyzed. The data collection form developed for this work proved adequate and can be used in future observations. This form is shown in Appendix 3. If the form is reprinted, a few minor changes are suggested. The column for feet can be eliminated. The initial intent was to mark the location of the conflict, but that proved to be impractical. The three columns for Slow- Vehicle can be reduced to two, so that only one column need be checked for each Slow-Vehicle Conflict. Actually, nearly all such conflicts involved an exiting vehicle. A third column for Weaves, from Lane 2 to Lane 3, would be helpful for freeways having more than two thru lanes. The use of video tape for obtaining data is recommended over the use of observers at the site. The video tape has two distinct advantages: it provides a more accurate count of the number of conflicts, and it provides the opportunity to review a sequence of events to determine exactly what conflicts occurred. Additional conflict and accident data, especially for ramps with either low or high conflict rates, may produce a higher level of significance and more representative regression line. In this work, four of the six ramps had conflict rates near the mean: this resulted in the wide range in the confidence interval shown in Figure 4. More values on the extremes of the conflict rates would narrow that interval. APPENDICES Conflict and Flow Data Test Dates: Test P Test P Test P Frequency Slow Per Min Vehiglg 0 49 1 23 2 12 3 4 4 0 S 2 6 7 R 9 10 ll 12 ”_ Totals 69 Mean 0.77 Stand. Ibv 1.08 Flow Rate Inside in vpm Lane Total 711 Mean 10.2 Stand Dev 6.5 32 SITE P.1 I 496 E8 to US 127 N8 Lansing 1 1 1.3 Wednesday, October 4, 1972 l 4 Friday, August 31, 1973 DISTRIRUTION OF CONFLICTS Tuesday, Septenber 12, 1972 3:30 - 4:00 pm 4:15 - 4:45 pm 4:15 - 4:45 pm Rear-End Wrong-Lane Late-Exit Total genflict Conflict Conflict Weave Confliqtg 50 43 80 83 16 18 36 8 5 22 l? 8 2 2 ll 5 2 15 3 l 8 1 10 1 4 2 0 0 0 1 ._ .1 .. _ ___1. 80 63 12 9 232 0.89 0.69 0.13 0.10 2.58 1.29 0.83 0.140 0037 2031‘ FLOW RATES 70 Minutes Outside Total Outside Total Lane Exits Terr}. Thru Flow 1694 1273 1132 421 2405 24.2 18.2 16.2 6.0 34.4 7.1 6.2 12.5 l 2 3 33 Site P.1 RECRESSION EQUATIONS all values per minute Total Conflicts =:f(Exits, Total Thru) min;* T Confl = 0.15 Ex + 0.076 T Th - 1.49 min -Neither variable significant- min) -Neither variable significant- Total Conflicts = fLExits, Outside Thru) min) -0utside Thru not significant- min) -0utside Thru not significant- min) - -0utside Thru not significant- Total Conflicts = f(Exits) min) T Confl = 0.195 x Ex - 1.08 min) T Confl = 0.204 x Ex - 1.18 min) T Confl = 0.218 x Ex - 1.47 Total Conflicts = fLTotsl Flow) min) T Confl = 0.104 x T Fl - 1.12 min) T Confl = 0.100 x T F1 - 0.87 min) T Confl = 0.103 x T Fl - 1.02 Total Conflicts = flexitsZ, Exits) min) T Confl = 0.0162 x Ex2 - 0.0476 x Ex + 5.13 Exits = fLTQtal Flow) min; Ex = 0.384 x T Fl +-4.98 min Ex = 0.379 x T F1 + 5.21 min) Ex = 0.380 x T Fl + 5.16 Exits = fLOutside Lane Flow min) Ex = 0.695 x 0L +-1.36 min) Ex = 0.776 x 0L - 0.59 min) Ex = 0.792 x 0L - 0.97 * Time increment used for computations Stand Err 2.12 3.85 2.69 2.10 3.70 2.28 1.59 34 SITE P.2 I 496 EB to Walnut St Lansing Test Dates: Test P.2.1 Thursday, September 21, 1972 7:25 - 7:55 am Test P.2.2 Wednesday, September 26, 1972 3:15 - 3:45 pm Test P.2.3 Wednesday, October 9, 1973 7:25 - 7:55 am DISTRIBUTION OF CONFLICTS Frequency Slow Rear-End Wrong-Lane Late-Exit Total Per Min vehicle Conflict Conflict Conflict Weave Conflicts 0 76 48 79 88 78 30 l 11 23 11 1 12 29 2 3 14 l 18 3 4 8 4 l 4 5 ._. __. __ __. __ .___ Totals 17 67 ll 3 12 110 Mean 0.19 0.74 0.12 0.03 0.13 1.22 Stand Dev 00“? 0095 0.33 0023 003“ 1.20 FLOW RATES 90 Minutes Flow Rate Inside Outside Total Outside Total in xpm' Lane Lane Exits Thru Thru Flow Total 1538 2025 1388 2175 817 3563 Mean 17009 22.50 15.42 24017 9.08 39059 Stand Dev 9090 9010 9.23 18005 (1 min)* (2 min) (3 min) (1 min) (2 min) (3 min) (1 min) (2 min) 3 min) 1 min 2 min 3 min (1 min) (1 ming (2 min (3 min) (2 min (3 min) 35 SITE P.2 REGRESSION EQUATIONS all values per minute Total Conflicts a f(Exits. Tota17Thru) -Tota1 Thru not significant- -Tota1 Thru not significant- -Tota1 Thru not significant- Total Conflicts a f(Exits, Outside Thrg)_ -0utside Thru not significant- T Confl = 0.0817 x Ex + 0.0754 x 0 Th - 1.14 T Confl = 0.0810 x Ex + 0.102 x 0 Th - 1.12 Total Conflicts = f(Exits) T Confl = 0.0809 x Ex - 0.03 T Confl = 0.0714 x Ex +-0.12 T Confl a 0.0723 x Ex + 0.11 Total Conflicts a f(Total Flow) T Confl = 0.0362 x T Fl - 0.21 T Confl = 0.0385 x T F1 - 0.30 T (:0an a 0.0388 x T F]. "' 003]. Total Conflicts = f(ExitsZL_Exits) 'T Confl a -0.0002 x Ex2 + 0.0879 Ex - 0.05 Exits = f(Tbtal Flow) Ex 3 0.466 x T Fl - 3.04 Ex = 0.469 x T F1 - 3.14 Ex = 0.475 x T Fl - 3.38 Exit§7= f(Outside Lane Flow) EX = 00959 X 0L " 6016 Ex = 0.986 x 0L - 6.76 EX 3 0.990 x 0L - 6.86 * Time increment used for computations Stand Err _§E_ 40 SITE T.2 US 10 NB to Meyers Detroit Test Dates: Test T.2.l Thursday, September 28, 1972 5:00 - 5:30 pm Test T.2.2 Thursday, September 20, 1973 2:00 - 2:30 pm Test T.2.3 Thursday, September 20, 1973 5:15 - 5:45 pm DISTRIBUTION OF CONFLICTS Frequency Slow Rear-End Wrong-Lane Late-Exit Total Par Min Vehicle Conflict Conflict Conflict Weave Conflicts 0 82 54 66 89 65 27 l 7 22 24 1 19 25 2 l 9 5 21 3 4 1 14 4 0 l 5 _ .1 __ __ _. _2 Totals 9 57 24 l 32 123 Than 0010 0.63 0027 0.01 0036 1037 Stand Dev 0.34 0.97 0.44 0.10 0.64 .22 FLOW RATES 70 Minutes Flow Rate Inside Outside Total Outside Total in gym Lanes Lane Exits Thru Thru Flow Total 3459 1714 923 7250 791 5173 Mean 49.41 24.49 13.19 60.71 11.30 73.90 Stand Dev 18.01 6.87 3.94 24.14 l 2 3 min min) min) min) min) min) min min min min) min min min min min min min;* 3? SITE P.3 REGRESSION EQUATIONS all values per minute Total Conflicts - fLExits. Total Thru) -Neither variable significant- -Neither variable significant- Total Conflicts . f(Exits. Outside Thru) ~Neither variable significant- -Neither variable significant- Total Conflicts - f(Exits) -R less than minimum for significance- -R2 less than minimum for significance- -R2 less than minimum for significance- Total Conflicts a f(Total Flow) -R2 less than minimum for significance- -R2 less than minimum for significance- -R2 less than minimum for significance- Total Conflicts = f(ExitsZ, Exits) T Confl - 0.0089 x Ex - 0.173 x Ex +-l.90 Exits = f(Tbtal Elgw) Ex - 0.173 x T Flow +-5.09 -R2 less than minimum for significance- Ex = 0.192 x T Flow 4'4.72 Exits = f(Outside Lane Flow) EX = 0.214 x 0L + 5.44 EX 3 0.289 X 0L +'4039 * Time increment used for computations Stand 0.00 0.03 0.05 0.02 0.0 0.06 U\ 38 SITE T.1 I 75 NB to 7-Mile Detroit Test Dates: Test T.1.2 Thursday, September 21, 1972 5:00 - 5:15 pm Test T.1.3 Wednesday, July 25, 1973 11:30 - 12:00 noon Test T.1.4 Tuesday, October 9, 1973 5:45 - 6:15 pm DISTRIBUTION OF CONFLICTS Frequency Slow Rear-End Wrong-Lane Late-Exit Total Per Min Vehicle Conflict Conflict Conflict Weave Conflicts O 83 69 85 9O 80 57 1 7 17 5 10 22 2 3 9 3 1 l a 0 5 __ __ __ __ _ .1. Tbtals 7 26 5 0 10 48 Mean 0.08 0.29 0.06 0.11 0.53 Stand mv 0.27 0.59 0.23 0.32 0.86 FLOW RATES 70 Minutes Flow Rate Inside Outside Tbtal Outside Total in ypm Lanes Lane Exits Thru Thrg__ __E;g§_ Total 4615 1348 609 5371 739 5980 Mean 65.93 19.26 8.70 76.73 10.56 85.43 Stand Dev 24.15 6.01 3.58 29.38 UNI-J M UNI-J (1 8 min)* min; min min min min min; min min) min) min) min) min) min min min min min (3 min 39 SITE T.1 REGRESSION EQUATIONS all values per minute Total Conflicts e f(Exits. Total Thru) -Exits not significant- -Exits not significant- -Neither significant- Total Conflicts = f(Exits, Outside Thru) T Confl = 0.0536 x Ex + 0.0671 x 0 Th - 0.70 -Exits not significant- -Outside Thru not significant- Total Conflicts = f(sxits3 T Confl = 0.0585 x Ex - 0.04 -R2 less than minimum for significance- T Confl = 0.148 x Ex - 0.82 Total Conflicts = f(Total Flow) T Confl = 0.0106 x T Fl - 0.44 T Confl = 0.0112 x T Fl - 0.48 T Confl = 0.0117 X T Fl - 0.53 Total Conflicts = f(ExitsA Exits) T Confl = 0.0052 x 8x2 - 0.0415 x Ex + 0.38 Exits = f(Total Elm Ex = 0.0584 x T Flow + 3.71 Ex = 0.0576 x T Flow + 3.78 Ex = 0.0568 x T Flow + 3.85 Exits = f(Outside Lane Flow) EX = 0.382 X 01.: + 1.35 Ex 0.354 x 0L + 1.88 Ex 0.330 X 0L + 2.33 * Time increment used for computations Stand Err 0.75 1.86 2.78 2.11 1.71 R2 0.18 141:. ACCIDENT MTA SITE P.1 COLLISION DIAGRAM IE" I/ EAST IL’—‘ ’496‘ [CAB—re \/ NOPTH ’6.) \OCIJVQU‘? 10 11 DETAILS OF ACCIDENTS Stated Reason Severity Light __ $1.; :I‘_i_m_e_ F PI PD Weather 9.92.9.2. __ 1971 9-18 0235 X Clear Dark 9-27 0730 X Rain Light 10-22 1710 X Rain Dusk 1972 4-19 1405 X Clear Light 5-13 1820 X Rain Light 6-14 0900 X Cloudy Light 6-29 1640 X Clear Light 7-22 2153 X Clear Dark 9-18 0730 Rain Light 10- 7 1150 X Clear Light 10— 7 1150 X Clear Light 10-30 1500 X Clear Light 12 Pavt for Accident Dry Avoid Vehicle Wet Too Close Wet Failure to Stop Dry Avoid Vehicle Wet Speed Dry Avoid Vehicle Dry Failure to Stop Dry Avoid Vehicle Wet Speed ’ Thy' Too Close Dry Failure to Stop Dry Speed 45 ACCIDENT DATA SITE P.2 COLLISION DIAGRAM E] EAST " “ 199 U') A. -_J\ C: a g o \ _J To Wain”, 3 DETAILS OF ACCIDENTS Acc Severity Light Stated Reason No. Date Time F PI PD Weather Cond. Pavt for Accident 1972 1 2-6 1005 X Snow Light Snow Speed 2 2-8 0802 X Clear Light Dry Failure to Stop 3 8-2 5 0040 X Clear Dark Dry Improper Lane Usage 4 11-30 0813 X Clear Light Dry Speed 46 ACCIDENT DATA SITE P.3 COLLISION DIAGRAM Clare St I 1% Hungerford Sf 1 / WEST €93 / ~\I TO V SOUTH DETAILS OF ACCIDENTS Acc Severity Light ' Stated Reason N9; ‘_Date Time F PI PD Weather Cond. Pavt for Accident 1971 1 3-2 2340 X Cloudy Dark Dry Avoid Vehicle Drinking 1972 2 8-22 1700 X Rain Light Net Failure to Stop a? ACCIDENT DATA SITE T.1 COLLISION DIAGRAM M99 rZ 9/ Q9 III-ia—9> ill [:1 ::><;: IIIJLwir—dr ‘\“‘\‘\\\\:TA/?QD l/e’pd DETAILS OF ACCIDENTS Acc Severity Light Stated Reason No. Date Time F PI PD Weather Cond. Pavt for Accident 1971 1 1-14 0337 Clear Dark Icy Failure to Stop 2 h-30 1750 Clear Dusk Dry Improper Lane Usage 3 9-20 0630 X Rain Light Net Lost Control, Too Close h 10-26 0012 X Clear Dawn Dry Failure to Stop 1972 5 1-11 10#5 X Clear Light Dry Improper Lane Usage 6 2-6 0400 X Snow Dark Snow Lost Control 7 10-9 0830 Clear Light Dry Improper Lane Usage 8 12-6 0930 X Cloudy Light Wet Lost Control 9 12-23 0225 Cloudy Dark Dry Failure to Stop 10 12-23 0230 X Cloudy Dark Dry Improper Lane Usage 11 12-23 0230 Cloudy Dark Dry Too Close 12 12-23 0230 Cloudy Dark Dry Failure to Stop 1+8 ACCIDENT DATA SITE T.2 COLLISION DIAGRAM 7FLJ:.£1_ [ES] 7‘” NOPTH .::Silz:<> III-419-——5£’1|I|n .L\% In -¥-a———uo [:11—a>—+- ‘\ / 4% ’0 4» DETAILS OF ACCIDENTS Acc Severity Light Stated Reason No. Date Time F PI PD Heather Cond. Pavt for Accident 1971 1 3-10 1230 X Rain Light Wet Lost Control 2 9-8 0700 X Clear Light Dry Lost Control 3 10-12 20U5 Clear Dark Dry Failure to Stop h 12-6 X Rain Dusk Net Tea Close 1972 5 6-16 0115 X Clear Dark Dry Lost Control 6 7-3 0010 X Rain Dark Wet Failure to Stop 7 10-29 1105 Rain Dark Wet Improper Lane Usage 8 10-29 0115 Rain Dark Wet Failure to Stop 9 10-29 0120 Rain Dark Net Failure to Stop 10 11-4 2200 X Cloudy Dark Dry Avoid vehicle 49 ACCIDENT DATA SITE T.3 COLLISION DIAGRAM I—‘l—H lI-La~—€>o B8 Acc tWNI—l 0(1)me 10 11 12 13 1h DETAILS OF ACCIDENTS Severity Light Stated Reason Date Time F PI PD Weather Cond. Pavt for Accident 1971 6-7 1530 Cloudy Light Dry Improper Lane Usage 6-11 2255 X Clear Dark Dry Improper Lane Usage 2-11 1595 Clear Light Dry Failure to Stop 12-7 1800 Clear Dark Dry Speed ' 1972 1-10 1300 Cloudy Light Dry Improper Backing 1-13 1&25 X Snow Light Snow Improper Lane Usage h-25 1535 X Clear Light Dry Speed 4-27 1730 X Clear Light Dry Speed 9-13 1730 X Cloudy Light Dry Too Close 9-20 0905 Clear Dark Dry Lost Control 10-24 17195 x Clear Dusk Dry Failure to Stop 11-20 1825 Cloudy Dark Dry. Avoid Vehicle 12—1 1930 Clear Light Dry Speed 12-26 0130 X Snow Dark Snow Failure to Stop 50 ACCIDENT RATES Accidents per loo-Million Vehicles (Ace in 2 years) x (100 000 000) Accident Rate =92 yr x 365 day/yr x (% Average Daily Traffic) _ (Ace in 2_years)_g — (Average Daily Traffic) X 273 973. M 1971 a: 1922 Accidents 1221 AD'I' Ace/100 NV 31. 12 3a 000 96.70 P.2 u 32 500 33.72 P.3 2 36 000 15.22 T.1 12 98 500 33.38 T.2 10 108 000 25.37 T.3 1h 63 000 60.88 51 TEST FOR LEVEL OF SIGNIFICANCE Accident Rate as a Function of Total Conflicts per Minute Regression Equation: Ace/100 MV = 36.0435 x (T Confl) - 5.5680 Null Hypothesis: Ho: b = b0 t309-19029: Si); Se 4' n 0‘ ll 36.0435 0 19.8750 13.2809 6 E3” flflflll 2.698 C‘- II The confidence level for t = 2.7 w/h DF is approximately 9h.3%. A P P E N D I X 3 Data Collection Form 'URBAN RAMP STUDY Location STATE OF MICHIGAN TRAFFIC A SAFETY D‘Vlf-.CP. ' __———.._.. . ...—...— Observcr __ live _TV_ 3 — r —- 1-.“ > -' r ‘ r- —'~~ -—-—- """-—v'-"~*'—"-'-~-——"— -—"' -.. 28.2.10» 19:19.. 9.....1 ”Fonfi’LaWe.iE2Lit.____. We nave ExiL End a ..a La Conflict Exit 193 anm;“~'<)n.‘1Drifq Z£m%pigiL-_l -CfieNfl .CDnilp. lit-gm... L‘uflngnhfév rarer La 21:51-11 --e...-_ .11.... 52 REFERENCES 1. 2. 3. 7. 8. 9. 10. R E F E R E N C E S Conklin, Robert D., "A Comparison of Vehicle Operating Character- istics Between Parallel Lane and Direct Taper Types of Freeway Off Ramps," Traffic Eggineerigg, Vol. 30, No. 3 (December 1959), Pp . 1.3-1? 0 Davis, Merritt M. and Williams, K. M., Vehicle ngratigg Character- istics on Outer Loop Deceleration Lanes of Interchagggs, Ontario Joint Highway Research Programme Number 43, University of Toronto, Toronto, Ontario, Canada (1968). Fisher, R. L., "Accident and Operating Experience at Interchanges,' Highwapresearch Board Bulletin 221, Washington, D.C. (1961), pp. 124-138. Fukutome, Ichiro and Moskowitz, Karl, ”Traffic Behavior and Off-Ramp Design," Hi hwa Research Record.2;, Highway Research Board, Washington, D.C. (1963 5, PP. 17‘310 Jouzy, Neddy C. and Michael, Harold L., "Use and Design of Accelera- tion and Deceleration Lanes in Indiana,” Hi hwa Research Rec rd , Highway Research Board, Washington, D.C. 219635, pp. 25-51. Lind, Bruce A. and Hong, Hyoungkey, ”Traffic Accident Study on Milwaukee Expressway," Journal of the Highwa Division of ASCE, Vol. 91, No. HHl, new York, new York (1965), pp. 25-u8. Martin, Darryl B., Newman, Leonard, and Johnson, Roger T., "Evaluation of Freeway Traffic Flow at Ramps, Collector Roads, and Lane Drops," Highwangesearch Record5432. Highway Research Board, Washington, D.C. (19735T‘pp. 25-31. Mercer, Donald J., Driyer;§§hayior at Rural Parallel and Ta r Ex t Ram s, Michigan Department of State Highways Report TSD~221-73, Lansing, Michigan (1973). Pahl, Juergen, ”Lane-Change Frequencies in Freeway Traffic Flow," Highwapresearch Record #09, National Academy of Sciences-National Research Council, Washington, D.C. (1972), pp. 17-25. Perkins, Stuard R. and Harris, Joseph I., Traffic Conflict Character- istics,_Freeway_Curve gnd Brit Area F1, General Motors Research Publication GMR-656, Warren, Michigan (1967). 53 59 ll. Perkins, Stuard R. and Harris, Joseph I., Traffic Conflict Character- istics. Accident Potential at Intersections, General Motors Research Publication GER-718, warren, Michigan (1968). 12. Perkins, Stuard R., GMR Traffic Conflicts Techni ue Procedures Manual, General Motors Research Publication OMB-895, Warren, Michigan (19695. 13. Pinnell, Charles and Reese, C.J., "Traffic Behavior and Freeway Ramp. Design," Journal of the Highwaprivision of ASCE. Vol. 86, No. HW3 (September 1960), New York, New York, pp. 41-58. 1“. Taylor, James I., Improving Traffic Operations and Safety at Exit Gore Areas, National Cooperative Highways Research Report 1&5, Highway Research Board, Washington, D.C. (1973). 15. Tiptcn, William E., Carrell, James D., and Pinnell, Charles, Effects of Off Ramps on Freeway Operations. Texas Transportation Institute Report 59-h, College Station, Texas (1965). O T H E R R E F E R E N C E S 16. Lundy, Richard A., "The Effect of Ramp Type and Geometry of Accidents," Highway_Research Record 163, Highway Research Board, Washington, D.C. (I967), pp. 80-119. 17. Miller, Irwin and Freund, John E., Probability and Statistics for En ineers, Prentice-Hall, Inc., Englewood, Cliffs, New Jersey (1965). 18. Paddock, Richard D., The Traffic Conflict§_Techniouex An Accident Prediction Method, Presented at Highway Research Board Annual Meeting, January 197u (Unpublished). 19. Unknown Author, Burroughs Advanced Statistical In ui S stem Reference Manual, Burroughs Corporation, Detroit, Michigan (1965). 20. Williams, James B., Olivetti Underwood Programma 101 Reference Manual, Statigtical Analysis, Olivetti Underwood Corp., New York, New York 1968 . 21. Wood, Donald L., "The Driver and the Road,” Trans rtation ineer n Journal of ASCE, Vol. 97. No. TEH, New York, New York (1971;, pp. 609-617. . H .jAI STATE UNIVERSITY L'BF II I III II; II III! V‘TV'V“ " 1r1‘ W“‘ --—.-_—-——