Geometric Categories as Intersection Safety Evaluation Tools John R. Campbell Ayres Associates N19 W24075 Riverwood Drive, Suite 300 Waukesha, WI 53188 campbellj@ayresassociates.com Keith K. Knapp Engineering Professional Development University of Wisconsin Madison 432 North Lake Street #713 Madison, WI 53706 knapp@epd.engr.wisc.edu ABSTRACT A number of operational and geometric factors impact the safety of at-grade intersections. Examples include the type of traffic control, the existence of left-turn lanes, and adequate sight distance. Determining whether an intersection of particular geometric design has unusual crash patterns could help identify those locations that need additional safety evaluation. During the last two years, the Wisconsin Department of Transportation and the University of Wisconsin- Madison created an intersection crash database segmented by area type (i.e., rural and urban), traffic control, traffic volumes, and general geometric categories. The overall database included crash report and entering volume information for more than 1,700 locations and 34,500 crashes. In addition, a series of 18 intersection categories were defined with respect to the number of intersection approach legs, the number of major road lanes, and the existence of a left-turn lane and/or a roadway median. More than 1,200 of the intersections were geometrically categorized, and the crash patterns at these summarized and evaluated. A portion of the results from the activities above is presented in this paper. More specifically, annual average crash frequencies and average crash rates are presented for intersections with varying operational and geometric characteristics. This integration of crash data and geometrics is not typically available for the general statewide safety management of intersections. These safety evaluation measures will be used to identify intersection locations of interest in Wisconsin, and are expected to be an invaluable resource for improving roadway safety in the state and possibly the region. Key words: crash rate geometry intersection safety Proceedings of the 2005 Mid-Continent Transportation Research Symposium, Ames, Iowa, August 2005. 2005 by Iowa State University. The contents of this paper reflect the views of the author(s), who are responsible for the facts and accuracy of the information presented herein.
PROBLEM STATEMENT Intersection crashes throughout the United States are a costly problem, both economically and in terms of the injuries and fatalities they produce. Limited public funds, however, require that the application of intersection safety improvements be efficient and effective. To reduce the number of intersection crashes, a systematic understanding of the intersection safety problem within a jurisdiction is needed. An understanding of the typical or expected intersection crash patterns can assist transportation professionals with identifying intersection safety problems. OBJECTIVE The objective of this paper is to provide intersection and crash rate statistics for intersections with varying operational and geometric characteristics. These statistics were calculated in the draft version of the Wisconsin intersection safety report (Knapp and Campbell 2004). The Wisconsin Department of Transportation (WisDOT) report is the first such safety summary for Wisconsin intersections and includes even more detailed intersection statistics than what is presented in this paper. It should assist WisDOT staff and other transportation professionals in their safety decision making process. The information should also help identify intersection locations of concern (e.g., locations with greater than average crash experience) and assist in the investigation of what might be the problem. RESEARCH METHODOLOGY The basis for the intersection statistics presented in this paper is three years (1998 to 2000) of intersection and intersection-related crash data from the WisDOT database. Only locations along the WisDOT state or connecting highway system (i.e., those sections of the state or United States highway system maintained by a local jurisdiction) were considered. In addition, only those rural intersection locations with at least three crashes in any one year from 1998 to 2000 were summarized. Urban intersections had to have at least five crashes in any one year. Urban intersections were located in incorporated areas, while rural intersections were located in unincorporated areas. The crash patterns at the intersection locations that met these minimum crash requirements are summarized in this paper. The intersection safety measures presented in this paper are from the draft version of the WisDOT intersection safety report in which crash data from the years 1998 to 2000 was analyzed. The final version of the report, which is scheduled for completion in July 2006, will be updated for the years 2001 to 2003. KEY FINDINGS Several intersection safety measures are presented in this paper. State highway intersection and intersection-related crash statistics (e.g., average annual and average crash rate) were calculated and summarized by area type (i.e., rural and urban), traffic control, and volume. Crash statistics were also calculated for intersections with different geometric characteristics, such as the number of approach legs, the existence of a median, the number of travel lanes on the major approaches, and the existence of left-turn lanes on the major approaches. Intersection Traffic Control and Volume Table 1 shows the average annual crash frequencies and average crash rates for the intersections in the database designated as rural and urban. The frequencies and rates presented are also grouped by intersection traffic control and by entering volumes, measured in vehicles per day (VPD). Crash rates Campbell, Knapp 2
were calculated per million entering vehicles (MEV). Comparisons of these crash statistics, however, need to account for the different minimum crash requirements used to identify the rural and urban intersections in the database. A few interesting data patterns and/or trends are shown in Table 1. Not surprisingly, the average annual crash frequencies at the urban intersections were always larger than those that occurred at the rural intersections. In addition, the trends in the crash frequencies for different traffic control and annual average daily entering volumes are similar for both rural and urban locations. The through-stop controlled intersections (i.e., minor-roadway stop-controlled) always had the lowest frequency of crashes and the signalized intersections the highest. Of course, the crash frequencies also increased with volume in both rural and urban areas. Table 1. Traffic control and entering volume Sample size Rural/Urban Rural Urban Rural Urban All intersections 592/1,147 3.49 8.26 0.95 0.98 Traffic control Signal 99/731 6.61 10.18 0.96 1.05 Through-stop 469/413 2.84 4.87 0.95 0.84 Four-way stop 24/3 3.28 8.00 0.88 1.29 Entering volume (VPD) < 15,000 466/226 2.83 4.85 1.01 1.25 15,000 to 25,000 91/441 4.82 6.53 0.72 0.92 > 25,000 35/440 8.80 12.06 0.71 0.88 The patterns shown in Table 1 for average crash rates are less consistent than those described for the crash frequencies. These differences are due to the introduction of entering volumes into the calculation, and the wide range of crash-volume combinations that can and do occur in the database. As expected, urban crash rates are almost always higher than their comparable rural crash rates. However, the urban crash rate calculated for through-stop-controlled intersections was smaller than that calculated for similar rural intersections. This outcome appears to be a result of the database containing data for a number of urban through-stop-controlled intersections with very large entering volumes (e.g., more than 55,000 VPD) but relatively few crashes (e.g., 10 or fewer in three years). In some cases a crash rate produced for this type of situation may not be a true measure of its safety. The high volumes may simply be restricting its use by minor roadway vehicles and subsequently reducing the potential for crashes. The four-way stopcontrolled crash rate in Table 1 should also be used with caution due to the small sample used in its calculation (only three intersections with these characteristics are in the database). Not surprisingly, the crash rates at both rural and urban intersections also decrease with volume (i.e., the volume using the intersections generally increases more quickly than the number of crashes at the intersections). Intersection Geometries In the WisDOT study, a total of 481 rural and 918 urban intersections were assigned one of 18 geometric categories. These totals represent about 80% of the locations in the intersection crash database. Four primary geometrics were used to differentiate the 18 geometric categories. These geometrics included the number of intersection approach legs, number of major roadway lanes, whether the major roadway had a Campbell, Knapp 3
median, and the existence of left-turn lanes on the major roadway. The intersections were grouped by these four geometrics for both rural and urban intersections. The and rate summary statistics for these groups were then calculated and are summarized in the following paragraphs. Number of Approach Legs The average annual crash frequencies and average crash rates calculated for four-legged intersections were higher than those calculated for three-legged intersections for both rural and urban intersections (see Table 2). This result is not unexpected, because the number of potential vehicle conflicts at four-legged intersections is more than three times that of three-legged intersections. At three-legged intersections, rural intersections observed a higher average crash rate than urban intersections. This outcome was believed to be the result of a series of very high volume urban intersections with very few crashes. At four-legged intersections, both rural and urban intersections observed similar average crash rates. Table 2. Approach legs approach legs Rural/Urban Rural Urban Rural Urban Three-Leg 105/103 2.63 5.82 0.88 0.72 Four-Leg 349/615 3.89 8.61 0.98 1.00 Number of Lanes Table 3 shows that the average annual crash frequencies calculated for intersections with a four-lane major roadway was higher than that calculated for intersections with a two-lane major roadway in both rural and urban locations. Since four-lane roadways typically have higher volumes than two-lane roadways, this result was not unexpected. Although intersections with four-lane major roadways generally observed more crashes than intersections with two-lane major roadways, these four-lane facilities averaged lower crash rates in both rural and urban locations. The lowest average crash rate was observed in rural intersections with a four-lane major roadway. Table 3. Number of lanes on the major roadway major roadway Rural/Urban Rural Urban Rural Urban Two-lane 319/178 3.12 5.51 1.03 1.07 Four-lane 143/570 4.72 9.04 0.79 0.99 Median Existence The average annual crash frequencies and average crash rates calculated for intersections with or without a median, in both rural and urban locations, are shown in Table 4. Since medians separate opposing traffic and sometimes enable the construction of exclusive left-turn lane storage bays, it was anticipated that intersections on divided roadways would have lower crash rates than intersections on undivided roadways. As expected, intersections on divided roadways had lower crash rates than intersections on undivided roadways. It appeared that rural intersections benefited most from the divided geometry and Campbell, Knapp 4
observed a crash rate of 0.79 crashes compared to the undivided rural intersections, which observed a crash rate of 1.01 crashes. Table 4. Existence of a median on the major roadway major roadway Rural/Urban Rural Urban Rural Urban Undivided 350/338 3.22 6.16 1.01 1.06 Divided 112/410 4.85 9.88 0.79 0.96 Left-Turn Lanes The addition of left-turn lanes to an intersection is a typical geometric improvement used to add capacity and improve safety. The average crash frequencies and rates for intersections with and without left-turn lanes on the major roadway are shown in Table 5 for both rural and urban locations. At rural intersections, the average crash rate for intersections without left-turn lanes was higher than for intersections with leftturn lanes. However, the difference in crash rates was relatively small. In urban locations, intersections with or without left-turn lanes observed nearly equivalent crash rates. While the results appear to indicate that the safety impact of a left-turn lane may be small or insignificant, it must be recognized that left-turn lanes are generally added based on capacity needs. Intersections with left-turn lanes may have been as crash prone as intersections without left-turn lanes simply because they had higher left-turn volumes, resulting in more potential vehicle conflicts. Table 5. Left-turn lanes on the major roadway major roadway Rural/Urban Rural Urban Rural Urban Left-turn lanes 261/514 4.17 9.21 0.92 1.00 No left-turn lane 201/234 2.90 5.98 0.99 1.01 CONCLUSIONS Limited public resources require the efficient and effective application of intersection safety improvements. Intersection locations that may need more detailed safety analysis and/or potential improvements must be identified. An understanding of the typical or expected intersection crash patterns within a jurisdiction can assist transportation professionals with this identification. The intersection safety measures presented in this paper are from the draft version of the WisDOT intersection safety report, in which crash data from the years 1998 to 2000 were analyzed. The final version of the report, scheduled for completion in July 2006, will be updated for the years 2001 to 2003 and will contain numerous typical and/or expected safety measures at intersections based on area type (i.e. rural or urban), traffic control (e.g. signal, through-stop, and four-way stop), traffic volume, and 18 different intersection geometric categories. In this paper, a portion of the intersection crash statistics calculated in the draft version of the WisDOT intersection safety report was presented. The crash statistics were summarized based on area type, traffic Campbell, Knapp 5
control, traffic volumes, and four primary geometrics: number of approach lanes, number of through lanes on the major roadway, existence of a median on the major roadway, and existence of left-turn lanes on the major roadway. All of these geometric characteristics are typically added as volumes increase. The following conclusions are based on the results of the intersection crash statistics calculations presented in this paper. The crash database included three years (1998 to 2000) of crash information from those intersections that met a predefined minimum crash requirement. The application of this type of filter is typical, and is normally applied to limit the scope of a safety evaluation to those facilities expected to be of interest. The database summarized in this report considered urban intersections if they had five or more crashes in any one year. Rural intersections were included in the database if they had three or more crashes in any one year. All the locations in the database were along the state highway or connecting highway system. The database included information about more than 34,000 crashes at more than 1,700 locations. The crash statistics for the intersections were calculated for rural and urban intersections. The average annual crash frequencies for the rural and urban intersections, respectively, were 3.49 and 8.26 crashes per year. The average rural and urban intersection crash rates, however, were determined to be 0.95 and 0.98 crashes, respectively. For more detailed safety evaluations, similar statistics were also provided for rural and urban intersections with different traffic control and annual average daily entering volumes. The patterns and trends found for the crash frequencies and crash rates at rural and urban intersections were generally as expected. Not surprisingly, the crash frequencies increased and the crash rates decreased with volume at both rural and urban locations. The average annual at signalized intersections was also greater than this measure at four-way stop-controlled intersections. Through-stop-controlled (i.e., minor-roadway stop-controlled) intersections had the smallest average annual. The crash rate patterns found for rural and urban intersections with different traffic control varied more than the previously described patterns. Rural signalized and through-stop-controlled intersections in the database had similar average crash rates, and the rural four-way stop-controlled intersections had the lowest crash rate of the three. At the urban intersections, through-stop-controlled locations exhibited the lowest average crash rate, but this outcome was believed to be the result of a series of very high volume intersections with very few crashes. In fact, the average crash rate calculated for urban through-stop-controlled intersections was unexpectedly smaller than the same measure for rural intersections. This characteristic of the database is believed to be one of its recognized weaknesses. The crash rate calculated for signalized and four-way stop-controlled urban intersections was greater than those for the rural intersections, and the four-way stop-controlled rate was the largest average urban crash rate calculated. Unfortunately, this four-way stop-controlled crash rate was also only based on data from three intersections, and it should be used with caution. The lack of data for four-way stop-controlled intersections is the second recognized weakness in the database. The average crash rate for four-legged intersections was higher than at three-legged intersections. The difference was more evident at urban intersections than at rural intersections. The average crash rate for intersections with a four-lane major roadway was lower than at intersections with a two-lane major roadway. The difference was more evident at rural intersections than at urban intersections. Campbell, Knapp 6
The average crash rate for intersections without a median on the major roadway was higher than at intersections with a median on the major roadway. The difference was more evident at rural intersections than at urban intersections. At rural intersections, the average crash rate for intersections without left-turn lanes on the major roadway was higher than at intersections with left-turn lanes on the major roadway. At urban intersections, nearly equivalent crash rates were observed regardless of the existence of left-turn lanes. The results appear to indicate that the safety impact of a left-turn lane may be small or insignificant. However, it must be recognized that left-turn lanes are generally added when warranted by capacity demands. Intersections with left-turn lanes may have been nearly as crash prone as intersections without left-turn lanes simply because they had more potential vehicle conflicts resulting from higher left-turn volumes. ACKNOWLEDGMENTS The authors thank the Wisconsin Department of Transportation for providing the funding and guidance necessary to complete the project used to create this paper. The opinions, findings, conclusions, and views expressed in this paper are those of the authors and not necessarily those of the Wisconsin Department of Transportation. REFERENCES Knapp, K. K., and J. R. Campbell. 2004. Intersection Crash Summary Statistics for Wisconsin: Draft Report. Madison, WI: Midwest Regional University Transportation Center, University of Wisconsin-Madison. Campbell, Knapp 7