6. Study Intersections

Based on the data summarized in Tables 2.2-4.4 and the discussions in Sections 4 and 5, 35 intersections were chosen for further analysis. Each of the 35 intersections appeared on several statewide, regional or district lists, such that each had both a large number of crashes and a high crash severity score. Also, as discussed in the text supporting Table 2.6, some of the intersections had a relatively large number of fatal crashes. The 35 intersections are listed in Table 6.1. Six are located in Region 1, 14 are in Region 2, six are in Region 3, and nine are in Region 4, with five in the Cedar City District, two in the Price District, and two in the Richfield District. Five of the intersections are in West Valley City, and four are in Provo. Six of the intersections are along US 89, eight are along SR 171, three are along SR 173, and three are along US 189. A total of 27 of the intersections are signalized; the other eight are not signalized, with the major street having the right-of-way (a stop sign faces cross-street traffic; i.e., two-way stop control).

6.1 Crash Types

The types of crashes that occurred at the 35 study intersections between 1994 and 2003 are summarized in Table 38. As expected, most of the crashes involved two or more motor vehicles (the research team did not distinguish between 2-, 3-, and 4-or-more vehicle crashes). In fact, of the 11,615 crashes that occurred, 10,910 (94 percent) involved motor vehicles only. The remaining 6 percent of the crashes involved a single vehicle and a pedestrian (23 percent of all single-vehicle crashes), bicyclist (28 percent), fixed object (13 percent), running off the road (18 percent), rollover (2 percent), animal (2 percent), or some other hazard. An intersection at which any of the crash types deviated from these averages may be problematic for that type. For example, at 800 North and State Street in Orem, bicyclist-vehicle crashes represented 61 percent of all single-vehicle incidents, well above the 35-intersection average of 28 percent. This intersection might have a heavier volume of bicyclists than the others; alternatively, there may be a need for improvements in the accommodations for bicycles at this location. For another example, at 5600 South and 1900 West in Roy, pedestrian-vehicle crashes represented 52 percent of all single-vehicle incidents, well above the average of 23 percent. There may be a need for improvements in the pedestrian facilities at this intersection. Also, at 5400 South and 5030 West in Kearns, single-vehicle crashes were 23 percent of all incidents, much higher than the average of 6 percent. It may be useful to closely examine this intersection for factors related to driver guidance, fixed objects, non-motorized highway users, and so forth.

6.2 Crash Severities

The crash severities at the 35 study intersections are listed in Table 6.3. Each of the intersections was selected for further study because of a large number of crashes and/or a large number of severe or fatal crashes. It is interesting to note that, at ten of the intersections, more than 10 percent of the crashes resulted in either an incapacitating injury or a fatality. These tended to be the (comparatively) "low-crash" intersections (i.e., none had more than 210 crashes between 1994 and 2003). The crash severities at one example of these intersections, 2400 South and 8000 South in Magna, are shown in Figure 6.1. At nearly all of the intersections - particularly the "high-crash" intersections - more than half of the collisions resulted in no injury. The crash severities at one example of these, Redwood Road and 5400 South in Taylorsville, are shown in Figure 6.2. At three of the intersections - Hinckley Drive and Pennsylvania Avenue in Ogden, State Street and Wood Avenue in Salt Lake City, and 3500 South and 4200 West in West Valley City - more than half of the collisions resulted in at least a "possible" injury. It is likely that there are high travel speeds on the major street at these intersections. A speed-reducing mitigation, such as a lower speed limit or traffic calming measure, may be needed. A turning restriction or prohibition from the minor street may also be a strategy. The research team did not look at crash severities by collision type - this would be a subject for further study - but it is likely that many of the minor crashes were rear-end incidents.

Figure 6.1 Crash Severities: 2400 South + 8000 West

Figure 6.2 Crash Severities: Redwood Rd + 5400 South

6.3 Direction of Travel

Table 6.4 summarizes the vehicles involved in crashes at the study intersection by direction of travel. In general, "Leg 1" is the northbound approach, Leg 2 is southbound, Leg 3 is eastbound, and Leg 4 is westbound. These data indicate the approach(es) along which crashes are concentrated at the given intersection. In many cases, the leg with the heaviest approach volume will have the most vehicular involvement. For example, at the intersection of SR 75, Main Street and 1400 North in Springville, 119 (44 percent) of the 270 vehicles involved in crashes were approaching from the south (i.e., northbound) on US 89 (Main Street). It is not readily clear why the number of vehicles involved in crashes on the northbound approach was 2.6 times that of the southbound approach. At Carbon Avenue and 100 North in Price, just over 90 percent of the vehicles involved in crashes were traveling along 100 North. About 60 percent of the vehicles entering this intersection were using 100 North, so it is not clear why a disproportionate number of these vehicles were involved in collisions. Also, as mentioned earlier, the research team did not examine multiple-vehicle crashes at any of the intersections. These would be subjects for further study.

6.4 Crash Rates

Crash rates at the 35 study intersections, based on 2001-2003 traffic volumes and crashes, are summarized in Table 6.5. The rates can be compared to those listed in Table 2.7, which features all intersections between state routes having a crash rate of two or more per million entering vehicles (MEV). Fourteen of the intersections in Table 6.5 had a crash rate of two or more per MEV; six of these intersections are between state routes, which also appear in Table 2.7. Traffic volumes were not available for one or two of the approaches to some of the intersections; the crash rates listed for these may be greater than actual. It would be useful to obtain traffic volumes on the cross-streets at these locations to verify the crash rates. There appears to be a correlation between crash rates and crash totals, although the relationship is unclear. Intersections with very large numbers of crashes appear to also have high crash rates; further study is needed to identify the relationship. Some intersections with few crashes, however, have high crash rates, as shown in Table 2.7.

To prioritize intersections for mitigation, it may be useful to develop a composite ranking based on crash occurrences, crash frequencies, and crash rates. The Iowa DOT, for example, was identifying high-crash locations according to the following procedure (Souleyrette et al., 2001):

1. Rank crash sites according to the total number of crashes in a five-year study period.
2. Rank crash sites according to the crash rate, based on five years of data.
3. Rank crash sites according to the crash "loss" (similar to this report's severity score).
4. For each site, add the three rankings; the cumulative "score" is used to compile a final ranking.

Note that the cumulative scores developed in step 4 do not "weight" any of the rankings; that is, a ranking based on a crash frequency is equivalent to that based on a crash rate or loss. As indicated earlier, the research team did not compile crash rates for intersections between state and non-state routes, except for those listed in Table 6.5. It would be useful to compute these rates, then apply the Iowa DOT method (or a suitable modification) to Utah's intersections. A reasonable study period would need to be selected - the Iowa DOT used five years, while this report used ten and three years. Hauer (1997) argued for using "as much crash data as possible," primarily because crashes are relatively infrequent events. He claimed that the effects of infrastructure changes (e.g., new roads, improvements, new traffic controls, etc.) should be reflected in the crash data. To effectively use this approach, the analyst would need to be aware of the types and dates of all important changes.

6.5 Collision Types

The collision types at the 35 study intersections, based on the numerical codes described in Table 6.6, are listed in Table 6.7. A total of 11,615 crashes occurred at these intersections between 1994 and 2003; 4,259 (37 percent) involved a left-turning vehicle, and 4,410 (38 percent) were rear-end incidents. A total of 825 incidents (7 percent) were side-swipe collisions, and 770 (7 percent) were right-angle incidents. Just over 1 percent of the crashes (165) involved a pedestrian, and just under 2 percent (196) involved a bicycle. Intersections at which the collision types did not "conform" to this distribution may present special strategic needs. For example, the greatest number of backing incidents (17) occurred at Redwood Road and 5400 South in Taylorsville. This intersection also had the most left-turn involvements (365) during the study period. The intersection of 800 North and State Street in Orem had the most bicycle-vehicle collisions (22), and the largest number of right-turn involvements (41) The greatest number of pedestrian-vehicle collisions (17) occurred at 700 East and 3300 South in South Salt Lake, and the largest number of right-angle crashes (67) occurred at 3300 South and 30 West, also in South Salt Lake. Both of these types of collisions tend to be severe, so further investigation would be worthwhile. The Institute of Transportation Engineers (ITE, 2004) reported, in fact, that 60 percent of all fatal intersection crashes are right-angle collisions. The highest number of single-vehicle incidents - excluding pedestrian-vehicle and bicycle-vehicle collisions - occurred at Bangerter Highway and 3500 South in West Valley City (17). This type of crash may be associated with high travel speeds and driver behavioral issues. ITE (2004) reported that about one-third of all fatal intersection crashes involved just one vehicle (and a pedestrian, bicycle, fixed object, or other single-vehicle factor). Center Street and 900 West in Provo had the largest number of side-swipe collisions with 86. Two intersections - 5400 South and 4460 West in Kearns, and 5600 South and 1900 West in Roy - experienced three head-on collisions. The greatest number of any type of crash at any of the study intersections was the 474 rear-end collisions that occurred at University Avenue and 900 North in Provo.

Intersection safety countermeasures, extracted from Hauer et al. (2002), are discussed in Sections 2 and 5, and are displayed in Figure 2.1. Countermeasures for backing and head-on collisions are not discussed in Hauer et al.; these were the two "least popular" types of collisions at the study intersections. Backing incidents tend to occur off-street, in parking lots and residential driveways, and are rare on roads and streets. It is likely that backing maneuvers are associated with on-street parking; on-site studies of intersections having recurrent backing crashes would be needed to properly identify the critical issues. Head-on collisions generally involve wrong-way travel (i.e., a median or centerline crossover) by one vehicle. Most head-on collisions occur away from intersections; 22 of the 35 study intersections, in fact, did not experience any head-on collisions. The development of mitigating strategies for head-on collisions has concentrated on non-junction crashes. The two intersections that had three head-on collisions indicate a potentially recurring problem, however, further study is suggested. A possible mitigation would be raised medians on the intersection approaches.

6.6 Functional Radius of Influence

The numbers of crashes by distance from the intersection, in 100 ft increments, are summarized in Table 6.8. As discussed in Section 1, this study used a 500-ft radius for all intersections, based on the findings in Stover (1996). A review of the statistics in Table 6.7 indicates, however, that the radius should probably be varied. For example, a 100-ft radius captured more than half of the crashes at 16 of the study intersections. The crash activity at one intersection that exhibited this pattern, 2400 South and 8000 West in Magna, is shown in Figure 6.3. At an additional nine intersections, the 100-ft radius captured more crashes than any other 100-ft increment. A 200-ft radius captured a large number of crashes in the 100- to 200-ft band at three of the intersections, including Hinckley Drive and Pennsylvania Avenue in Ogden (Figure 6.4). Similarly, a 300-ft radius captured a large number of crashes in the 200- to 300-ft "doughnut" at one intersection (5400 South and 4460 West in Kearns); while, using the same method, a 400-ft radius applied to 3500 South and 4200 West in West Valley City. A 500-ft radius appeared to apply to only four of the intersections; at these locations, the 400- to 500-ft band was very active. Further study is needed of these radii before a conclusion can be drawn. For example, it may be useful to examine 50- or 25-ft increments. The presence of adjacent intersections may also be a factor. At University Avenue and 900 North in Provo, for example, intersections with 880 North, 940 North, and 960 North are nearby. The intersection with 960 North is, in fact, signalized, and there may be occasional spillover that affects the 900 North intersection. A heavy volume of non-motorized Brigham Young University traffic (i.e., not crossing at the intersections) may also be a factor.

One impact of overstating an intersection's functional area may be to overestimate the number of crashes occurring at that intersection. In some cases, crashes that should be attributed to an upstream or downstream intersection may be "falsely" attributed. Given that driveways proximate to an intersection can be an additional contributing factor, the challenges of pinpointing the functional area are evident. The most direct technique would be to examine an intersection in the field, taking special note of the locations of conflict points. Another technique would be to closely examine accident reports, along with accident reconstruction studies, to determine the pre-crash events and driver intentions. In a general analysis such as in this study, the best approach may be to identify functional areas by varying the radius of influence at each intersection. This would be a time-consuming exercise that may be most efficiently applied to a county or city, rather than an entire district, region, or state.

Figure 6.3. Crashes by Radius of Influence: 2400 South + 8000 West, Magna, 1994-2003

Figure 6.4. Crashes by Radius of Influence: Hinckley Drive + Pennsylvania Avenue, Ogden, 1994-2003