• Federal Highway Administration >
• Publications >
• Research Publications >
• 99207 >
• 05.Cfm >
• Prediction of the Expected Safety Performance of Rural Two-Lane Highways

Prediction of the Expected Safety Performance of Rural Two-Lane Highways

5. SENSITIVITY ANALYSIS RESULTS

A sensitivity analysis was performed to illustrate the estimated effect on safety of the various parameters and factors in the accident prediction algorithm. The results of this sensitivity analysis are present in this section.

All of the sensitivity analyses presented here were performed with the calibration factors for both roadway segments (C_r) and intersections (C_i) set equal to 1.00. Thus, these results do not represent the safety conditions experienced by any particular highway agency and should be interpreted primarily in a relative sense.

Roadway Segments

The sensitivity analysis for roadway segments first established the variation of accident frequency and accident rate with ADT for the base or nominal condition. Then specific AMFs in the prediction algorithm for roadway segments were varied one at a time. Finally, the predicted accident frequencies and accident rates for combinations of AMFs with extremely high and extremely low accident experience were determined.

Nominal or Base Condition

The nominal or base condition for evaluating roadway segments consists of the following combination of conditions:

• Lanes that are 3.6-m (12-ft).
• Paved shoulders that are 1.8-m (6-ft).
• Tangent roadway.
• Level grade (0 percent).
• Three driveways per km (5 driveways per mi).
• Roadside hazard rating = 3.
• No passing lanes or short four-lane sections.

Table 6 illustrates the variation of accident frequency and accident rate with the roadway segment ADT for the nominal or base condition. The table shows that for the nominal or base condition the accident frequency per mile per year increases linearly with increasing ADT, while the accident rate per million veh-mi remains constant.

Table 6 and the other tables in this section of the report are presented in conventional units because all of the equations and AMFs on which they are based are in conventional units (see sections 3 and 4).

Table 6. Sensitivity of Safety to ADT for Nominal Conditions for Roadway Segments.

Conversion: 1 mi = 1.61 km

Lane Width

Table 7 presents the sensitivity of safety to lane width while all other factors are held at their nominal or base conditions. The table shows that under low-volume conditions there is very limited sensitivity of safety to lane width, while the sensitivity is larger at higher volume levels. For ADTs above 2,000 veh/day, accident frequency is 16.5 percent higher for 2.7 m (9 ft) lanes than for 3.6 m (12 ft) lanes.

Table 7. Sensitivity of Safety to Lane Width on Roadway Segments.

Conversion: 1 mi = 1.61 km; 1 ft = 0.305 m

Shoulder Type and Width

Table 8 presents the sensitivity of safety to shoulder type and width while all other factors are held at their nominal or base condition. Like the lane width effect, there is very limited sensitivity of safety to shoulder type and width at low volume levels. For ADTs above 2,000 veh/day, accident frequency can differ by a maximum of 25 percent among various combinations of shoulder type and width.

Table 8. Sensitivity of Safety to Shoulder Type and Width on Roadway Segments.

Conversion: 1 ft = 0.305 m; 1 mi = 1.61 km

Horizontal Curvature

Tables 9 and 10 present the sensitivity of safety to factors related to horizontal curvature. Table 9 compares the safety performance of a tangent roadway with various combinations of horizontal curve length and radius with and without spiral transitions.

The values in the table are computed with all factors other than horizontal curvature set to their nominal or base conditions. The table shows that the safety performance of long flat curves is only slightly worse than the safety performance of a tangent roadway. However, short sharp curves can have much higher accident rates. A horizontal curve with a length of 31 m (100 ft) and a radius of 31 m (100 ft) on a roadway segment would be expected to have an accident rate over 28 times as high as a tangent section on the same roadway. Addition of spiral transition curves can reduce accident frequencies up to a maximum of 6.6 percent for the curves selected for table 9.

Table 10 shows the expected effect on safety of superelevation deficiencies for horizontal curves. the table shows, as indicated in equation (21) that a superelevation deficiency of 0.02 increases accidents on the curve by 6 percent and a deficiency of 0.04 increases accidents by 12 percent.

Table 9. Sensitivity of Safety to Horizontal Curve Length and Radius on Roadway Segments.

Conversion: 1 ft = 0.305 m; 1 mi = 1.61 km

Table 10. Sensitivity of Safety to Horizontal Curve Superelevation Deficiency on Roadway Segments.

Conversion: 1 ft = 0.305 m; 1 mi = 1.61 km

Grade

Table 11 illustrates the sensitivity of safety to roadway grades. The table shows that, as also indicated in table 4, steeper grades increase accidents by 1.6 percent per 1-percent increase in grade.

Table 11. Sensitivity of Safety to Percent Grade on Roadway Segments.

Conversion: 1 mi = 1.61 km

Driveway Density

Table 12 presents the sensitivity of safety to driveway density for roadway segments while all other factors remain at their nominal or base conditions. The table shows that a roadway segment with 19 driveways per km (30 driveways per mi) can experience up to four times as many accidents as a similar roadway segment with no driveways. The sensitivity of safety to driveway density is greater at lower ADTs than at higher ADTs, although the absolute magnitudes of the predicted accident frequencies at low ADT are very low. Nevertheless, it might be more reasonable to expect greater sensitivity of accidents to driveways at higher ADTs than at lower ADTs. Further research on this issue would be desirable.

Table 12 also shows the predicted accident frequency and accident rate for two-lane highway sections with two-way left-turn lanes (TWLTLs). The AMF for TWLTLs is based on equations (23) and (24). The accident reduction effectiveness of a TWLTL ranges from 2 to 23 percent as a function of driveway density.

Table 12. Sensitivity of Safety to Driveway Density on Roadway Segments.

Conversion: 1 mi = 1.61 km

Passing Lanes

Table 13 presents the sensitivity of safety to passing lanes and short four-lane sections on roadway segments. The table shows that, as explained in section 4 of this report, installation of passing lanes to increase passing opportunities reduces accidents by 25 percent and installation of short four-lane sections to increase passing opportunities reduces accidents by 35 percent.

Table 13. Sensitivity of Safety to Presence of Passing Lanes and Short Four-Lane Sections on Roadway Segments.

Conversion: 1 mi = 1.61 km

Roadside Design

Table 14 presents the sensitivity of safety to roadside hazard rating on roadway segments while all other factors are held at their nominal or base conditions. The table shows that roadside hazard rating can increase total accident frequency by up to 50 percent over the full range of roadside hazard ratings.

Table 14. Sensitivity of Safety to Roadside Hazard Rating on Roadway Segments.

Conversion:1 mi = 1.61 km

Combinations of Geometric Design and Traffic Control Features

Table 15 presents the sensitivity of safety to extreme combinations of geometric design and traffic control features. The low accident frequency combination represents the “best” combination of features considered in the previous sensitivity analyses. Specifically, this low accident frequency combination includes:

• Lanes that are 3.6-m (12-ft).
• Paved shoulders that are 2.4-m (8-ft).
• Tangent roadway.
• Level grade (0 percent).
• No driveways.
• Roadside hazard rating = 1.
• Short four-lane sections used to increase passing opportunities.

The accident frequencies and rates shown in the table represent levels that are unlikely to be improved further through geometric design or traffic control modifications.

By contrast, the high accident frequency combination represents the “worst” combination of features considered in the previous sensitivity analyses. Specifically, the high accident frequency combination includes:

• Lanes that are 2.7-m (9-ft).
• No shoulders.
• Horizontal curve with length of 31 m (100 ft), radius of 31 m (100 ft), no spiral transition curve, and a superelevation deficiency of 0.04.
• An 8-percent grade.
• Nineteen driveways per km (30 driveways per mi).
• Roadside hazard rating = 7.
• No passing lanes or short four-lane sections.

The accident frequencies and rates shown in the table are extremely high, but the combination of geometric and traffic control features they represent is so extreme that it is unlikely to exist in the real world.

Table 15. Sensitivity of Safety to Extreme Combinations of Geometric Design and Traffic Control Features.

^{Conversion: 1 mi = 1.61 km}

^a These combinations of geometric design and traffic control features are defined in the accompanying text.

Three-Leg STOP-Controlled Intersections

A sensitivity analysis was performed with the accident prediction algorithm for three-leg STOP-controlled intersections. The nominal or base condition for this analysis consisted of the following geometric design conditions:

• No major-road left- or right-turn lanes.
• No skew angle (90-degree intersection angles).
• No intersection sight distance deficiencies.

The accident frequencies per year for this condition for various combinations of major- and minor-road ADT are shown in table 16. Table 16 also shows the predicted accident frequencies for various combinations of major-road left- and right-turn lanes. As indicated in table 5 installation of a major-road left-turn lane at a three-leg STOP-controlled intersection is expected to reduce accident frequency by 22 percent. Installation of a major-road right-turn lane is expected to reduce accident frequency by 5 percent.

Table 17 presents the sensitivity of safety to skew-angle for three-leg STOP-controlled intersections. A skew angle of 10 degrees results in an accident frequency 4 percent higher than a 90-degree intersection, while a skew angle of 45 degrees results in an accident frequency 20 percent higher than a 90-degree intersection.

Table 18 presents the sensitivity of safety to of intersection sight distance limitations at three-let STOP-controlled intersections. As indicated in section 4 of this report, intersection sight distance limitations can increase accident frequency by 5 percent per quadrant.

Table 16. Sensitivity of Safety to Major-Road Turn Lanes at Three-Leg STOP-Controlled Intersections.

Note: TL=turn lane; LTL=left-turn lane; RTL=right-turn lane.

Table 17. Sensitivity to Safety to Skew Angles at Three-Leg STOP-Controlled Intersections.

Table 18. Sensitivity of Safety to Limited Intersection Sight Distance at Three-Leg STOP-Controlled Intersections.

Note: ISD = intersection sight distance.

Four-Leg STOP-Controlled Intersections

A sensitivity analysis was performed with the accident prediction algorithm for four-leg STOP-controlled intersections. The nominal or base condition for this analysis consisted of the following geometric design conditions:

• No major-road left- or right-turn lanes.
• No skew-angle (90-degree intersection angle).
• No intersection sight distance deficiencies.

The accident frequencies per year for this condition for various combinations of major- and minor-road ADT are shown in table 19. Table 19 also shows the predicted accident frequencies for various combinations of major-road left- and right turn lanes. As indicated in Table 5, a single major-road left-turn lane at a four-leg STOP-controlled intersection is expected to reduce accident frequency by 24 percent and two major-road left turn lanes are expected to reduce accident frequency by 42 percent. A single major-road right-turn lane would reduce accident frequency by 5 percent and two major-road right-turn lanes would reduce accident frequency by 10 percent.

Table 20 presents the sensitivity of safety to intersection skew angle for four-leg STOP-controlled intersections. A skew angle of 10 degrees results in an accident frequency 6 percent higher than a 90-degree intersection, while a skew angle of 45 degrees results in an accident frequency 28 percent higher than a 90-degree intersection.

Table 21 presents the sensitivity of safety to deficiencies of intersection sight distance at four-leg STOP-controlled intersections. As indicated in section 4 of this report, intersection sight distance deficiencies can increase accident frequency by 5 percent per quadrant.

Four-Leg Signalized Intersections

A sensitivity was performed with the accident prediction algorithm for four-leg signalized intersections. The nominal or base condition for this analysis consisted of a four-leg signalized intersection with no major-road left- or right-turn lanes. The accident frequencies per year for this condition for various combinations of major- and minor-road ADT are shown in table 22. Table 22 also shows predicted accident frequencies for various combinations of major-road left- and right-turn lanes. As indicated in table 5, a single major-road left-turn lane at a four-leg signalized intersection is expected to reduce accident frequency by 18 percent and two major-road left-turn lanes are expected to reduce accident frequency by 33 percent. By contrast, a single major-road right-turn lane would reduce accident frequency by 2.5 percent and two major-road right-turn lanes would reduce accident frequency by 5 percent.

The predicted accident frequency at a four-leg signalized intersection is not sensitive to intersection skew angle or intersection sight distance limitations.

Table 19. Sensitivity of Safety to Major-Road Turn Lanes at Four-Leg STOP-Controlled Intersections.

Note: TL=turn lane; LTL=left-turn lane; RTL=right-turn lane.

Table 20. Sensitivity of Safety to Skew Angle at Four-Leg STOP-Controlled Intersections.

Table 21. Sensitivity of Safety to Limited Intersection Sight Distance Deficiencies at Four-Leg STOP-Controlled Intersections.

Table 22. Sensitivity of Safety to Major-Road Turn Lanes at Four-Leg Signalized Intersections.

Note: TL=turn lane; LTL=left-turn lane; RTL=right-turn lane.

Previous | Table of Contents | Next