Lecture 20: Identifying Hazardous Locations

1
Highway Traffic and Safety Analyses
Lecture 20 Identifying Hazardous Locations
Purdue University School of Civil
Engineering West Lafayette
2
Reading Assignment

• Chapter 4, pages 9-26, in Guidelines for Highway
  Safety Improvements in Indiana, access via HAT
  prototype software distributed in class

3
Safety Management SystemDecision-making Process

• Identify highway hazard
• Determine causes
• Determine countermeasures
• Develop safety projects
• Select projects for implementation
• Evaluate projects effectiveness

4
Identifying Hazardous Locations

• Identification task
• Criteria for selection of locations
• Statistical quality control
• Policy-based criterion
• Exposure-based criterion
• Rank list
• Other methods
• Safety index
• Safety priority index

5
Identification Task

• OBJECTIVE
• Select sites that can have safety considerably
  improved with cost-effective remedial actions
• ISSUES
• Many candidate locations (sites)
• Limited information about the sites
• CONSEQUENCES OF INCORRECT SELECTION
• Selecting safe locations causes costs wasted on
  detail analysis of these locations
• Not selecting hazardous location eliminates this
  location from further consideration

6
Selection Criteria

• Excessive crash frequency
• For example, a gt am
• Promotes the most cost-effective mitigation of
  hazard (system perspective)
• Excessive risk faced by users
• For example, r gt rm, where r a/E
• Promotes fairness of the highway system by
  reducing the differences in risk experienced by
  users (user perspective)

7
Statistical Quality ControlPolicy-based am
If am is assumed by policy, how large crash
count c in n years has to be to indicate that the
unknown crash frequency is higher than am (hazard
is statistically evident)?
Hazard is statistically evident at the
significance level p if Pr(Cc mean nam)
p p. The lowest such c is control limit Lp
L0.10
L0.05
L0.01
8
System Perspective Example 1 Policy-based am

• Maximum frequency am 5 crashes per year
• p 0.10
• Number of crashes during one year c 8
• Does the location deserve attention?
• p Pr(C8mean5) 1- Poisson(8-1,5,1)
• p 0.133 gt p
• The location does not deserve attention

9
System Perspective Example 1 Policy-based am
p Pr(C8)mean5) 0.133 p gt p
10
System Perspective Example 2 Policy-based am

• Maximum frequency am 5 crashes per year
• p 0.10
• Number of crashes during the last three year c
  21
• Does the location deserve attention?
• p Pr(C21mean35) 1- Poisson(21-1,15,1)
  0.083
• p lt p
• The location deserves attention

11
User Perspective Example Policy-based rm

• The maximum crash rate for two-lane rural
  segments is rm 1 crash/million VMT
• A road segment has AADT 10,000 veh/day and is
  L2.3 miles long
• Observed number of crashes c during the last
  three years (n3) is 35
• The required significance level is p 0.05
• Does this segment require attention?

12
User Perspective Example Policy-based rm

• am rmE (rm)(AADTL365n/b)
• am (1.0)(10,0002.33653/1,000,000) 25.2
• p 1-Poisson(35-1, 25.2, 1) 0.037
• p lt p
• The segment requires attention

13
Statistical Quality ControlExposure-based am
Estimate

• A safety performance function that includes only
  exposure variables estimates the crash frequency
  an conditioned on the exposure
• The location deserves attention if the crash
  frequency is higher than the expected one
• Past research indicates that this criterion
  balances the user and system perspectives

14
Statistical Quality ControlExposure-based am
Estimate

• A signalized intersection with known volumes and
  with 46 crashes last year is analyzed at the
  p0.05 significance level
• a 25.4 crashes/year from a safety performance
  function with over-dispersion ? 0.2
• p Pr(Cc) 1 - ?x0..c-1 NegBinomDist(x, 1/?,
  1/(1?a))
• Useful equivalence
• ?x0..c NegBinomDist(x, s, f) BetaDist(f, s,
  c1)
• p 1- BetaDist(1/(1?a), 1/?, c)
• p 1- BetaDist(1/(10.225.4), 1/0.2, 46)
  0.062
• p gt p
• The site does not require attention

15
Index of Crash FrequencyIntroduction
16
Index of Crash Frequency(Policy-based am 5, n
1)
IF Index of Crash Frequency IF (c
nam)/sigma Var var c var (nam) 8 IF
(8-5)/81/2 1.06
17
Index of Crash Frequency(Exposure-based
estimated am 5, a 0.2, n 1)
IF (c nam)/sigma Var ca(nam)2
80.2(15)2 13 IF (8-5)/131/2 0.83
18
Index of Crash Cost IC
where wi is the cost of crash of severity
i. Remark wi can be any equivalency factor for
any crash category i. The name of the calculated
index should reflect its meaning. RoadHAT
19
Rank List

• Roads sorted by evidence of safety problems
  (crash counts, calculated p, certain safety
  index)
• An agency selects top candidates according to
  available safety resources

20
Other MethodsSafety Index

• Equivalent Property Damage Only crashes (EPDO)
• EPDO PDO wIINJURY wFFATAL where wI is the
  weight for an injury crash, and wF is the weight
  for a fatal crash
• Weights may reflect the differences between the
  average costs of crashes

21
Other MethodsSafety Priority Index System
(Oregon DOT)
Crash counts recorded for urban segments of 0.05
mi, and for rural segments of 0.10 mi. n number
of years, c total crash count.