Download presentation
Presentation is loading. Please wait.
Published byBritton Kelley Modified over 9 years ago
1
Continuous Risk Profile: A Simple Method for Identifying Sites for Safety Investigation. Koohong Chung, Ph.D. California Department of Transportation Highway Operations
2
Outline 1. Motivation and Background 2. Continuous Risk Profile 4. Discussion 5. Concluding Remarks 3. Findings
3
1. Motivation and Background
4
“Sliding Moving Window” Approach 0.2 mile roadway the reference value the number of collision with the window
5
1. Motivation and Background “Sliding Moving Window” Approach 0.2 mile roadway the reference value slide the window by small increment of 0.1 mile and repeat the same analysis 0.01 mile the number of collision with the window <
6
1. Motivation and Background “Sliding Moving Window” Approach 0.2 mile roadway The site will be reported it to Table-C or Wet Table-C and move the window to the next 0.2 mile segment the reference value the number of collision with the window >
7
Task Force (2002) conducted survey among 44 safety engineers A. Identify sites that are adjacent to each other as one site B. High false positive rate for both Table-C and Wet Table-C 1. Motivation and Background
8
Direction of traffic Pattern I: Collision causative factor can reside outside of 0.2 mile window.
9
1. Motivation and Background Pattern II: Collisions can accompany secondary collisions in the vicinity.
10
1. Motivation and Background The collision data on freeways were often spatially correlated. Direction of traffic Reference Rate
11
2. Continuous Risk Profile (CRP)
12
2. Continuous Risk Profile Direction of traffic Cumulative number of Collisions B(d) A(d)
13
2. Continuous Risk Profile Rescaled Cumulative Collision Count Curve (I-880 Northbound, Alameda County, California, 2003)
14
2. Continuous Risk Profile M(d) = Where d 0 = beginning postmile d end = ending postmile K, are integers f(d) = A(d) – B(d-d o ) D start < D end 2L = size of the moving average l = increment For and CRP
15
2. Continuous Risk Profile M(d) = Where d 0 = beginning postmile d end = ending postmile K, are integers f(d) = A(d) – B(d-d o ) D start < D end 2L = size of the moving average l = increment For and CRP A Method for Generating a Continuous Risk Profile for Highway Collisions (2007) Chung and Ragland To be Determined, (working paper) Chung, Ragland and Madanat
16
2. Continuous Risk Profile postmile By dividing the above CRP by AADT, the unit can be converted to number of collisions per vehicle miles.
17
3. Findings
18
Comment from hydraulic division We were thinking that a plot like these presented to Hydraulics prior to a major rehabilitation project would be ideal in assisting us evaluate and upgrade drainage at the high accident locations as necessary. …Could I encourage you to have a discussion at the end of your report recommending that Caltrans generate such plots? It (CRP plot) would help us out immeasurably during design. -Joseph Peterson, Office Chief,District 4 Hydraulic-
19
3. Findings CRP can be used to identify freeway sites that display high collision rate only under wet pavement condition. Findings 1:
20
DRY WET WET ONLY
21
DRY WET WET ONLY “Identification of High Collision Concentration Locations Under Wet Weather Conditions”, Hwang, Chung, Ragland, and Chan
22
3. Findings Findings 2: CRP are reproducible over the years and can proactively monitor traffic collisions.
27
3. Findings Findings 3: CRP plots can be used to capture the “spill over benefit”.
28
Postmile
29
Project Completed in 2001 Postmile
30
Spillover Benefit Postmile
31
3. Findings Findings 4: Using CRP, you can save time in site investigation.
32
Direction of Traffic 2003 2002 2001 2000 1999 ON OFF Access PM 18.1
33
PM 17.887PM 18.141PM 18.3
34
Accidents Rate (Accidents/Mile) (SR- 91W) 4 Times Higher
35
Accidents Data Analysis (PDO) 2 Times Higher
36
Accidents Data Analysis (INJURY) 3 Times Higher
37
Due to the inclined freeway, drivers tend to accelerate
38
Heavy Vegetations
39
1) Inclined On-Ramp 2) Heavy vegetations
40
Map of HCCL (SR-91 W) 1) Inclined On-Ramp 2) Heavy vegetations
41
3. Findings More Findings: “Comparison of Collisions on HOV facilities with Limited and Continuous Access during Peak Hours”, Jang, Chung, Ragland, and Chan “Identification of High Collision Concentration Locations Under Wet Weather Conditions”, Hwang, Chung, Ragland, and Chan
42
4. Discussion
43
Highways Intersections Ramp YES (SafetyAnalyst)
44
Accidents Per Mile Per Year AADT +1.5 б -1.5 б LOSS -I LOSS -II LOSS -III LOSS -IV SPF (“Level of Service of Safety”, Kononov and Allery) 4. Discussion
45
Accidents Per Mile Per Year AADT +1.5 б -1.5 б LOSS -I LOSS -II LOSS -III LOSS -IV SPF (“Level of Service of Safety”, Kononov and Allery) 4. Discussion
46
“The Analysis of Count data: overdispersion and autocorrelation”, Barron “.. ML estimation of both Poisson and negative binomial regression typically requires independent observations. This assumption will often not be true in time-series data, and Poisson and negative binomial regression are then problematic.” 4. Discussion
47
Accidents Per Mile Per Year AADT Unbiased SPF biased SPF 4. Discussion
48
5. Concluding Remark
49
CRP is simple to use and provides overview of collision rates of extended segment of freeways over the years. CRP can identify sites that display high collision rates only under certain condition. (ex: wet hot spots) CRP can proactively monitor traffic collision rates. CRP can be used to capture “spill over benefit” of countermeasure. Spatial correlation is not an issue in constructing CRP
50
5. Concluding Remark In future research, III. Expand CRP approach for CALTRANS intersections and ramp. I. Continue exploring different areas where CRP can be used. II. Friendly interface CALTRANS
51
Thank you! Q & A
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.