Highway crash data with average of 39 thousand fatalities and 2.4 million nonfatal injuries per year have repetitive and predictable patterns, and may benefit from statistical predictive models to enhance highway safety and operation efforts to reduce crash fatalities/injuries. Highway crashes have patterns that repeat over fixed periods of time within the data set for crashes such as motorcycle, bicycles, pedestrians, nighttime, fixed object, weekend, and winter crashes. In some States, these crashes are weekly, monthly, or seasonally. Contributing factors such as: age category, light condition, weather, weekday, underlying state of the economy, and others impact these variations.
Science 7 - LAND and SEA BREEZE and its Characteristics
Summer Program on Transportation Statistics, What governs Highway Crashes Recurring Patterns? - Roya Amjadi, Aug 14, 2017
1. “What Governs
Highway Crashes Recurring
Patterns?”
Roya Amjadi,
FHWA, Turner-Fairbank Research Center
SAMSI, Summer Program on Transportation
August, 14, 2017
1
2. Overview
• Purpose of Presentation
– a crash course in highway safety
• Highway Safety Concerns
• FHWA and NHTSA roles in safety
• Literature Review for Seasonal Adjustments
• Seasonal and Recurring Patterns in Highway
Crashes
• Selected Crash Topics for Recurring Patterns
• Summary
• Discussion/Questions
2
3. Purpose of Presentation
1. Outreach and Communicate the recurring
patterns of highway crashes.
2. Identify the need for using Time Series
Statistical methodologies for highway planning,
design, operation, and safety improvements.
3. Propose development of “highway crash
specific” Time Series methodologies using
crash data.
3
4. Evaluation of Low Cost Safety Countermeasures
Is a 40 State Departments of Transportations study
4
5. Death Timing May Be Predictable
Canadian Statistics’ Richard Trudeau Health
Reports, Summer 1997, stated:
• While death may happen at any time as a result
of illness or accident, its timing is predictable to
some extent.
• Deaths due to specific causes tend to follow a
yearly cycle.
• Because seasonal upsurges of deaths from
specific causes are predictable, preventive
health and safety measures may reduce the toll.
5
7. Why Seasonal Adjustment?
Long Term Goal:
1. Eliminate cyclical/seasonal influences for development
of reliable crash models.
• Identify major concerns for highway crashes, and advance
research for Safer design and operation.
Short Term Goal:
1. Identify causes behind crash upsurges in certain time
of day, day of week, month of year, or other time
periods.
• Determine “effective” low-cost safety countermeasures .
7
8. Highway Crash
Major Contributing Factors:
1. Driver (i.e. age, gender, behavior, intoxication)
2. Vehicle (i.e. body type)
3. Road geometric design (i.e. number of lanes, speed
limit, curve, grade, and median)
4. Roadside Design (i.e. structure, culvert, ditch, curb,
sign, light pole, and fence)
5. Environment (i.e. weather, light, and trees)
6. Other major factors (i.e. economy, and TBD)
8
9. FHWA and NHTSA
Roles In Highway Safety
FHWA:
• Works with State and responsible for:
– the design, construction, and maintenance of the
roads
– Responsible for “roads” safety
NHTSA:
• Directs the highway safety and consumer programs for:
– Driver
– Vehicle
9
10. FATALITY ANALYSIS REPORTING SYSTEM (FARS)
Crashes with Minimum One Fatality
• Since 1994, highway crashes have caused:
– Average 39,348 fatalities per year
– Over 2.4 million nonfatal injuries per year
• Cost over $200 billion to our national economy per year
10
11. Literature on Seasonal Adjustment
Very Limited!
• A. Karlsson and K. Willero, Time Series Analysis of
Fatalities in the Traffic, Swedish National Road and
Transport Research Institute, 2005
– used time series models to explain correlation
between “observed” and “estimated” number of crash
fatalities.
• The predictions showed that Swedish “Zero Vision”
will not be fulfilled without radical changes in
highway safety and operation.
• Predicted number of fatalities will stay at the same
level as in recent years.
12. NHTSA
Fatality Analysis Reporting System (FARS)
FARS is a database for crashes with Minimum One Fatality.
• Used data for 2010 to 2015 (6 years.)
– Isolated data for:
1. Vehicle “Number 1” AND
2. Person Type “Driver” for vehicle number 1.
Extracted:
I. Non-Intersection cases= 125,680 crashes
II. Intersection cases= 1,948 crashes
Only used Non-Intersection data
12
13. Intersections
• Where two or more roads meet with different traffic
volumes.
• May be stop controlled or traffic signal controlled.
• By design have different road geometrics, speed (if not
stopped), lighting, and pedestrian volume.
• Traffic flow is not all straight forward, and vehicles may
turn in different directions.
• Have many accesses, or access control to business or
other properties located around intersection that may
flow dynamics.
• Other different characteristics from straight or curved
segment of the road.
13
14. Recurring Highway Crash
FARS Data, 2010-2015
Selected crash highlights:
1. Hourly, weekly, and monthly crash reoccurring
patterns
2. Driver Age
3. Cause of Crash (first harmful events)
a. Motor Vehicle in Transport
b. Pedestrian
c. Curb
4. Vehicle (Motorcycle)
5. Vehicle Speed
14
15. All NON_INTERSECTION
FIRST HARMFULL EVENT
15
Event Frequency Percent
Number Total 125,680 100
1 Motor Vehicle In-Transport 33,097 26.3
2 Pedestrian 17,755 14.1
3 Rollover/Overturn 14,839 11.8
4 Tree (Standing Only) 13,123 10.4
5 Ditch 5,738 4.6
6 Embankment 5,463 4.3
7 Curb 4,127 3.3
8 Utility Pole/Light Support 3,644 2.9
9 Guardrail Face 3,643 2.9
10 Culvert 2,403 1.9
11 Pedalcyclist 2,385 1.9
12 Fence 2,166 1.7
13 Traffic Sign Support 1,798 1.4
14 Mail Box 1,629 1.3
15 Parked Motor Vehicle 1,548 1.2
16 Other Post, Other Pole or Other Supports 1,329 1.1
17 Other Fixed Object 1,323 1.1
18 Concrete Traffic Barrier 1,101 0.9
21. 21
Age category crash percentages for; 15-19 through 50-54 are
higher than their population percentages.
22. Driver Age category 18-29
Crashes have 24 hour, weekly, and monthly patterns.
22
- 42% of crashes occurred from 7 pm-3 am.
- 55% of crashes occurred on Friday, Saturday, and Sunday.
23. SPEED RELELATED CRASH
Sample= 61,327
Odds of Having a Fatal Crash
for Age 18-29 Compared to Age 40-50
Yes, Racing 5.73
Yes, Exceeded Speed Limit 1.68
Yes 1.39
Yes, Specifics Unknown 1.38
Yes, Too Fast for Conditions 1.31
Unknown 0.94
No 0.82
23
0%
2%
4%
6%
8%
10%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
CRASH PERCENTAGE PER HOUR
OF DAY
2010-2015
Age 18-29 Age 40-50
FIRST HARMFULL EVENT Odds
Curb 1.57
Utility Pole/Light Support 1.46
Tree (Standing Only) 1.26
Mail Box 1.21
Fence 1.16
Other Fixed Object 1.15
Embankment 1.09
Concrete Traffic Barrier 1.09
Guardrail End 1.07
Traffic Sign Support 1.06
Parked Motor Vehicle 1.04
Other Post, Pole or
Supports
1.03
24. Driver Age category 18-29
Vehicle Occupant Number and Safety
24
NUMBER OF
OCCUPANTS
Sample= 61,327
Age 18-29
Cashes
(40,169)
Age 40-50
Crashes
(21,158)
Odds of Having a Fatal Crash
Age 18-29
Compared to
Age 40-50
1 24,960 15,949 0.82
2 8,731 3,410 1.35
3 3,418 929 1.94
4 1,768 438 2.13
5 804 198 2.14
6 253 100 1.33
7 106 50 1.12
8 55 23 1.26
9 23 20 0.61
10 12 9 0.70
25. Motor Vehicle in Transport Crashes
25
Have a 24 hour recurring pattern.
- 36% of crashes occurred between 12 pm to 5 pm (5 hour peak.)
- Crashes were 2.12 times more likely to occur in period of 12 pm-5 pm compared to
crashes that occurred in rest of the hours.
26. Pedestrian Crashes
26
- 61% of crashes occurred between 5 pm-12 am (8 hour peak)
- Crashes were 3.16 times more likely to occur in period of 5 pm-12 am
compared to crashes that occurred in rest of the hours.
27. Curb Crashes
27
- 66% of crashes occurred from 7:00 pm to 4:00 am.
- Crashes were 3.3 times more likely to occur in period of 7 pm-4 am
compared to crashes that occurred in rest of the hours.
28. Motorcycle crashes
28
Have 24 hour, weekly, and monthly patterns.
- 359% more crashes occurred between 12 pm-2 am compared to other
hours.
- 56% more crashes occurred in Fri-Sun compared to Mon-Thu.
- 71% more crashes occurred in May-Sept compared to Oct-April.
29. Speed Related Crashes
Low Hanging Fruit
33% of non-intersection related crashes were speed related
Newton’s Second Law applies for constant mass
F= m dv/dt = ma
• Compared to non-speed related crashes, speed
related involved:
1.43 times more fatalities
1.42 times more sever injuries
2.00 times more crashes on curves
29
31. Speed Related Crash
SAFETY IMPROVEMENTS?
1. Low Cost (short term)
In segment with high percent of crashes:
• Lower speed limit and advisory speed
• Install variable speed limit
• Remove/relocate fixed objects
• Increase pavement friction
• Enforce speed at night
• Public education/communication
• Nighttime visibility (markers)
2. Design (long term)
• Increase curve radius and super elevation
• Implement “clear zone”
3. Other (policy, vehicle design)
31
32. Selected Crashes
with Minimum One Fatality
• Possible low cost solutions for:
– Driver age 18-29
– Night time
– Speed
– Weekends
– Pedestrian
– Curb
– Other
What governs crashes recurring patterns?
32
33. Opportunities to Learn From History!
• If highway crashes look chaotic and difficult to predict,
that means we have not recognized their patterns yet.
• “Those who fail to learn from history are doomed to repeat
it” Sir Winston Churchill
Highway crashes are predictable and preventable
Propose development of “highway crash specific” Time
Series methodologies for outstanding safety concerns
using crash data.
33
35. CRASH DATA RESOURCES
• Despite the limited statistical methodologies available, USDOT
agencies such as the FHWA, NHTSA, and FMCSA have succeeded in
developing and maintaining comprehensive and reliable data resources
(i.e., Highway Safety Information System, Fatality Analysis Reporting
System, and Motor Carriers Management Information System) that
have historical crash and roadway data for multiple years.
• The Transportation Research Board’s second Strategic Highway
Research Program (SHRP 2) (2006-2015) has collected an
unprecedented amount of real -time data on driver behavior and the
driving context.
• Most State DOTs collect and manage their own crash highway data for
roads, traffic, operation, planning, construction, work zone,
maintenance, pavement, bridges, structures, roadside objects, and
anything within the right-of-way, including weather and pavement
temperature.
35
36. Other Available Data for Highway Safety Research
• FHWA’s Highway Performance Management System
(other highway data)
• Bureau of Economic Analysis
• U.S. Censes Bureau (i.e. population, education, age)
• Bureau of Labor Statistics (i.e. employment)
• Navy’s Astronomical Information Center (i.e. hour of day)
• National Oceanic and Atmospheric Administration (NOAA)
National Centers for Environmental Information (i.e.
temperature)
• Department of Motor Vehicles (DMV)
• American Automobile Associations (AAA)
• Insurance companies
• Others
36
37. FARS Links
• FARS data at; https://www-
fars.nhtsa.dot.gov/Main/index.aspx
• FARS data user manual at:
https://crashstats.nhtsa.dot.gov/Api/Public/
ViewPublication/812315
37