Incorporating Safety into the Incorporating Safety into the Highway Design ProcessHighway Design Process

What is Meant by “Safe”?What is Meant by “Safe”?

• Is This Road Safe?

What is Meant by “Safety”?What is Meant by “Safety”?• Is This Road Safe?

– Is a “Yes” or “No” answer sufficient?

– Would your answer change if you were told...• The road averages 1 crash in 10 years? or...• The road averages 100 crashes in 10 years?

Kinds of SafetyKinds of Safety

• Nominal Safety– A road that conforms to the agency’s policy,

guidelines, and warrants is “nominally” safe

– A road either is, or is not, nominally safe

• Substantive Safety– The performance of a roadway, as defined by

its “expected” crash frequency (i.e., long run average)

– Substantive safety is a continuous variable

– Useful to compare one site with “typical” site

Kinds of SafetyKinds of Safety• Safety Comparison (NCHRP Report 480)

Safety-Conscious DesignSafety-Conscious Design

• AASHTO Guidance– “Consistent adherence to minimum [design

criteria] values is not advisable”

– “Minimum design criteria may not ensure adequate levels of safety in all situations”

– “The challenge to the designer is to achieve the highest level of safety within the physical and financial constraints of a project”

• Highway Safety Design and Operations Guide, 1997

Highway CrashesHighway Crashes• Contributing Factors

– Driver • Age, gender, skill, fatigue level, alcohol, etc.

– Vehicle • Type, age, maintenance, etc.

– Environment • Light conditions, weather, precipitation, fog, etc.

– Roadway• Geometric design, traffic control, etc.

• Focus of current research– Geometric design of the roadway

Quantifying SafetyQuantifying Safety

• Safety Prediction Model– C = base crash rate × volume × length × AMF

• Accident Modification Factor (AMF)– AMF used to estimate change in crashes due

to a change in geometry (AMF = Cwith/Cwithout)

– Example: • AMFadd bay = 0.70

• Cno bay = 10 crash/yr

• Cwith bay = Cno bay × AMFadd bay = 7 crashes/yr

– Crash reduction factor (CRF) = 1 - AMF

Crash DataCrash Data• Existing Crash Databases

– Texas Department of Public Safety (DPS)– Local databases

• Severity Scale– K: Fatal– A: Incapacitating injury– B: Non-incapacitating injury– C: Possible injury– PDO: property damage only

• Reporting Threshold – $1000, informally varies among agencies

Research focus

Crash Data VariabilityCrash Data Variability• Examination of Crash History

– Annual crash counts: 2, 3, 1, 1, 7, 5, 2...

– Count in any one year is effectively random

– Variability year to year is LARGE

– So large that...• It is very difficult to determine if the change in count

from year to year is due to a change in geometry, traffic volume, or traffic control device

• It can frustrate efforts to reduce crashes (a change was made but crashes increased)

• It can fool us into thinking a change that we made significantly reduced crashes (when it really did not)

Crash Data VariabilityCrash Data Variability

• Questions– What is the true mean crash frequency at

this site?

– Is a 3-year average reliable?

0

2

4

6

8

0 5 10 15 20 25 30 35

Year

Cra

sh

Fre

qu

enc

y, c

ras

he

s/y

r

Each data point represents 1 year of crash data at one site

Crash Data VariabilityCrash Data Variability• Observations

– The average of 3 years (= 6 crashes)...• 2.0 crashes/yr• 0.7 to 4.3 crashes/yr (± 115%)

– The average of 35 years (= 100 crashes)…• 2.8 crashes/yr • 2.2 to 3.3 (± 20%)

0

2

4

6

8

0 5 10 15 20 25 30 35

Year

Cra

sh

Fre

qu

en

cy

, cra

sh

es

/yr

Running Average

Upper Limit (95% confidence interval)

Lower Limit

– One site rarely has enough crashes to yield an average with a precision of ± 20%

Influence of DesignInfluence of Design• Question

– 15 intersections have left-turn bays added

– Research shows bays reduce crashes by 20%

0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25 30 35

Site

Cra

sh

Fre

qu

en

cy

, cra

sh

es

/yr

0 5 10 15

Before Bay After Bay

Site 4

– What crash frequency do you expect for site 4 after the bay is installed?

Each data point represents 1 year of crash data

Average = 10

Influence of DesignInfluence of Design• Observations

– Random variation makes trend difficult to see

– Most sites show crash reduction

– Site 4, and a few other sites, had more crashes

– This does not mean bay won’t be effective in long run 0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25 30 35

Site

Cra

sh

Fre

qu

en

cy

, cra

sh

es

/yr

0 5 10 15

Before Bay After Bay

Site 4

Site 4

Influence of DesignInfluence of Design• Observations

– Distribution of crash change for sites with average of 10 crashes/yr and 20% reduction

– When reduction is small, random variation will let crash frequency increase at some sites in the year after

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

-15.00 -10.00 -5.00 0.00 5.00 10.00

Change in Annual Crash Frequency

Pro

ba

bili

ty

32% of sites experience an increase in crashes in the year after treatment due to random variation

Overcoming VariabilityOvercoming Variability• Large variability makes it difficult to

observe a change in crash frequency due to change in geometry at one site

• Large variability in crash data may frustrate attempts to confirm expected change

• Large databases needed to overcome large variability in crash data

• Statistics must be used to accurately quantify effect

Background ResearchBackground Research• National Research Sources

– Safety design guidelines• NCHRP Report 500: Guidelines for

Implementing the AASHTO Strategic Highway Safety Plan

– Vol. 5: Unsignalized intersections– Vol. 7: Horizontal curves– Vol. 8: Utility poles– Vol. 12: Signalized intersections– Vol. 13: Heavy trucks

• Volumes can be found at:

http://safety.transportation.org/guides.aspx

Background ResearchBackground Research

• National Research Sources– Safety evaluation tools

• Interactive Highway Safety Design Model

• Safety Analyst (forthcoming)

• Highway Safety Manual (forthcoming)

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

FHWA

FHWAFHWA

NCHRP

Background ResearchBackground Research• TxDOT Project 0-4703

– “Incorporating Safety into the Highway Design Process”

– Project Director: • Elizabeth Hilton

– Main products:• Roadway Safety Design Synthesis

(Report 0-4703-P1)• Interim Roadway Safety Design

Workbook (Report 0-4703-P4)

Available at: tcd.tamu.edu, click on “Products”

Facility TypesFacility Types• IHSDM

– Two lane highways

• Highway Safety Manual– Two lane highways

(& intersections)– Rural multilane

highways (& intersections)

– Urban streets (& intersections)

• TxDOT 0-4703– Freeways– Rural highways

• Multilane rural• Two lane rural

– Urban streets– Freeway ramps– Urban intersections– Rural intersections

Safety Prediction ProceduresSafety Prediction Procedures• Overview

– Six steps to procedure

– Evaluate a specific roadway segment or intersection (i.e., facility component)

– Same basic technique for all methods (IHSDM, HSM, TxDOT 4703)

• Output– Estimate of crash frequency for segment or

intersection

Step 1Step 1

• Identify Roadway Section– Define limits of roadway section of interest

• Limits of design project

• Portion of highway with safety issue or concern

– May include one or more components

Step 2Step 2

• Divide Section into Components– Analysis based on facility components

• One intersection or

• One interchange ramp or

• One roadway segment

– Each component analyzed individually in Steps 3 and 4

1

2

34

Homogeneous SegmentHomogeneous Segment

• Definition– A homogeneous segment has the same

basic character for its full length• Lane width

• Shoulder width

• Number of lanes

• Curvature

• Grade

• Horizontal clearance

Step 3Step 3• Gather Data for Subject Component

– Data may include• Roadway geometry (lane width, etc.)

• Traffic (ADT, truck percentage, etc.)

• Traffic control devices (stop sign, signal)

– What data do I need?• It depends on the component…

Step 4Step 4

• Compute Expected Crash Frequency– Use safety prediction model

• Model Components– Base model

– Accident modification factors

Volume Lane Width

Expected Crash Frequency

Base ModelBase Model• Relationship

– Cb = base crash rate × volume × length

– Injury (plus fatal) crash frequency

• Calibration– Analyst can adjust crash rate to local

conditions

• Application– Crash frequency for “typical” segment

– Typical: 12 ft lanes, 8 ft outside shoulder, etc.

Accident Modification FactorsAccident Modification Factors

• Definition– Change in crash frequency for a specific

change in geometry

– Adapts base model to non-base conditions

– One AMF per design element (e.g., lane width)

• Example: Two-lane highway

– Base condition: 12 ft lanes

– Roadway has 10 ft lanes

– AMF = 1.12

Steps 5 & 6Steps 5 & 6

• Repeat Steps 3 and 4 for Each Component• Add Results for Roadway Section

– Add crash estimates for all components

– Sum represents the expected crash frequency for the roadway section

• If there are multiple alternatives, repeat Steps 1 through 6 for each alternative

Questions?Questions?

More InformationMore Information

• Safety Resources from Project 0-4703– Workbook

– Synthesis

– Procedures Guide

– Texas Roadway Safety Design Software

• Web Address – http:// tcd.tamu.edu/documents/rsd.htm

– Also link from DES-PD site CROSSROADS

– Check periodically for updates

(Coming soon…)