Thesis Proposal

DETERMINING THE EXTENT AND CHARACTERISTICS OF OVERREPRESENTATION OF LARGE TRUCK CRASHES IN DAYTIME AND NIGHTTIME WORK ZONES

A thesis proposal by Naveen Mokkapati

Submitted to the office of graduate studies of Texas A & M University

In partial fulfilment of the requirements for the degree of MASTER OF SCIENCE

April 2007

Major Subject: Civil Engineering

Naveen Mokkapati Thesis Proposal

Page 1

INTRODUCTION

The population of United States is growing at a rapid rate resulting in an ever-growing increase in travel demand. However, the highway lane miles available to meet that demand have almost remained a constant over the last two decades. The statistics compiled by the Federal Highway Administration (FHWA) showed that vehicle miles of travel increased by 79 percent from 1982 to 2002, while the lane miles grew by 3 percent (1). The increased traffic demand accelerates roadway deterioration, leading to the need for repair and rehabilitation. According to the FHWA, roadway improvement projects covered an average of 23,745 miles of road per year from 1997 to 2001 (1). Furthermore, 3,110 work zones were active on the National Highway System (NHS) during summer of 2001. With the advent of new highway reauthorization bill ‘Safe Accountable Flexible Efficient Transportation Equity Act – A Legacy for Users’ (SAFETEA-LU) which provides a significant increase in the funding for construction and maintenance projects, work zones are expected to grow in the future years.

The effect of work zones on road users is significant. The FHWA estimates that 24 percent of non-recurring delay to the motorists is due to the presence of road maintenance and construction work projects (2). One of the FHWA surveys indicated that work zones are the second highest rated attribute for motorist dissatisfaction, the first being traffic flow (3). Unfortunately, delay and traffic congestion are not the only problems associated with increases in work zones. There are also safety related issues due to the presence of work zones. According to Fatality Analysis Report (FARS) and General Estimates System (GES) data, 1,028 fatalities and 41,000 injuries were reported in work zones for the year 2003 versus 693 fatalities and 36,000 injuries during 1997 (4, 5). This shows the increase in crashes in work zones over the recent years. Though this doesn’t prove that crashes will increase due to presence of work zones, it does give an inclination for the researchers to look into the safety effects of work zones on the road users. Various studies are available in the literature to corroborate that work zones, in fact, increase the crash likelihood of the driving public (6, 7, 8, 9, 10, 11, 12).

This proposed thesis will focus on large truck crashes, which could be more likely to be involved in a work zone related crash compared to automobiles due to their large size, limited


Page 2

maneuverability and narrow lanes in work areas. Furthermore, large truck crashes typically are severe in nature, with a good chance of being a newsworthy event, if fatalities are involved. Also, economic losses due to large truck crashes are expected to be higher than automobile crashes. As a result, it is desirable to deploy effective countermeasures to reduce truck crashes. The Federal Motor Carrier Safety Administration (FMCSA) has collected data and found that 24 percent of the fatal crashes occurring in the work zones are large truck crashes (13). With commercial trucks representing around 10.3 percent of all motor vehicles registered and 16.1 percent of total vehicle miles travelled, one might be inclined to say that the large truck crashes may be overrepresented in the work zones.

PROBLEM STATEMENT

Previous studies have indicated that large truck crashes are overrepresented in work zones. However, none of the studies could show if the overrepresentation is due to differences in exposure between large trucks and automobiles or greater crash risk for large trucks in work zones (14). Furthermore, there is a need to better understand the possible underlying causes behind the higher truck crash rate in work zones, if it truly exists. Finally, there is still limited knowledge on whether the characteristics of work zone involved truck crashes are similar to that of the total work zone crashes. Without a clear understanding of these characteristics and possible causes, it is not possible to identify the potential countermeasures that could be employed to reduce the large truck crashes in work zones.

RESEARCH OBJECTIVES

The main objective of this thesis is to identify appropriate countermeasures to reduce large truck crashes in work zones based on the extent and characteristics of such crashes. This will be achieved by assuming that the truck crashes are truly overrepresented in work zones. This study uses the data collected for National Cooperative Highway Research Program (NCHRP) project 17-30 ‘Traffic Safety Evaluation of Nighttime and Daytime work zones’. This data includes work zone project files and Highway Safety Information System (HSIS) crash data for 19


Page 3

construction or maintenance projects in North Carolina. The objective of this study is subdivided into different parts and listed below: •

Compare the truck and automobile crash characteristics in work zones with that of the non work zones.

•

Identify the possible reasons for any disproportionate change in truck to auto crash characteristics between work zones and non work zones.

•

Recommend potential countermeasures to reduce truck crashes in work zones using the list of possible reasons.

BACKGROUND

Over the past three decades, many studies were conducted to examine the effect of work zones on traffic safety (6, 7, 8, 9, 10, 11, 12). Most of them focused on long term work zone projects on freeway facilities. Typically, the studies were comparing the crash rates from the preconstruction and during work zone periods. The following are the general trends of the work zone characteristics observed by the earlier researchers from their studies: •

Crash rates increased during work zone operation compared to non-work zone period.

•

Rear-end and side swipe crashes occurred more frequently in work zones than non-work zones.

•

There were mixed results for the effect of work zone on crash severity. Some researchers (6, 10) have concluded that crashes were more severe in work zone, but others (7, 8, 9) indicated there is not much change in severity. Few others also found that work zone crashes were less severe than non-work zones (12).

•

Most crashes were observed to occur within the buffer and activity areas of the work zone.

•

Large truck crashes were overrepresented in work zone crashes.

The last of these truck crash trends is of primary interest for this thesis. The following is a discussion on some of the research, in chronological order, which examined the effect of work zones on truck crashes.


Page 4

In 1989, Hall et al. conducted a study analyzing the crashes in New Mexico construction zones for a 3 year period to improve the safety of the region (9). In spite of depending on crash report codes for identifying a work zone crash, the authors used the locations and durations of construction activity on rural state highways. Both the crash data and construction activity information were merged to obtain the crashes before and during the construction period. The results conclude that the truck involved crashes increased by an average of 44 percent compared to 23.8 percent increase for non-truck involved crashes. These results have to be interpreted carefully as the study did not use a control section accounting for the external factors.

Pigman et al. also did a similar analysis using accident data from 1983 to 1986 in Kentucky (10). They assumed work zone crashes as those with the contributing factor coded as “road under construction.” The results of this naive before and after study showed that the large trucks comprised 25.7 percent of work crashes, compared to only 9.6 percent of crashes outside the work zones.

In their study on truck related work zone accidents in New Jersey, Hong Lin et al. found that the frequency of truck accidents increased in work zones compared to non-work zones (7). However, when compared using each covariate, the researchers could not identify a significant difference in the severity and frequency of the truck crashes by roadway system, roadway character, day of week, time of day, weather condition, type of vehicles involved, lighting conditions, and the location of the first crash. But they found that sideswipe crashes followed by rear end crashes are the predominant collision types in work zones, especially at the time when lanes were closed. The researchers also indicated that 70 percent of truck accidents are due to driver’s inattention, followed by improper lane change and failure to yield right of way.

Studies by Daniel et al. (11) and Walker et al. (12) found the same result that the large truck crashes are overrepresented in work zones. A summary of the five studies is given in Table 1. The large variation in the truck crash percentages for different states might be due to geographic location affecting the work zone frequency, difference in the police crash reporting and the variation of work zone design guidelines and specifications.


Page 5

Table 1. Summary of truck related work zone crash studies Study

Year

Location

Hall et al.

1989

New Mexico

Conclusions on large truck crashes in work zones Work zone truck crashes increased an average of 44 percent compared to an average 23.8 percent increase for non truck involved crashes.

Pigman et al.

1990

Kentucky

Trucks represent 25.7 percent of work zone crashes, compared to only 9.6 percent of non work zone crashes.

Hong Lin et al.

1997

New Jersey

Trucks represent 25.9 percent of work zone crashes, compared to only 10.5 percent of total state-wide crashes.

Daniel et al.

2000

Georgia

Trucks represent 20.0 percent of fatal work zone crashes, compared to only 13.0 percent of non work zone fatal crashes.

Walker et al.

1999

Arizona

Trucks represent 5.5 percent of work zone crashes, compared to only 2.8 percent of non work zone crashes.

In 1999, Benekohal et al. conducted a survey in Illinois to identify truck drivers’ concerns on work zones and their assessment of work zone features (15). The survey indicated that nearly 90 percent of truck drivers find work zones to be more hazardous than normal road sections. Interestingly, half of the drivers admitted that they were exceeding the work zone speed limits. Furthermore, most drivers expressed their primary concerns in work zones as, the pavement edge drop off, loose construction materials, lack of shoulder, narrow lane width, and visibility and clarity of the flagger’s message. However, they considered traffic control devices to be helpful with impact attenuators as the most preferred while concrete barriers were the least.

Khattak and Targa used the North Carolina work zone crash data containing explicit codes for location, work activity state, type of activity and others in the year 2000 (6). They developed ordered probit models for the total harm, combining severity and frequency with economic values to injuries. The researchers identified three potential situations where most harmful truck crashes occur: firstly, at a full roadway closure requiring a detour on the opposite side; secondly, before the actual work area rather than in the advance warning area or adjacent to the actual work area; and finally, on the two-way undivided roads.


Page 6

In order to get a better perspective on potential reasons for truck crashes, the researcher has also looked at a few studies relating to truck safety on normal roadways.

In 2002, Pigman and Agent investigated the impact of trucks on interstate highway safety (16). They identified various countermeasures to improve truck safety based on discussion with trucking industry as well as Kentucky crash data for the years 1998 to 2000. Each of the countermeasures is grouped into roadway, truck and driver characteristics. For the roadways, the study recommended additional parking facilities, audible rumble strips, increased use of median barriers, ITS devices (like Changeable message signs (CMS), Highway Advisory Radio (HAR) and citizen band (CB) radio) for real time congestion and weather information, lane use restrictions for higher number of lane roads, and truck climbing lanes for steep upgrades. In the case of trucks, the study recommended proper rear end protection, adequate lighting and reflective material on the rear of the truck, and use of ITS technologies to warn drivers about closeness to an object, drowsiness, and any other impending dangers. Finally, in the case of drivers, the study recommended mandatory truck driving school providing important information on various frequently occurring hazards, use of seat belt, and strict laws on trucking companies to assign proper driving schedules.

Recently in 2006, FHWA and National Highway Traffic Safety Administration (NHTSA) jointly conducted a nation wide study of large truck crash causal factors (17). A sample data of 967 crashes, along with 1000 characteristics for each crash, has been collected at 24 sites in 17 states from 2001 to 2003. The results indicated that driver recognition and decision errors were the most common type of driver mistakes coded for both trucks and passenger vehicles. However, truck drivers have less frequent driving performance problems (e.g. asleep, sick, fatigue) than passenger vehicle drivers. Furthermore, the study found that, in crashes between trucks and passenger vehicles, fatigue is more frequent in passenger vehicle drivers and speeding in truck drivers. Finally, brake problems were found to be coded for 30 percent of truck crashes compared to 5 percent in passenger vehicles.

In summary, both the national data from FMCSA and state data confirm that truck crashes appear to be overrepresented in highway work zones. Though the earlier studies had their own


Page 7

limitations in proving the overrepresentation, the researcher assumes that truck crashes are higher in work zones for the purpose of this study. Furthermore, there is a limited research done on identifying the potential concerns of truck safety in work zones. This thesis will try to identify these concerns and provide appropriate countermeasures to reduce work zone truck crashes.

RESEARCH STUDY TASKS

In order to achieve the aforementioned objectives, the researcher will complete the following tasks:

1. Review of the Past Literature: The researcher will conduct an in-depth study of the past research done on truck crashes in work zones. To be more specific, the researcher will place emphasis on identifying the flaws in the earlier research and try to avoid them in this thesis work.

2. Study Sites: This study will focus on both urban and rural freeway facilities with long term work zone projects (duration greater than 2 months) completed in the past 5 to 6 years. The sites will be selected so as to encompass both day and night work activities. This provides the opportunity to see the change in truck crash characteristics due to work zones at both day and night type of projects.

3. Data Collection: There are three sources of data for this study:

a. Field Data Collection: As part of NCHRP Project 17-30, North Carolina Department of Transportation (NC DOT ) has been contacted and potential work zone projects were identified (18). Furthermore, the project diaries, traffic control plans and other documentation have been reviewed and the following necessary data have been extracted: •

beginning and end mile point limits of the project

•

project start and end dates

•

general type of work performed

•

daily information on o when work actually occurred


Page 8

o number and direction of lanes closed

b. HSIS Data: From the HSIS database, crash, vehicle and roadway inventory data will be obtained for each of the projects using the mile points for the years 1995 to 2004. The obtained HSIS mile points will be matched with that of the project section limits.

In order to identify the control sites with similar roadway and traffic conditions for each of the work zone projects, the researcher will select roadway sections upstream of the work zone on the same highway to minimize the traffic volume pattern changes, weather effects, etc. The researcher believes that the downstream control section might not work well, as the traffic volume through the downstream section depends on the work zone bottleneck. Furthermore, in order to ensure queues from the work zone and the crashes associated with them not extend into upstream control sections, work zone coded crashes in HSIS data will be checked. If there are unusually high number of work zone crashes in a selected control section, that site will not be used for further analysis and new control sites will be explored. Finally, a geometric check will be conducted to make sure that the numbers of lanes do not change in both control and treatment sections as well as before and during periods. If a geometric change is observed, the corresponding roadway sections will not be used for the time period when the number of lanes is changed.

For the purpose of requesting data from HSIS, a list of work zone projects will be prepared with control and work zone section limits along with the specified years of interest. From the HSIS data obtained, the researcher will extract the following necessary data: •

Annual Average Daily Traffic (AADT)

•

Vehicle type

•

Number of lanes

•

Severity of crash

•

Number of vehicles

•

Surface Condition

•

Road Curvature

Naveen Mokkapati Thesis Proposal •

Lighting Condition

•

Weather

•

Contributing factors

•

Manner of collision

Page 9

4. Data Reduction and Analysis: First, the crash data will be reduced into before and during periods as well as control and work zone sections. Then, each crash from the HSIS database will be matched with that of the work zone project information and following accident, roadway and work zone variables will be identified and categorized:

I. Accident Characteristics

a. Vehicle Type: In order to study the differences between change in crash likelihood due to work zones for trucks and automobiles, the variable ‘vehicle type’ will be obtained for each of the crashes showing whether the crash involves truck or not. Therefore, there will be two vehicle types: truck and non-truck. Using the HSIS codes for different vehicle types, the researcher has defined a truck as any of these vehicles: single unit truck with two or more axles, truck/trailer, truck/tractor, tractor/semi-trailer, tractor/doubles, and unknown heavy truck.

b. Time of Day: The researcher believes that the crash likelihood changes largely with the time of the day. Keeping in mind the sample size restrictions, three times of day are selected in this study and defined in the Table 2. Table 2. Three categories for time of day Day Night Twilight

Start time Sun rise time + half an hour Sun set time + half an hour Sun set time - half an hour Sun rise time - half an hour

End time Sun set time - half an hour Sun rise time - half an hour Sun set time + half an hour Sun rise time + half an hour

It is a well know fact that the sunrise and sunset times will change every day. But it would be hard to estimate these timings each day and separate the crashes into day, night and twilight categories. So, an average sun rise and sun set times will be identified for the 19 work zone projects for each month using the U.S. Naval Observatory website (19). Then, crashes will be


Page 10

divided into day, night and twilight times based on different sun rise and sun set timings of each month.

c. Severity: A typical KABCO scale (Killed, A injury, B injury, C injury, Only property damage) will be used in this study. Though the Property Damage crashes (PDO) are not reported consistently, they will still be used in this analysis because of the small sample sizes. The researcher found from Highway Safety Research Centre (HSRC) that NC DOT has changed its PDO crash threshold from $500 to $1000 in 1996. As the analysis period for this study is 1995 to 2004, there would be higher PDO crashes in the year 1995 which creates uncertainty in the analysis. Therefore, the researcher will conduct separate analysis with and without PDO crashes. If the results with and without PDO crashes are similar, then PDO crashes will be used in the analysis. But if they are different, non-PDO crash analysis results will be used for further conclusions.

d. AADT per lane: Many earlier studies showed that the AADT of a roadway has a large impact on crash likelihood. Therefore, the researcher will use the AADT per lane as a categorical variable. The criteria used for categorizing will be based on volume to capacity ratio, by considering both over-saturated and free flow conditions.

e. Number of Vehicles: According to Rouphail et al (20), multi-vehicle crashes, particularly rear end crashes are overrepresented in work zones compared to normal roadway sections. This trend is attributed to the increased speed variations between the lane closure and upstream segments. For this study, this variable will be categorized into auto – auto collision, truck – truck collision and truck – auto collision, and collisions involving more than two vehicles.

f. Surface Condition: The researcher believes that surface condition affects the braking characteristics of a vehicle. A wet road will have higher braking distance compared to a dry road. This distance will increase even more during ice and snow conditions. Furthermore, work zones will not be active during rain, ice or snowy conditions. In this study, three ‘surface condition’ categories will be taken: dry, wet and ice/snow.


Page 11

g. Road Curvature: Horizontal and vertical curves have a negative impact on operational characteristics of trucks. When these curves are present in a work zone, the effect on trucks might be worse. Therefore, the researcher will take the roadway curvature as a covariate and see how truck and auto crashes behave with different roadway curvatures. This study considers four categories for ‘road curvature’: Straight-level, Straight-grade, Curve-level, and Curve-grade.

h. Lighting Condition: It is evident that a lighted roadway will improve the safety at night. However, it would be interesting to see whether this improvement has a similar proportion of effect on trucks and automobiles. For this purpose, this variable will be categorized into two subsets: lighted roadway at night and roadway not lighted at night.

i. Weather: Adverse climatic conditions increase the crash frequency, especially for trucks. Trucks typically have a long haul. Therefore, it is hard for the drivers to plan ahead for the weather changes. This study considers 5 categories for weather: Clear, Cloudy, Rain, Snow, and Sleet/Hail.

j. Manner of Collision and Contributing factors: As the primary objective of this study is to identify the underlying causes for the increase in truck crashes during work zone, both the variables ‘manner of collision’ and ‘contributing factors’ will be obtained for each of the truck crashes. The researcher will specifically look at crashes involving rear end, side swipe, run off the road, fixed object and angle crashes while the contributing factors considered will be speeding, improper lane change, careless driving, failure to yield Right of Way (ROW), follow too closely, operating defective vehicle, disregard traffic control, improper passing, and alcohol use. Unfortunately, some of the contributing factors like driver inattention, visibility obstructed, over steering, have been coded starting from the year 2000. As the analysis period for this study begins from 1995, it is not possible to use these contributing factors. But the researcher can still use these factors for projects starting and ending in 2003 and 2004. As the sample sizes are limited, the results obtained for these factors might not be fully accurate. All the above categories were selected because the earlier studies have identified them as the major crash types and contributing factors in work zones.


Page 12

II. Roadway Characteristics

a. Speed Limit: Earlier studies observed that speeding is the most significant contributing factor in work zone crashes. Therefore, selecting an appropriate speed limit and providing adequate enforcement to ensure that drivers follow that speed limit is important. An unusually low speed limit should be avoided so as to maintain the credibility of drivers on work zone traffic signs. The researcher will attempt to identify whether the reduction of speed limit has a varied effect on trucks compared to automobiles. Speed limits are categorized into four subsets: 55 mph, 60 mph, 65 mph, and 70 mph.

b. Number of lanes: For a higher number of lanes, sideswipe crashes will increase due to more points of contact between vehicles. This trend is expected to be more in trucks due to their massive size and limited maneuverability. This study considers three categories for ‘number of lanes’: four lanes, six lanes, and eight lanes.

c. Functional class: It is not unusual to expect that crash frequency and characteristics differ in urban and rural freeways. Some of these differences can be attributed to higher traffic volumes, more lanes, and better roadway lighting conditions in urban areas compared to rural. Therefore, the crashes will be analyzed separately for urban and rural freeways.

III. Work Zone Characteristics

a. Work Zone Type: Each work zone project will be divided into two major types, based on the Ullman et al. study (21). The first category is that of a road work, where temporary traffic control devices used for lane closure will be removed after completion of work activity. The majority of the projects in this category are typical pavement repair and rehabilitation works. On the other hand, the second category contains projects involving major reconstruction and pavement widening. For these projects, long term roadway geometric changes like lane shifts, median crossovers, and shoulder, ramp or acceleration/deceleration lane closures, etc. exist throughout the project. In statistical terms, this categorical variable has two possible values: projects with and without long term geometric changes.


Page 13

b. Work Activity: It is of particular interest for this study to verify whether the change in crash likelihood for trucks is due to the work activity or the change in work zone geometrics (lane shifts, median crossovers, and shoulder, ramp or acceleration/deceleration lane closures). In order to analyze these effects separately, the ‘work activity’ variable is being used. It has two possible values: •

work is occurring at the time of crash

•

work is inactive at the time of crash

The researcher believes that the location of the work activity plays a vital role in the crash occurrence. In fact, work activity on the traffic side of a barrier can have significant impact on the traffic flow compared to the work activity behind a barrier. However, due to lack of traffic control plans for each of the work zones, it is hard to identify precisely the location where a work activity is taking place. But the researcher does have information on whether the work zone is a major reconstruction or a pavement rehabilitation type of project. And it is widely believed that most of the major reconstruction projects will have a barrier, with work activity being done behind the barrier. On the other hand, pavement rehabilitation projects typically do not have a barrier and work activity can sometimes take place on the traffic side, which adversely affects the traffic flow. The researcher will compare the proportions of truck and auto crashes in work zones from that of non work zones, separately for major reconstruction and pavement rehabilitation projects. The comparison of conditional odds ratio between both these types of projects will indicate the significance of safety issues when work is being done on traffic side of the barrier.

c. Number of lanes closed: A lane closure in a work zone has an adverse effect on traffic. Therefore, lanes are typically closed at night when the traffic volumes are lower. Irrespective of whether the lanes are closed during day or night, travellers have to change their normal path and merge or cross over in order to traverse through the work zone. This has an additional work load on drivers and leads to higher crash likelihood. Moreover, the direction of lane closure is important; as it gives a better understanding of whether the crash has occurred due to the presence of a work zone lane closure. Hence, this study will categorize this variable into 2 subsets: no lane closure and lane closure, separately for both the directions.


Page 14

5. Methodology: A two by two contingency analysis will be used for this study. For each of the crash characteristic, two contingency tables will be developed separately for work zone and control sections, as shown in Tables 2, 3.

Table 2. 2*2 Contingency table for work zone section Trucks

Autos

Work zone section - before

n111

n112

n11+

Work zone section - during

n121

n122

n12+

Total

n1+1

n1+ 2

n1

Table 3. 2*2 Contingency table for control section Trucks

Autos

Control section - before

n211

n212

n21+

Control section - during

n221

n222

n22+

Total

n2+1

n2+ 2

n2

Firstly, Breslow day (B-D) test will be conducted to verify the homogeneity of odds ratios. The odds ratio for Table 2 can be defined as the odds of having a truck crash in before work zone period to the during work zone period. The null hypothesis for B-D test can be written as (22): OR 1= OR 2

where OR 1 =

n111 * n 221 n * n 222 and OR 2 = 112 n121 * n 211 n122 * n 212

The breslow day test helps us know whether the higher odds ratio is due to the presence of work zone or other external factors.


Page 15

If the null hypothesis of the Breslow day test is not rejected, then Cochran Mantel Haenszel (CMH) test will be conducted to test whether the conditional odds ratio of the work zone section (accounting for external factors) equals 1. The null hypothesis for CMH test can be written as: Conditional odds ratio,ψ 0 = 1

If the p-value of CMH is