IDENTIFICATION OF HOTSPOT LOCATIONS ALONG THE I-95 JFK MEMORIAL HIGHWAY. Fathy Elgendi, Master of Science, March 2015

Transcription

1 Title of Thesis: IDENTIFICATION OF HOTSPOT LOCATIONS ALONG THE I-95 JFK MEMORIAL HIGHWAY Fathy Elgendi, Master of Science, March 2015 Thesis Chair: Celeste Chavis, Ph.D. Department of Transportation and Urban Infrastructure Studies This study identifies the locations and causes of crashes along the I-95 JFK Corridor between Baltimore and the Delaware border. After identifying hotspot crash locations, this study determines the major crash types, explores the relationship between roadway geometry and crashes, and determines other contributing factors to speed and non-speed crashes. Key segments are analyzed by crash type and predominant causes. Possible solutions in reducing these crashes and improving safety are then offered. Through an analysis of speed profiles, the study found that there was no correlation between average vehicle speeds and the rates of accidents. Other factors such as roadway geometry and variance in speed were more significant. The actual crash rates were lower than those which are predicted using the Highway Safety Manual predictive analysis methodology.

2 IDENTIFICATION OF HOTSPOT LOCATIONS ALONG THE I-95 JFK MEMORIAL HIGHWAY by Fathy Elgendi A Thesis Submitted in Partial Fulfillment Of the Requirements for the Degree Master of Science MORGAN STATE UNIVERSITY May 2015

3 IDENTIFICATION OF HOTSPOT LOCATIONS ALONG THE I-95 JFK MEMORIAL HIGHWAY by Fathy Elgendi has been approved March 2015 Celeste Chavis, Ph.D., Chair Young-Jae Lee, Ph.D. Anthony Saka, Ph.D. ii

4 In dedication to my family for their love and support. iii

5 I would like the take the time to thank my advisor Dr. Celeste Chavis for the long hours dedicated to reviewing my paper and making sure that all edits and information were accurate and correct. I really appreciate all she has done to help me. I would have never been able to complete this thesis without her guidance. In addition, I would like to thank Dr. Young-Jae Lee for his support and guidance and for making sure that I incorporated a relevant methodology to improve my project as well as for all his support throughout the completion of my undergraduate and graduate studies. I would like to thank Dr. Anthony Saka for his help and support in guiding me through the program. I would like to thank Dr. Jeihani for providing me with all the support and guidance I needed while I was in this program. Also, I would like to thank Ms. Alice Williams for addressing all my needs and being very helpful anytime I needed it. I would like to thank my parents Hossam Abdelfatah and Azza Suliman for motivating me and helping me to finish my studies. iv

6 List of Tables....iix List of Figures... ix List of Abbreviations... xi 1 Introduction Overview Facility Overview Problem Statement Objectives Research Questions Thesis Outline Literature review Speed limit compliance Surface conditions Roadway Geometry Safety Analysis Studies Effect of Speed on Safety and Crashes Methods on Identifying Hotspot Crashes Highway Safety Manual Methodology Corridor Analysis Overview v

7 3.2 Crashes by Time of Day AM Peak PM Peak Off Peak AM Peak Speeds (7AM-9AM) PM Peak Speeds Off Peak Speeds (9AM-11AM, 1PM-3PM, 8PM-10PM) Effect of Drop in Speed at Toll Plaza in the Northbound Direction Site Analysis MD-155 to MD Crash Types Weather and Seasonality Crashes by Time of Day Analyzing Hotspot Crashes using HSM ( ) MD-222 to MD Crash Type Weather and Seasonality Crashes by Time of Day Summary and Conclusions MD-155 to MD MM 90.9 to MM MM 91.9 to MM vi

8 5.2 MD-222 to MD MM 94.7 to MM 95.2 (Winch Road Overpass) Recommendations Study Implications Study Limitation & Future Work References Appendix A JFK Facility Map vii

9 Table 1: Daily Crash Rates by Segment Table 2: Crash Rate by Time of Day Table 3: Average Speeds by Segment during the AM and PM Peak Table 4: Average Speeds by Segment during the Off Peak Table 5: F-Test of Speeds Before and After the Toll Plaza Table 6: T-Test (Equal Means) of Speeds Before and After the Toll Plaza Table 7: Rear-end Crashes by Probable Cause Table 8: Probable Causes of Fixed Object Crashes along MD-155 to MD Table 9: Calculated Crash Rates using 2013 Crash Data Table 10: IHSDN Analysis with the Tydings Bridge Table 11: Predicted vs. Actual Crash Severity Table 12: Types of Collisions Along MD-222 to MD Table 13: Probable Causes of Crashes along MD-222 to MD Table 14: Probable cause of Wet Crashes viii

10 Figure 1: Crash Rates along the JFK Corridor Figure 2: Crash rates during AM Peak by Direction Figure 3: Crash Rates during the PM Peak by Direction Figure 4: Crash Rates in the Off-Peak by Direction Figure 5: Northbound AM Peak Speed Runs Figure 6: Southbound AM Peak Speed Runs Figure 7: Distribution of AM Peak Speeds in the NB Direction Figure 8: Distribution of AM Peak Speeds in the SB Direction Figure 9: Northbound PM Peak Speed Runs Figure 10: Southbound PM Peak Speed Runs Figure 11: Distribution of PM Peak Speeds in the NB Direction Figure 12: Distribution of PM Peak Speeds in the SB Direction Figure 13: MD-155 to MD-222 Segment with Toll Plaza Approach Figure 14: Northbound Off-peak Speed Runs (9AM-11AM) Figure 15: Southbound Off-peak Speed Runs (9AM - 11AM) Figure 16: Northbound Off Peak Speed Runs (1PM-3PM) Figure 17: Southbound Off-peak Speed Runs (1PM-3PM) Figure 18: Northbound Off-peak Speed Runs (8PM-10PM) Figure 19: Southbound Off-peak Speed Runs (8PM-10PM) Figure 20: Frequency of Crashes along MD-155 to MD Figure 21: Map of Segment MM Figure 22: Map of Segment MM Figure 23: Crash Types along MD-155 to MD Figure 24: Location of Rear-end Crashes Along MD-155 to MD Figure 25: Horizontal Curve at MM ix

11 Figure 26: MM NB Figure 27: MM SB Figure 28: Type of Crashes during Wet Conditions along MD-155 to MD Figure 29: Probable Causes of Crashes in Wet Conditions along MD-155 to MD Figure 30: Distribution of Wet Crashes along MD-155 to MD Figure 31: MM Figure 32: MM 91.1 Tydings Bridge Figure 33: Crashes by Time of the Day along MD-155 to MD Figure 34: Nighttime Crash Types along MD-155 to MD Figure 35: Probable Nighttime Crashes along MD-155 to MD Figure 36: Distribution of Night Crashes along MD-155 to MD Figure 37: Scope of HSM Study Area Figure 38: Map of Segment between MD-222 and MD Figure 39: Distribution of Crashes along MD-222 to MD Figure 40: Rear-end Crashes Cluster at MM Figure 41: Location of Followed too Closely Crashes along MD-222 to MD Figure 42: Too Fast for Conditions Crashes along MD-222 to MD Figure 43: Crashes by Time of Day along MD-222 to MD Figure 44: Types of Nighttime Collisions along MD-222 to MD Figure 45: Distribution of Nighttime Crashes along MD-222 to MD Figure 46: SB at MM Figure 47: NB at MM Figure 48: Corridor Map of JFK Memorial Highway (I-95) x

12 AASHTO ADT CMF CR DE DOT FXOBJ HSM JFK MD MM NB PARKD PDO RREND SB SDSWP VMT American Association of State Highway Transportation Officials Average Daily Trips Crash Modification Factor Crash rate (crashes per 100 million vehicle- miles traveled) Delaware Department of Transportation Fixed-Object Crash Highway Safety Manual John F. Kennedy Maryland Mile Marker Northbound Crash with parked vehicle Property Damage Only Rear-end Crash Southbound Sideswipe Crash Vehicle Miles Traveled xi

13 Highways are designed to move vehicles as efficiently as possible while maintaining minimum safety standards. The State Department of Transportations (DOTs) utilizes a design policy in which a selected speed is used to determine the various geometric designs (Kanellaidis, 1996). The geometric design of highway facilities is based on the desired operating speed. Speed limits, on the other hand, are not set at design speed, but rather are based on the speeds at which drivers actually travel. This is a function of drivers perception of safety along a particular roadway segment. Roadway segments with horizontal or vertical curvatures are often controlled by advisory speeds and warning signs to maintain drivers safety on the roadway. In the future, these roadway segments will need to be mitigated in order to meet current roadway standards and reduce travel times for drivers. Safety analysis study is important for understanding a driver s perception of the roadway geometrics during different times of the day as well as for assessing whether a driver s behavior and perception changes during different time periods and under different conditions on the freeway. This gives transportation agencies and policy makers the opportunity to evaluate systemic improvements and mitigations that could be implemented to reduce the possibility of crashes occurring in the segment. The general concern is the relationship between crash types, probable causes, speed, volume, density, and road conditions. As speed increases, drivers perceived reaction time also increases Charles (2001). Kockelman (2010) found that when the posted speed limit is near that of the 85th percentile, the variance in speed is lower and therefore there are less speed related crashes. Due to drivers desires to travel at or above the speed limit, an increased speed limit has been shown to result in a higher speed in the 85th percentile. In fact, studies 1

14 have shown that the 50th percentile operating speed is typically at or above the posted speed limit (Garber, 1988). Harwood (2000) found that nearly 50% of state and city traffic engineers use design speed as the posted speed limit for new designed highways until there is a measure of the 85th percentile speed, which determines the newly posted speed limit. However, there may be safety implications when the posted speed is significantly higher than the design speed. This study will investigate the factors associated with speed related crashes along the John F. Kennedy (JFK) Memorial highway. The John F. Kennedy Memorial Highway is a section of I-95 that runs from northern Baltimore City to the Delaware border. A map of the facility is provided in Appendix A. The segment is a 50-mile section with tolls collected in the northbound direction. JFK Highway began operating in November Prior to this, US 40 was the only northbound/southbound route through Maryland. Today the John F. Kennedy Memorial Highway carries approximately 29.3 million vehicles per year. Tolls are only collected in the northbound direction at the toll plaza located one mile north of the Millard Tydings Memorial Bridge over the Susquehanna River in northeast Maryland. The corridor connects travelers who are traveling from Washington D.C and Baltimore to Philadelphia and New York City. The JFK corridor includes two service areas: the Maryland House located in Aberdeen, Maryland and the Chesapeake house located near the Susquehanna River. The Maryland Transportation Authority Police operate the I-95 Commercial Vehicle Weigh Station located at JFK toll plaza. In order to reduce crashes and improve safety, one needs to examine, compare, and cross classify collision types with probable causes, geometric alignments, and 2

15 roadway conditions that have been shown to contribute to crashes. Further examination is needed for speed related crashes where these crashes are occurring and roadway geometrics may be contributing to the speed related crashes. A study by Fitzpatrick (2003) of 40 state DOTs participating in a mail survey found that thirty-eight percent of agencies use anticipated speed limits while fifty-eight percent use legal speed limits in AASHTO to select design speed. The anticipated speed limits are better due to a higher compliance and less variance in speed limits. Due to states using the speed limit that does not represent most drivers, the geometric design is much less than what is desired Fitzpatrick (2003). The anticipated speed limits are better to use due to a higher compliance and less variance in speed limits. The purpose of this study is to identify how speed limits contribute to crashes as well as determine any differences in roadway geometry that may be leading to these crashes. Correlation between safety related characteristics and speed related crashes will be identified in order to determine possible solutions that will improve safety and reduce speed and non-speed crashes. These improvements may be a combination of corridor wide improvements and segment improvements depending on the characteristics of the segment. The study focuses on the operational effects of geometric alignment and customers compliance to the speed limit in order to investigate the influence of operational effects on highway safety. The goal of this research is to identify primary segments of the I-95 JFK corridor where a high frequency of accidents occur, determine predominant crash types, identify probable causes, and compare the primary segments to identify trends that might be occurring based on season, time of day, light conditions, geometric conditions, peakvolumes, and surface conditions. The objective is to identify how speed, geometric design, and other related factors are contributing to crash occurrences. The study will 3

16 focus on speed related crashes; crashes solely attributed to driver error are outside the scope of this study. There are many speed related reasons for crashes that may be a function of roadway geometry, weather, and time of day. An evaluation of the causes needs to be conducted in order to identify what improvements need to be made to reduce crashes along the corridor. According to HCM (2000), a reasonable speed limit is one at or below which the majority of drivers (85%) desire to travel. When the 85th percentile speed and posted speed limit are quite different this may increase the probability of accidents as drivers may deem the speed limit unrealistic and thus may travel at faster speeds. This study will identify sections of the JFK Highway where a high frequency of crashes occur, determine probable causes, and propose solutions. The following research questions will be addressed. 1. At what locations along the JFK Corridor is there a high probability of accidents? 2. What are the predominant crash types and probable causes for speed related crashes at various locations along the corridor? 3. How does speed and speed limit compliance affect the likelihood of crashes? 4. What improvements may be implemented to reduce the number of speed-related crashes? 5. How can the results for the JFK Highway be generalized to other highway corridors? (This last question will offer an answer related to the transferability of the study). Based on these questions, the thesis is organized as follows. Chapter 2 will present relevant literature. The study will explore numerous studies conducted to 4

17 identify different methods used to discover hotspot crash locations throughout the segment. Chapter 3 will examine the entire JFK corridor to determine crash rate trends as a function of time of day, location, and speed. Chapter 4 will review two hotspots in detail in order to identify segments that have a concentration of reoccurring crashes and to determine the probable cause for these crashes. Lastly, Chapter 5 will provide recommendations for facility improvements and future work. 5

18 The literature has shown that crashes may often be attributed to several factors: driver error or neglect, drivers speed profiles, surface conditions, and roadway geometry. As this study focuses on the geometric and operational causes of crashes along the JFK Corridor, everything with the exception of driver error will be discussed in the following sections. Studies have shown that the 85th percentile speed alone is not a determinant of accident rates. The Texas DOT found that raising speed limits did not increase the number of crashes (Parker, 1992). However, increased speeds cause accident severity to rise. Larger variances in speeds occur due to vehicles incompliance with posted speed limits, which raises accident rates on freeways. Therefore, an ideal speed limit would be one that users follow and one that is safe under various roadway conditions and geometrics. According to Parker (1992), lowering speed limits did not have an effect on safety due to a lack of compliance. As Wilmot (1999) noted, motorists usually drive at a speed they feel is safe for them and that newly posted speed regulations will not necessary reduce speeds and increase safety. A review of the literature shows that there is an inconclusive relationship between accident rates and speed limits. The goal of determining the right speed limit must adhere to improving safety and the flow of traffic. Therefore, this research will determine what factors have an effect on the speed of vehicles, and in turn, the likelihood of accidents. Speed limits should be based on geometry, traffic density, and volume. Speed variation is a major cause of crash occurrences and a determinant of crash severity (Kockelman & Jianming, 2010). A Furthermore, Futheringhan (2008) stated that a higher speed variance causes unsmooth freeway operation due to the difference in the speeds of road users and types of vehicles on the roadway. It is fundamentally 6

19 important to reduce speed variation in order to reduce accidents. Parker (1992) stated that changing speeds by increasing the speed limit to 10 mph above the 85th percentile or reducing the speed limit to 10 miles below the 85th percentile does not reduce the variance of speeds and thus does not increase or decrease accident rates. According to Lee (2006) crash rates were minimal near the average speed of traffic. Lee observed that when speed variation increased, crash rates increased. Therefore, change in speed is not a cause of vehicle accidents along a freeway. Consequently, if a state wants to reduce accident severity and the likelihood of crashes, then reduced speed limits must be coupled with additional law enforcement to help reduce the severity of crashes and improve safety. Drivers risk being involved in a crash when driving in wet conditions. According to Pisano ( 2008), over 87% of surveyed drivers stated that they reduce their speeds by 11% during rainfall. This means that drivers do not feel safe driving at the normal speed when it is raining. According to Smile (2009) crashes occurring in wet conditions are higher by 70% than under normal conditions. It is very important to understand that rain reduces the ability of drivers to control vehicles and the ability of drivers to stop vehicles. Studies have shown that speed-flow density relationship has shown a capacity reduction in heavy rain (Khattak, 1997). Wet conditions, along with geometric conditions, may cause crashes to occur. Visibility could be reduced due to heavy rain conditions, which may cause more crashes. Heavy rain conditions, along with drivers behavior, could increase the likelihood of crashes. Slippery road conditions are dependent on raining and the low temperatures during rain that may increase the probability and risk of crashes. According to Kazruzzaman (2013), wet conditions do not affect crashes and vehicles skidding off the roadway; however, rainfall causes visibility issues for drivers who are driving in the segment. On the other hand, 7

20 Mannering (1996) stated that wet surface conditions increase severe injuries for females and older men in crashes under wet conditions. This explains that rain only reduces the visibility for drivers and may cause accidents. If this is added to driving during nighttime when visibility issues occur due to the reduction of light, then it can be concluded that crashes under wet conditions occur at nighttime more than those occurring during the day (Jiang, 2012). Another important element is the contribution of roadway geometry to the visibility of roadway during wet condition that may be causing a higher chance of crashes to occur. Rutting is more hazardous in wet weather when there is an accumulation of water during the rut path; this may lead to hydroplaning (Jiang, 2012). Using the ZIMB model to estimate the safety effects of roadway geometric design and traffic features of interest, it was shown that crash frequency increases as AADT and segment length gradually increase (Jiang, 2012). The road segments with more than two lanes have lower frequency to crash than the segments that only have two lanes. Also, the width of shoulders outside and median shoulders do not impact the number of vehicle crashes. When width of travel lanes increased the crash rates were much lower than what they are currently. Mannering (1996) stated that vertical and horizontal alignments are critical elements that predict highway safety. This means that crashes are more likely to occur on freeway sections that have horizontal and vertical crest curve than a straight freeway segment with no curvature. Vertical and horizontal curves may limit a safe stopping sight distance on a freeway segment. A study by Mannering (1996) showed that sixtynine percent through lane crashes occurring in an expressway located in Dallas were happening on the vertical crest and sag curve. They did not find any direction correlation between crashes and vertical curves. However, they did come with a 8

21 conclusion that sight distance was a major leading cause of high crash rate on the freeway. It was also found that horizontal curve accident has a much higher crash rate than when the freeway segment is straight and tangent. The crash rate ranged from 1.5 to four times higher than straight roadway. Other factors may contribute to crash occurrence such as rainfall that will cause additional restraints that may restrict drivers visibility. A horizontal curve results in a high occurrence of sideswipe, fixed-object, and night time crashes than those on level terrain segments (Hummer, 2010). It is evident that these crashes are due to limited sight distance and drivers not being able to determine roadway characteristics due to the change in roadway geometry. Horizontal curvature multiplies the causes of crashes occurring at the segment when it is already existing if the segment was a tangent segment. It has a multiple effect on the amount of crashes and probability that they would occur on the changing geometry. Weather conditions will add to the problem and will cause crashes to occur often is vehicles do not reduce speed at these conditions. According to Hadi (1995) lane widening had the highest benefit on freeways causing reduction in the frequency of crashes. Therefore, when lanes are narrow due to shifting during construction it could possibly lead to higher crash occurrence on workzone segments. In addition, increasing highway paved shoulder width decreased the rates of crashes occurring on freeway segment. This is evident at work-zone sections where Jersey barriers protecting construction zones often eliminate shoulder width and remove some of the lane width that causes a higher rate of crashes. The higher Average Annual Daily Traffic (AADT), the higher the frequency of crashes will be. This is why it is important to monitor crash rates during peak hours with high congestion. Areas that have limited shoulder width, lane width, and higher AADT need to be evaluated for trends of crashes occurring at segments with these characteristics. 9

22 The larger the horizontal curve radius, the safer the roadway (Bonneson, 2009). Drivers are more capable of seeing the distance ahead of them. When the radius of the horizontal curve is large, it reduces the probability of vehicles sliding or rolling over. However, the curve radius does not necessarily change operating speeds. A change in the speed limit rarely influences operating speeds. Designers need to consider the embankments and retaining walls when measuring sight distance. The Accident Modification Factor could be used to evaluate the safety performance of the roadway. Other ways geometric design is examined is when design changes take place in a given segment of the road. The changes are examined as a before and after study to capture the changes in the number of crashes and safety effects. Before and after studies can help find the AMF, which will give a description of the crash frequency and the design factor. The AMF should not be used to change the speed limit; however, it should be used to identify changes in drivers operating speeds. The effect of crash counts could be dependent on the design and size of different element such as the widths of lanes and shoulders. The cross section study is a common technique to compare frequencies in crashes that have different size range and ways to control the variances in segment lengths or volumes. Accident Modification Factor (AMF) could be used to evaluate the relationship between design geometric factors corresponding to accident severity under certain values and criteria. The Accident Modification Factor should be a factor of 1 with the expectation that it is calibrated in a certain local area. AMF is used in order to conduct an evaluation that will determine the relationship between the degree of design change and the change in the frequency of crashes with severities such as injuries and fatalities. 2.4 Measures of safety are categorized under two different groups. The first group is different factors that are a contributing cause to crashes and that include road factors, 10

23 vehicles, and road users. The other group includes changes in roadway and safety characteristics as well as driver risk and exposure. One major reason that these crashes are occurring is due to reliability statistics. Studies have shown that countries and municipalities that use reliability statistics have noticed a reduction of risk over time (Haque, 2012). The possible solution to improving safety is to reduce the exposure of high-risk vehicles and only allow low-risk vehicles to use the roadway system. Other studies have addressed the number of crashes that occurred involving tractor-trailer trucks and the severity and risks of these crashes. A study by to Lemp (2011) which emphasizes all crashes involving trucks (10,000 pounds+) with passenger vehicles showing the effect of the crashes showed that, at the end of 1999, crashes involving large trucks have caused nearly 4000 deaths involving cars and 750 deaths involving trucks. In addition, multi-vehicle truck crashes were shown to have a very close relationship to the number of trucks on the highway segment when the crash occurred (Srinivasan, 2007). The single vehicle truck crashes occur parallel to the number of trucks on the highway segment during hours when it is dark outside. Speed has a risk of causing crashes however it is not a major contributor to crashes. The reason speed crashes are often under reported is due to when police arrive after a crash occurred they do not usually state that the driver was operating at a high speed because of it being nearly impossible to determine the operating speed during the crash. They could report certain actions that may have caused the crash. According to Graettinger (2005), GIS may be used to identify hotspot crashes occurring along the segment. Also, it is used to identify the pattern of crashes occurring at the segment in relation to type period when crashes occur, the severity of crashes occurring at the segment, the location of crashes occurring at the segment and the level 11

24 of severity of crashes occurring at the segment. GIS was used to analyze hotspots with crashes of high severity and used that location to find if crash and speed are related to one another. According to Xiaobo (2014) the land transportation authorities focus on reducing the number of crashes with high-risk and not the overall frequency of crashes; this is due do a limitation in budget. Safety studies conducted on freeways in Singapore have shown that crash frequencies in the Central district area were high, but severity of crashes was low due to lower speeds in higher volumes at the CBD. The Singapore Land Transportation authority has broken crashes into three categories of severity: slight injuries, serious injuries, and fatalities. The Singapore Land Transportation authority has used an approach called the Hotspot identification (HSID) that identifies hotspots based on the frequency of crashes, potential crash reduction, potential crash reduction, and other criteria without accounting for crash severities and number of crashes. Farnsworth (2013) describes two types of safety analysis methods that can be done in order to identify hotspot areas and evaluate improvements needed to improve hotspot locations: Traditional Descriptive safety analysis and Predictive safety analysis. The goal of traditional descriptive analysis is to summarize, analyze, and quantify crash data. This method is used to identify locations that are a priority and require attention as far as identifying top priorities and improvements. The second method is predictive analysis; it uses a statistical model in order to address the regression to mean. The method includes safety performance functions, ordinary least square regression, and crash reduction factors. According to Geedipally (2010) separating single vehicle crashes from multiple vehicle crashes helps reduce false positives and negatives and then modeling both types of crashes together to show inaccuracies. The distinctive model was used in Texas along with other methods to identify hotspots. There are six different performance measures used to measure the effectiveness of this model in relation to accuracy in identifying the true location of hotspots. The 12

25 first performance measure is the false discovery rate (FDR) which identifies the type one error within the hotspots identified. The smaller the FDR rate, the less error results. The second rate is the false negativity rate; it is a small number if the method performs well in identifying the hotspot. Sensitivity is the correct proportion segments that were correctly identified as hotspots. Specificity (SPEC) the correct or true number of hotspots that were classified as non-hotspots. Risk (RISK) is the number of errors and the number of sites under analysis. Therefore, when the value of risk is closer to 0, the model is performing very well. One way to assess the Hotspot identification method is to rank data in accordance to false positives indicating that the crash hotspot is present at a certain location when it is not and a false negative indicating that the hotspot location is not present when it is a hotspot (Cheng, 2005). They also state that using accident rates to identify the hotspots is one of the methods for identifying hotspots, but is the least effective of the top four tests in accurately identifying the locations of the hotspots. Therefore, it is used in combination with accident reduction potential, accident rate ranking, accident frequency ranking, and empirical bayes can best detect the hotspot location. The Empirical Bayes is the most effective of the four tests as it accounts for all the uncertainties of data. The Empirical Bayes method covers important problems when it comes to the safety estimation. It creates an increase in precision when it comes to estimates that are the best to use when the investigator is limited to 1-3 years data (Wu, 2014). Wu (2014) stated that the Interactive Highway Safety Design Model (IHSDM) use the Empirical Bayer method as part of the highway safety model. HSM (2010) provides the guidance, tools, and methodologies for evaluating safety and determining safety measures which indicate the severity level of injuries occurring on the roadway. HSM provides a methodology for forecasting the frequency 13

26 and severity level of crashes of a proposed roadway modification. This manual provides quantitative information for evaluation and crash analysis. HSM provides transportation professionals the ability to make decisions and solutions in order to improve safety and to evaluate and interpret related statistical data. The predictive analysis methodology forecasts the severity and number of crashes that are expected to occur in the future. HSM could be a good resource in a variety of applications. One benefit of HSM is it helps to rank projects by economic priorities. Also, it helps identify design model alternatives and its potential effect on the frequency and severity of crashes. The Crash Modification Factor (CMF) is a good indicator to quantify the safety benefit of reducing the crash frequency when doing mitigations and improvements to a roadway segment or a particular sight. For freeways and arterials, the predictive model is a method to identify average number of crashes, the frequency of crashes, and the crash types that are occurring at the segment knowing the geometrical condition of the segment. The predictive model combines three factors: safety performance factor, crash modification factor, and calibration factor. Safety Performance Factor (SPF) is the function which is used as an application to come up with a way of predicting frequencies of crashes. The Safety Performance Factor is a factor which is calibrated in order to meet local or state jurisdictional and geographic conditions. Harwood (2007) stated that HSM improves safety predictions. It provides procedural guidelines in order to interpret results such as rates of accidents. Also, it gives a prediction of the frequency of crashes when certain mitigations and improvements are made on the roadway. The predictive analysis will be applied to the study in order to compare to current crash rates and also determine the number of injury related and property damage only crashes. 14

27 This study analyzed the rate of crashes along the JFK corridor during The study area is broken into eight segments between major interchanges; see Appendix A. The crash rate ( ), crashes per 100 million vehicle miles traveled, for each segment,, can be determined by the formula below (1) where = the average daily trips along the corridor and = the length of segment. Table 1 shows the ranking of each segment by crash rate. 1 MD-155 to MD MD-222 to MD MD-43 to MD MD-152 to MD MD-272 to MD/DE 17 6 MD-24 to MD MD-22 to MD MD 543 to MD-22 3 The crash rate is particularly high from MD-155 to MD-272. Table 1 shows the crash rates for each segment by direction. The MD-155 to MD-222 and MD-222 to MD-272 experience a higher rate of crashes in the northbound direction. MD-43 to MD-152 has nearly double the crash rate in the southbound direction than northbound. Therefore, three segments warrant further investigation due their high crash rates, the segments between: MD-43 to MD-152, MD-155 to MD-222 and MD-222 to MD-272. The following sections will investigate the rate of crashes further. The subsequent section will: analyze crashes by time of day and weather condition. In addition, an analysis of vehicle speeds will also be presented. 15

28 MD-43 to 152 MD-152 to MD-24 to MD-543 to MD-22 to 155 MD-155 to 222 MD-222 to 272 MD-272 to DE Line Northbound Southbound The JFK corridor of I-95 is used for a variety of purposes: commuting, travel along the east coast, and freight Table 2 shows the corridor-wide crash rates by time of day. The rate of crashes is approximately three times higher during peak periods than the off-peak period. Many people use the corridor to commute to the DMV area. However, the numbers of DMV commuters diminish the further north. In addition, people that live in Maryland may commute to Wilmington, Delaware or even Philadelphia, Pennsylvania. The following sections explores crash rate for each segment by time of day. 7A-9A p-6p All other

29 As shown in Figure 2 southbound crash rates are more consistent across segments than in the northbound direction. During the AM peak period, the high crash rate of the first 7 miles of travel (MD-43 to MD-152) reflect the higher number of drivers commuting to work in the Baltimore-Washington Metropolitan area. The crash rate for the AM Peak southbound direction of MD-43 to MD-152 is nearly three times the northbound direction during the same period. At the other end of the corridor, along the Delaware state line, the rate of crashes were much higher in the northbound direction than in the southbound direction despite the majority more commuters traveling northbound. In the northbound direction, there is great variability in the rate of crashes between segments. The section from MD-155 to MD-222 has the highest AM peak crash rate followed by MD 222 to MD-272. These crash rates are alarmingly high when compared to an average corridor-wide crash rate of 20 crashes per 100 million miles traveled. The rate of crashes is over 10 times that of the average corridor-wide crash rate. Chapter 4 will explore the causes of crashes in these segments in detail. The variability between northbound and southbound crash rates was less during the PM Peak. Over the study period, little to no crashes was reported from MD-24 to MD-155. For all other segments with the exception of MD-152 to MD-24 the trends remained consistent. The southbound crash rate remained higher from MD-43 to MD As during the AM peak, there is still a spike in the northbound crash rates at MD- 17

30 Crash Rate (10^8 Crashes / VMT) Crash Rate (10^8 Crashes / VMT) MD-43 to MD-152 to MD-24 to MD-543 to MD-22 to MD-155 to MD-222 to MD-272 to DE Line Northbound Southbound MD-43 to 152 MD-152 to MD-24 to MD-543 to MD-22 to 155 MD-155 to 222 MD-222 to MD-272 to DE Line Northbound Southbound 18

31 Crash Rate (10^8 Crashes / VMT) MD-43 to 152 MD-152 to MD-24 to 543 MD-543 to 22 7 MD-22 to MD-155 to MD-222 to MD-272 to DE Line Northbound Southbound 155 interchange onward to MD-272. In addition, the northbound crash rate for these two segments was also high. The highest southbound PM peak crash rates occur between MD-222 to MD-272. The crash rate in the northbound and southbound directions was 186 and 128 crashes per 100 million VMT, respectively. Another segment which has experienced high crash rate and is the 2 nd highest crash rate in the northbound segment is MD-155 to MD-222 with a crash rate of 136 northbound and 82 and southbound. As shown in Table 2, the off-peak crash rate of 16 crashes per 100 million vehmiles of travel is much less than during the peak periods. The crash rates were relatively consistent by direction. Keeping with the trends described above, the segment from 19

32 MD-155 to MD-272 had the highest off-peak crash rates though they were significantly less than during peak hours. The whole segment from the beginning at the MD-155 interchanges going north past the JFK toll plaza towards the Delaware state line needs to be evaluated in order to identify the causes for these crashes and to determine possible solutions in reducing crashes occurring along these segments. As shown in the literature review, speed variation is a major cause of crash occurrence and factor in the severity of crashes Kockelman (2010). Due to differences in design speed and posted speed, it is a very important to look at the speeds in which motorist are driving. The design speed of the JFK Corridor is 70 mph; at most locations the posted speed limit is 65 mph. Northbound, the speed limit drops temporarily to 55 mph near the toll plaza approach between interchanges MD-155 and MD-222. This data was taken during March 2014 on weekdays free of non-reoccurring incidents. The speed runs were taken using a probe vehicle with 4 runs in each direction during AM peak (7AM-9AM), PM peak (4PM-6PM), and non-peak (9AM-11AM, 1PM-3PM, and 8PM-10PM). Average speeds were taken as the probe vehicle travel between interchanges. 20

33 Table 3 shows the average speeds traversed along each segment during the AM and PM peak periods. For much of the corridor during most time periods, the average speeds exceed the speed limit which implies a lack of congestion. In fact the segments with the lowest crash rate, MD 152 through MD-155, have the highest speeds. This supports the literature which states that high speeds alone do not increase the probability of an accident. 21

34 Location AM Peak (7a-9a) PM Peak (4p-6p) NB SB NB SB MD-43 to MD-152 to MD-24 to MD-543 to MD-22 to MD-155 to * * 68.1 MD-222 to MD-272 to DE Speed Limit 65 mph *Speed Limit 55 mph at MD-155 to 222, NB only Figure 5 shows the speed collected on each speed run in the AM Peak northbound direction. The lower limit of the gray band is the 65 mph posted speed limit and upper limit represents the 70 mph design speed. Average speed is highest at the between the MD-152 and MD-24 interchanges at 71.9 mph. The 55 mph speed limit along toll approach, is reflected in the lower average speed of approximately 62 mph along the MD-155 to MD-222 segment. As shown in Figure 6, the speed limit is 65 mph throughout the corridor in the southbound direction. The average run speed is at or above the design speed limit for most of the corridor. However, the freeway is severely congested near Baltimore at MD-43 and the average speed drops to 32 mph. Figure 7 shows the distribution in travel speeds for the probe vehicle during its runs northbound. The majority of the time, the vehicle was traveling between 65 and 75 mph (10 miles per hour the speed limit in most sections). When the speed limit drops at MD-155 to 222 to 55 mph, there is a wide distribution in speeds along the segment. Nearly 60% of the time the probe vehicle was traveling between 65 and 75 mph, well over the posted speed limit. At this segment 22% of the time vehicle was traveling between 55 and 65 mph and below 55 mph for roughly 17% of the time. It is 22

35 Speed (mph) Speed (mph) hypothesized wide variance in speeds after Tydings Bridge towards the toll plaza is a major cause for the extremely high crash rate in the AM Peak in the northbound direction. Figure 13 shows the map of the toll plaza; the 55 mph zone is marked in red MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to 222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to 222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed 23

36 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% > 75 mph mph mph 55 mph 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% > 75 mph mph mph 55 mph 24

37 As shown in Figure 8, the difference in the northbound and southbound crash rate at MD-272 to the DE State Line during the AM peak can be explained by looking at the distribution of speeds. There is a high level of congestion along this segment. For only a small portion of the segment, the probe vehicle was traveling at or above the speed limit but the majority of the time vehicle was traveling well below the speed limit. Also note that from MD-155 to MD-22, the speeds were always above the speed limit. However, this segment has a relatively low crash rate of 23 crashes per 100 million VMT. 25

38 Table 3 shows the average speeds of the probe vehicle during the northbound and southbound runs. The speeds in the northbound direction remained consistent with the AM Peak. The speeds were higher in the PM Peak at the toll plaza when compared to the AM Peak. Near the Delaware border, the average speed was about 5 mph under than of the AM peak in the same direction. In the southbound direction, with the exception of the MD-43 to MD-152 segment, speeds were generally lower in the PM peak. Southbound PM volumes are higher than the southbound AM volumes from MD/DE line up to the Chesapeake house and before the MD-222 interchange; see. The volumes From MD-222 to MD-43 interchanges are lower than the southbound AM peak period. Figure 9 shows the four northbound speed runs. The runs were similar to that of the AM peak; however, the toll plaza segment, MD-155 to 222, maintained higher speeds and average run speeds are still considerably 5 miles higher than the 65 mph speed limit. The southbound speed runs are shown in Figure 10. Figure 11 shows the distribution of the probe vehicle s speed in the northbound direction. The results were consistent with the AM Peak analysis. MD-155 to MD-222 still experiences a wide range of speeds due to the speed limit drop at the toll plaza. However, the crash rate was nearly half of that of the AM Peak so further investigation is required. Moreover a similar crash rate was seen on the MD-222 to MD-272 segment despite maintaining relatively constant speed. In the southbound direction, the probe vehicle traveled below 55 mph at the northern end of the segment more often than in 26

39 Speed (mph) Speed (mph) the AM Peak; see Figure 12. Conversely at the southernmost end near MD-43, vehicle speeds were generally maintained at mph MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to 222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to 222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed 27

40 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% > 75 mph mph mph 55 mph 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% > 75 mph mph mph 55 mph 28

41 As the freeway is predominantly in free flow during the off-peak hours, analyzing the speed profiles during these hours will provide insight to speed-related causes of incidents in the absence of congestion. Three off-peak periods 9 am to 11 am, 1 pm to 3 pm, and 8 pm to 10 pm, each with four runs, are analyzed in the section below. For this analysis, the speeds for the MD-155 to MD-222 segment are separated for the segment prior to the toll plaza and the segment after the toll plaza. Figure 13 shows the portion of the corridor in detail. The 55 mph segment is in red. Table 4 shows the average speeds traversed along each segment. Speeds were higher in the off peak than during the AM and PM peak periods. Generally the highest speeds were observed between MD 152 and MD 155 the segments with the lower crash rates. 29

42 Location Off Peak (9a-11a) Off Peak (1p-3p) Off Peak (8p-10p) NB SB NB SB NB SB MD-43 to MD-152 to MD-24 to MD-543 to MD-22 to MD-155 to Toll Toll to MD MD-222 to MD-272 to DE Speed Limit 65 mph *Speed Limit 55 mph at MD-155 to 222, NB only 9AM 11AM Figure 14 shows the four speed runs that were conducted northbound during between 9 am and 11 am and the dip in the average speed between the toll plaza and MD-222 where the speed limit changes from 65mph to 55mph. It appears that during the first run, there was still some morning rush hour congestion at the plaza. At all other locations, the average run speed exceeds the 70 mph design speed. For southbound direction the speed limit is set at 65mph throughout. The speeds were above 70 mph along the entire corridor with the highest speeds along MD-22 to MD

43 Speed (mph) Speed (mph) MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed 31

44 1PM 3 PM In the northbound direction speeds were a little lower from 1 to 3 PM than there were from 9 AM to 1 PM along the southern end of the corridor. Southbound, there is a drop in speed at the northern (Delaware) end of the corridor. Speeds increase through to MD-24 where there is a drop in speed as the Baltimore suburbs are approached. Irrespective, the average speeds tend to be high during the time from 1pm to 3pm when compared to the peak period. Since there are fewer vehicles on the road, drivers may feel safe to drive at a higher speed. 8PM 10PM During the nighttime, from 8-10 pm, high speeds were maintained along the corridor. In the northbound direction, the speeds were comparable to those from 1 to 3 and slightly lower than the 9-11 runs. Perhaps, drivers reduced their speed some while driving at night. In the southbound direction, the average run speed remained high with all speeds greater than the 70 mph design speed. Table 5) and followed by a T-test assuming equal means. The T-test shows that at a 95% confidence level, there was a significant reduction in speeds at the toll plaza. However it is not clear if that reduction was due to the drop speed limit or the presence of the toll plaza it is hypothesized the later. 32

45 Speed (mph) Speed (mph) MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed 33

46 Speed (mph) Speed (mph) MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed MD-43 to 152 MD-152 to 24 MD-24 to 543 MD-543 to 22 MD-22 to 155 MD-155 to Toll Toll to MD-222 MD-222 to 272 MD-272 to DE Run #1 Run #2 Run #3 Run #4 Avg speed 34

47 Mean Variance Observations P(F<=f) one-tail 0.20 Pooled Variance 20.3 df 22 t Stat P(T<=t) one-tail E-06 The corridor analysis has shown that speed is not the primary predictor of accidents along the corridor. Chapter 4 will explore other causes of accidents along the following segments: MD-155 to MD-222 and MD-222 to MD-272 as those segments have an extremely high crash rate. 35

48 Based on the corridor-level analysis the 8.5 mile stretch from MD-155 to MD-272 requires further analysis due to their high crash rates. In Chapter 3, it was determined that high speeds are not a predictor of high crash rates along this corridor. Other causes will be explored through a site analysis. Figure 23 shows the location of crashes along MD-155 to MD-222. There are two segments with high cluster of crashes in a half-mile segment to account for changes in geometry. The first segment is between MM and the second cluster is between MM and the second cluster is from MM Both segments incurred high crashes at a half-mile segment. As shown in Figure 21 and Figure 22, both of these locations occur along segments with horizontal curves. Further investigation of the cause of crashes at these locations is provided in the subsequent sections. Frequency

49 37

50 The most predominant crashes occurring in this segment are rear-end crash which accounted for over 30 percent (37 out of 115) of the crashes occurring at the segment and fixed object crashes which accounted for 36 crashes both making up 63 percent of total crashes occurring at the segment; see Figure 23 Fixed object crashes are crashes in which a vehicle runs into an inanimate roadway object FXOBJ OTHER PARKD PED RREND SDSWP Rear-End Crashes The highest probable cause of rear-end crashes occurring at this segment are following too slowly crashes which accounted for 35 percent (13 out of 37) of crashes occurring at the segment. This may be due to drivers following too closely for the speed in which they are traveling thus not having inadequate available stopping distance. The rest of the probable causes occurred sporadically. Also, wet condition crashes accounted for about 30 percent (11 out of 37) crashes occurring at the segment and 33 percent (11 out of 33) of total wet crashes were rear-end wet condition crashes; see Table 7. 38

51 35% 13 11% 4 8% 3 8% 3 5% 2 5% 2 27% Figure 24 shows the number of crashes by half-mile segment. There are two-half mile segments with a cluster of crashes. The first segment is MM and the second segment is MM , the latter which occurs along the toll plaza approach. Between MM rear-ends crashes occur along the toll plaza approach. There are 8 northbound crashes and 4 southbound crashes at this segment. The toll is collected northbound and the speed limit drops from 65mph to 55 mph. It is hypothesize that there are two predominant causes in accidents at this location: 1) not all people are reducing speeds which may result in a larger variance in speeds across vehicles and 2) the segment lies along a horizontal curve. Crashes are relatively spread out throughout the half-mile segment. The second cluster of crashes occur between MM with 9 crashes occurring at the segment, 7 occurring in the northbound direction and 2 occurring at the southbound direction. There is a hotspot location at MM with 5 crashes occurring, 4 of which are northbound crashes and 1 southbound crash. There is a horizontal curve at this location which may be causing crashes because speed limit is 65 mph at this location; see Figure 25. There is an advisory speed which is advising drivers of 55 MPH speed limit ahead. Drivers are not reducing their speed at the curve. Therefore, there were a cluster of rear-end crashes occurring at that segment. 39

52

53 Fixed Object Crashes Fixed object crashes are also a prominent crash type occurring at the segment. The most prominent probable cause of crashes occurring at the segment is Fail to drive in single lane which accounts for 39 percent of crashes (14 out of 36). The second most predominant probable cause is Too fast for conditions which accounts for 19 percent of crashes (7 out of 36) occurring at the segment. Lastly, Failed to give full attention accounted for 17 percent of crashes (6 out of 17) occurring along the segment. Fail to drive in single lane Too fast for conditions 19 7 Fell asleep, fainted, etc. 8 3 Fail to obey other control 6 2 Vehicle defect 6 2 Improper lane change 3 1 Under influence of drugs 3 1 Grand Total A cluster of fixed object crashes occurred at Mile Marker with 4 crashes, 2 occurring northbound and 2 occurring southbound at the segment. Three of the crashes that occurred at the cluster were Failed to drive in a single and one Too fast for conditions. The crashes are occurring right before the bridge at Frenchtown road approaching the toll plaza where there is a slight horizontal curve. Fail to drive in single lane and hitting a fixed object means that the vehicle is hitting the guardrails. This is probably due to high speeds at the segment and vehicle front driving slowly and suddenly hitting brakes causing the other vehicle to swerve off lane or skidding causing a fixed object crash. In the southbound direction at MM there is a merge during the horizontal curve which may also be cause of crashes. See Figure 26 and Figure 27 for 41

54 the existing condition on the northbound and southbound side of MM 92.19, respectively. The majority of crashes (48%) along this segment occurred in the summer when traffic volumes are at a high. The remaining seasons, Fall, Winter and Spring nearly split the remaining crashes at 17.4%, 16.5%, and 18.3%, respectively. Thirty-two percent of crashes along the MD-155 to MD-222 corridor occur during inclement weather. 42

55 The most predominant crash type occurring during wet condition crashes is fixed object crashes accounting for 48 percent of crashes occurring during wet conditions. Also rear-end crashes are pretty high during wet condition crashes accounting for 33 percent of crashes occurring at the segment; see Figure 28. The highest probable causes for crashes occurring at the segment are Too fast for conditions accounting for 30 percent of crashes occurring at the segment. Failed to drive in single lane accounted for 15 percent of crashes occurring at the segment. Other probable causes of crashes are occurring sporadically during wet condition crashes. See Figure 29 for a complete list of probable causes FXOBJ OTHER PED RREND SDSWP 43

56 Exceeded speed limit Fail to drive in single lane Fail to give full attention Fail to obey other control 10 3 Fail to obey traffic signal Fell asleep, fainted, etc Followed too closely Improper lane change Other or Unknown Stopping in lane roadway Too fast for conditions Probable Cause of Crashes during Wet Conditions Figure 30 shows the distribution of wet crashes along MD-155 to MD-222. Segment with 11 crashes has the highest number of crashes within a halfmile segment..there are two hotspots within this segment. The first hotspot is at MM where four crashes occurred, 3 northbound and 1 southbound. The location is at the toll plaza approach going northbound at Frenchtown overpass and is shown below in Figure 26 and Figure 27. The crashes may have occurred due to horizontal curvature and vehicles skidding of roadway. The second hotspot is at MM 92.4 where 6 crashes occurred, 3 of which are northbound crashes and 3 southbound crashes. This location in on a horizontal and vertical curvature and is just before the crest curve. Crashes at this location may be occurring due to limited sight distance and drivers traveling too fast and skidding off the roadway. The second highest cluster of crashes by half-mile segment occurred at MM where 6 crashes occurred. There was only 1 hotspot at that segment. The hotspot is at MM 92.6 with 4 crashes occurring at that point all occurring in the 44

57 northbound direction. The location is at the toll plaza approach. The probable cause of the crash varied from one crash to another. Crashes at that area may be occurring at that location due to pavement markings and vertical joints that are not consistent with one another. At that location we go from 3 lanes to 12 lanes. Also during wet conditions drivers might be confused on which lane to follow during the wet conditions and that may cause a crash. The third cluster of crashes occurred at segment MM to with a total of 5 crashes and 1 hotspot accounting for 4 out of the five crashes. The hotspot is at MM 92.0 which is.7 miles south of the Tydings Bridge. There are 2 northbound crashes and 2 southbound crashes at that segment. The location is right after speeds are reduced from 65mph to 55mph. The location is at a horizontal curvature and there are two crashes with the probable cause of Too fast for conditions, 1 Fail to drive in single lane and 1 fail to obey traffic signal. Drivers need to be alert to the change in roadway conditions that may cause them to crash. Figure 26 shows MM92.00 and location of crashes occurring at the segment. Lastly there is a cluster occurring at segment with all 3 crashes occurring at one hotspot on MM northbound crashes and 1 southbound crashes are occurring at that point. The probable cause for this crash varies from one crash to the other. One probable cause for the crash was due to wet conditions. These crashes are occurring on the Tydings Bridge. 45

58

59 Figure 33 shows the number of crashes along this segment by time of day. A high number of crashes occur in the morning between 6AM and 10 AM and again the early afternoon. Another period of high crashes appeared around midnight. Further investigation of nighttime crashes is provided below

60 FXOBJ OTHER RREND SDSWP Nighttime Crashes There are 39 night crashes occurring along the segment. Of the 39 crashes, 26 crashes occur in the northbound direction and 13 crashes occur in the southbound direction. The majority of crashes along the segment, 31% or 12 of 39, are with fixed object crashes; see Figure 34. The probable causes for the fixed object crashes at this segment are due to drivers failing to drive in single lanes, fail to give full attention, and too fast for condition crashes. This may be due to driver s reduced vision during night time. Other types of crashes occurring at the segment are rear-end crashes which accounts for 11 crashes occurring at the segment. For rear-end crashes the highest probable cause is following too closely crashes which accounted for 4 crashes occurring at the segment. The highest probable cause of night crashes occurring at the segment is Fail to drive in single lane which accounts for 18 percent (7 out of 39) crashes occurring at the segment. The second highest probable cause of crashes along the segment are Fail to give full attention and Too fast for condition crashes with 5 crashes occurring at each one and both accounting for about 26 percent (10 out of 39) crashes occuring at the segment; see Figure

61 There are two locations where there are a cluster of crashes occurring at the segment; see Figure 36. The first area is between MM 92.4 to MM 92.9 with 11 crashes occurring at the half-mile segment. At that segment there is a hotspot area with 5 crashes occurring at MM 92.6, 3 of which occur in the southbound direction and 2 occur in the northbound direction. In the southbound section MM 92.6, the crashes may be occurring due to limited view. First, there are vehicles entering traffic from weigh station and administration building and along that there is a horizontal curve occurring at the segment. (Note: This is the same area that is a hotspot for wet condition crashes and it appears to be a sight distance issue). For, the northbound section at MM 92.6 the JFK toll plaza approach at the point where we pick up more toll lanes leading to the toll booth. The joints and markings are different which may cause drivers to be confused. 49

62 Predictive analysis has been conducted using the Interactive Highway Safety Design Model (IHSDN) software to predict the expected number of crashes for the evaluation period and the crash rate in one year. The total predicted crash rate results from IHSDN is 67.0 per 100 million VMT while calculated crash rate calculations are 66.5 per 100 million VMT. Therefore, the predictive analysis results are very close to the actual crashes rates calculated in The crash rates including the Tydings Bridge calculations are higher than calculated crash rate accounting for 90.6 crashes per 100 million VMT while the calculated crash rates are 62.8 crashes per 100 million VMT. Therefore, HSM is predicting a higher crash rate than what is calculated for 2013; see Table 9. For a map of the study area see Figure 37. As shown in Table 10, Site No. 5 experiences a high number of crashes along Tydings Bridge. Due to the limited shoulder clearance and presence of a concrete median several sideswipe crashes were noted. Site No. 2 is the location with the horizontal curve. Several rear-ended crashes were noted at this location. 50

63 - Not Including Tydings Bridge (MM ) - Including Tydings Bridge (MM ) Site No. Effective Length (mi) Expected No. Crashes for Evaluation Period Crash Rate (crashes/mi/yr) Travel Crash Rate (crashes/100 million veh-mi) * *Site 5 contains Tydings Bridge 51

64 Severity of Crashes Another finding which is of significance is the predicted number of injury related and property damage crashes. The predicted property damage only crashes are higher than the number of property damage crashes occurring on the roadway. On the other hand, the number of injury related crashes were founded to be a little higher than what was predicted using IHSDN software. PDO Injury Limitations of Analysis Though the study presented above provides insight, the study was done with uncalibrated predictions which may slightly affect output results. In addition, there is an entrance ramp which at the weigh station on the southbound direction which has not been counted for which caused an assumption to distance end to entry decreasing to be measured from the segment to the MD-222 interchange. This portion of the JFK corridor has several horizontal alignment changes; see Figure 38. There are 4 cluster of crashes occurring at this segment that require detailed analyses to predict the types of crashes occurring at the segment, probable causes and different countermeasures that need to be taken in order to reduce crashes occurring at the segment. Figure 39 shows the locations of these crashes. 52

65 The most prominent crash type occurring at this segment is rear-end crashes which accounts of 38 percent (72 out of 188) of crashes and fixed object crashes which account for 24 percent (46 out of 188) of crashes occurring at the segment. Also, sideswipes were a little high at the segment making 19 percent (36 out of 188) of 53

66 crashes occurring at the segment. The rest of the crashes occurring were other/unknown crashes. Types of Crashes Percentage # of crashes RREND FXOBJ SDSWP Other PARKD 2 3 Total Rear-end Crashes Of the total rear-end crashes that have occurred 46 percent (33 out of 72) crashes occurred in the northbound direction and 54 percent have occurred in the southbound direction. Also, only 25 percent of rear-end crashes occurred at night. The two main probable causes for rear-end crashes are followed too closely which accounts for 54 percent (39 out of 72) of crashes and too fast for conditions accounting for 27 percent( 20 out of 72) crashes. A cluster of crashes occurred at Mile Marker 98.1 (Segment 4.9) with 6 crashes occurring at that area; see Figure 40. The cluster of rearend crashes occur along a horizontal curve. The top two probable causes for this crash type are fail to drive in single and following too closely. Fail to drive in a single lane accounted for over 25 percent (48 out of 188) of crashes. Following too closely accounted for 22 percent (42 out of 188) of crashes occurring at the segment. 54

67 Fail to drive in single lane Followed too closely Too fast for conditions Fail to give full attention 9 5 Fail to obey other control 1 1 Fail to yield right-of-way 3 2 Fell asleep, fainted, etc. 5 3 Illegally in roadway 1 1 Improper backing 1 1 Improper lane change 13 7 Improper passing 1 1 Improper turn 1 1 Other or Unknown 7 4 Passenger interference/obstruction 1 1 Rain,snow 1 1 Stopping in lane roadway 1 1 Under influence of alcohol 5 3 Under influence of drugs 2 1 Vehicle defect 7 4 Wet 2 1 Animal 1 1 Debris or Obstruction 2 1 Exceeded speed limit 1 1 Total

68 Followed Too Closely As expected when one follows too closely the primary collision type will be a rear end collision here at 93% (39 out of 42) of all collision types. There is a cluster of crashes occurring at the half-mile segment between MM 97.7-MM 98.2 with 10 crashes occurring at the segment. At that segment there is only one hotspot at MM 98.1 with 5 crashes occurring at that point, 4 occurring southbound and 1 occurring northbound. All the crashes occurring at that hotspot were rear-end crashes. The location has a horizontal curve along with a steep median which may be causing limited sight distance. See Figure Too Fast for Conditions Crashes This is the third highest probable cause occurring at the segment. The most predominant crash type occurring is rear-end crashes which is accounting for 61 percent (20 out of 33) crashes occurring at the segment. This may be due to the vehicle not being able to brake on time causing it to rear-end the vehicle ahead of it. Vehicles are driving at a high speed and losing control. The second most predominant crash type 56

69 is fixed object crashes which accounts for 27 percent (9 out of 33) crashes occurring at the segment; see Figure 42. Crashes Inclement Weather Wet condition crashes make up 20 percent (39 out of 188) of crashes occurring at the segment. The most predominant crash type occurring at the segment rear-end, fixed object, and sideswipe crashes making up 33 percent (13 out of 39) crashes, 31 percent (12 out of 39) crashes, and 23 percent (9 out of 39) crashes occurring on the segment respectively. The other 13 percent (5 out of 39) crashes are other/unknown crashes. The highest probable cause for crashes occurring at this segment is Followed too closely which accounts for 31 percent (12 out of 39) crashes occurring at the segment. The second highest probable cause of crashes is fail to drive in single lane which accounts for 23 percent (9 out of 39) crashes occurring at the segment and too fast for condition crashes which also accounted for 23 percent (9 out of 39) crashes occurring at the segment; see Table

70 Followed too closely Fail to drive in single lane 9 23 Too fast for conditions 9 23 Wet 2 5 Improper lane change 2 5 Other or Unknown 1 3 Rain, snow 1 3 Fail to yield right-of-way 1 3 Under influence of alcohol 1 3 Fail to give full attention As shown in Figure 43: Crashes by Time of Day, the highest number of crashes occurred in the early afternoon and evening, with the highest number of crashes occurring between 4 pm and 5pm during the evening peak. That means that most 58

71 crashes occur during the PM peak when people are going back home from work. Further investigation would be necessary to determine the cause of these crashes. One reason could include the high traffic volume during the PM peak. However, since the number of crashes was low during the AM peak there may be other influences. Nighttime Crashes Thirty-two percent (61 out of 188) crashes occurred during night time. Crashes were not very severe during the nighttime. Only 26 percent (16 out of 62) of crashes were injury related. There are a cluster of night crashes occurring at MM 94.7 The most predominant crash type is rear-end crashes which accounts for about 30 percent (18 out of 61) crashes occurring at the segment. Drivers may be driving at an excessive speed which is causing the rear-end crashes. Also, vehicles may be tailgating which is causing them not to have time to stop and causing them to rear-end the vehicle ahead. There are about the same amount of fixed object and sideswipe crashes with 13 fixed object crashes and 14 sideswipe crashes both accounting for 44 percent of night crashes. The other 26 percent (16 out of 61) crashes are unknown crashes FXOBJ OTHER RREND SDSWP 59

72 Crash As shown in Figure 44, there are two clusters of crashes by half mile segment. The first cluster is occurring between MM 98.7 and MM9 9.2 with 11 crashes, 5 northbound crashes and 6 southbound crashes. There are two hotspots along this half time segment. The first hotspot is located on MM with three crashes occurring on the segment, 1 northbound and 2 southbound crashes. Two of the crashes were rear-end crashes. The location is located right at a crossover and vehicles may tend to slow down quickly and check is there is a police vehicle at night failing to get caught driving at a high speeds; see Figure

73 The second highest half-mile segment of crashes is located between MM 94.7-MM 95.2 with 10 crashes occurring at the segment. There is one hotspot with 8 out of 10 crashes occurring at MM Of the 8 crashes, 6 occurred northbound and 2 occurred southbound. There is no lighting in that location and there are no reflective markers on 61

74 the guardrails. During night time it may be difficult to see the roadways clearly. Please see Figure 47 for more details. Probable Causes of Nighttime Crashes The highest probable cause of night crashes occurring at the segment is Fail to drive in single lane crashes which accounts for 25 percent (16 out of 61) crashes occurring at the segment. This may be due to drivers reduced vision during the night time which may cause him/her not to distinct between different lanes on the roadway. The second highest probable cause of crash is too fast for condition crashes which accounts for 18 percent (11 out of 61) crashes occurring on the segment. Driver may not view the roadway as clearly at night and driving faster does not help him/her Perceive vehicles ahead and react in a timely manner. Also, the probable cause of following too closely made up 15 percent (9 out of 61) crashes occurring at the segment. This is also due to drivers driving at a high speed and leaving adequate stopping sight distance. 62

75 The overall corridor crash rates are below statewide average when it comes to severity of crashes occurring throughout the whole corridor. The rate of injury crashes are fairly below the statewide average. All crash types are fairly below statewide averages. Crashes occurring during night time are slightly higher than the statewide average running at 33 percent night crashes while state average runs at 31 percent. Wet surface crashes are running at the same percentage as statewide with 21 percent of crashes occurring during wet condition. As supported by the literature, high speeds are not a predictor in crash rates. Along this corridor other factors such as geometry, season, and time of the day are proved more important. The 8.5 mile segment from MD-155 to MD-272 was the most problematic. MD- 155 to MD-222 was the segment with the highest crash rate. In particular, the northbound direction along the toll plaza approach identified as a hotspot. The most predominant crash type occurring at this segment is rear-end crash and fixed object crashes with about equal number of crashes occurring for both. Though frequent, the crashes occurring at the segment are not severe. In fact 70 percent of crashes occurring at the segment are property damage only and the remaining are injury related. There were no fatal crashes that occurred on the segment. The two primary contributing probable causes are failure to drive in a single lane and driving too fast for roadway condition crashes. The drop in speed along the toll plaza approach coupled with a roadway profile that incurred significant geometrical changes with a horizontal curving tying in with a vertical curve along with a crossover at the beginning of the vertical curve at MM 91.6 led to frequent crashes. Two hot spot locations where identified along this segment. 63

76 The crash rates found along the segment were compared to the rates predicted through the Highway Safety Manual predictive analysis methodology. It was found that the predicted crash rates were higher than actual crash rates for A high number of crashes occurred along the eastern part of the bridge approaching a horizontal curve. The major cause of crashes along this segment is the steep horizontal curve just past the bridge. Vehicles approach at a high speed, lose control, and swerve off the road and crash. Over 30% of the crashes occurring along this segment are at night when visibility is limited. This half mile segment has the highest cluster of crashes occurring along the segment. The most predominant crash types are rear-end and single vehicle crashes. Due to the toll plaza approach, the speed limit drops from 65 mph to 55 mph. Over 60 percent of crashes occurred in the northbound direction the approach in which the toll is levied. A high number of crashes occurred northbound during the AM peak. These crashes were due to congestion and resulted in vehicles rear-ending one another or swerving off road to avoid hitting other vehicles. The subsequent segment along the JFK Corridor, MD-222 to MD-272 experiences the second highest crash rate. This segment is just past the toll plaza where the speed limit returns to 65mph. Roadway geometry changes throughout the segment. The most prominent crash type occurring at this segment is rear-end crashes which accounts of 38 percent of crashes and fixed object crashes which account for 24 percent of crashes occurring at the segment. Also, sideswipes were a little high at the segment accounting for 19 percent of crashes occurring at the segment. The top two probable causes for these crashes are fail to drive in single lane which accounts for over 25 percent of 64

77 crashes and following too closely which accounts for 22 percent of crashes occurring at the segment. Thirty-two percent of crashes occurring at the segment are nighttime crashes. There were no fatal crashes occurring at this segment. The crashes were not severe crashes; in fact, 69 percent of crashes were property damage only crashes and 31 percent crashes were injury related crashes. MM One location in particular was problematic in this segment the Winch Road overpass. Here there were 24 total crashes equally disbursed in both directions. There is a vertical curve prior to Winch Avenue overpass along with a horizontal curve tying in with a vertical curve at Winch Avenue overpass. Fifty-eight percent of crashes occurring at this segment were night time crashes. In addition, there was a concentration of crashes which occurred at MM 94.7 at the crest curve in the northbound direction. Throughout this analysis some systematic issues were identified. Vehicles exceeded speed limits throughout. This resulted in rear-end and fixed object crashes at horizontal and vertical curves. Other issues that were noted were the failure to drive in single lane which could be a function of operator error as well as roadway geometry and too fast for conditions. In order to increase awareness of the horizontal curves and to reduce speeds at these approaches, chevron alignment signs as well as advisory speed signs should be installed to notify drivers of changes in roadway alignment. In addition, flashing signal warning indicators may be used to warn drivers of changes in traffic conditions ahead. In order to reduce the occurrence of driving into other lanes, the lane widths may be increased along curves where able and raised pavement markers may be used to increase visibility. It is suggested that marking which can be plowed over such as recessed pavement markers be used. Dynamic messaging signs may be used to notify 65

78 drivers of non-reoccurring incidents and inclement weather conditions. Lastly, increased speed enforcement either through the presence of cops or speed camera may be beneficial along the straight approaches leading to sudden changes in alignment. This study is useful to the state DOT decisions makers. This study can be used for the implementation of safety measures along the corridor and to prioritize investment. In addition, as the state decides to raise or reduce speed limits on portions of the corridor in the future, this study would be helpful in providing performance information under current speed limits and conditions. The speed study was limited to the speed profile of a single probe vehicle with four runs per period. Future work would benefit from a spot speed study at the hotspot locations. In addition, more discrete volume information could provide more insights to the operation of the corridors. Other next steps would include an analysis of locations in need advisory speed limits due to the roadway geometry. Future work may also include looking at the relationship between design speed, posted speed and actual travel speeds in order to provide recommendations for the design of new facility as well as the proper setting of speed limits. Also, applying HSM using historical crash data will be done in future study. 66

79 Abdel-Aty, M. (2005). Bonneson, J. A. (2009)... Cheng, W. a. (2005). 37,. Corben, B. (2001). Technology to enhance speed limit compliance" Proceeding of the 2001., 4.) Dutta. (2005)... Farnsworth, J. S. (2013). Hot Spot Identification and Analysis Methodology. Fitzpatrick. (2003). Design Speed, Operating Speed, and Posted Speed practices.. Fitzpatrick, K. (2003). Design speed, operating speed, and posted speed practices.. Garber, N. J. (1988). "Speed Variance and Its Influence on Accidents. Geedipally, S. R. (2010). Investigating the effect of modeling single-vehicle and multivehicle crashes seperately on confidence intervals of Poisson gamma models. 42. Gerald. (2012). Methods and Practices for Setting Speed Limits.. Graettinger, A. (2005)... Hadi, M. A.-F. (1995). 67

80 Haghani, A., Hamedi, M., Fish, R., & Nouruzi, A. (2013). Maryland State Highway Administration. Haque. (2012). Harwood, D. W. (2000)... Harwood, D. W. (2007)... Hegyi. (2013). Hummer, J. E. (2010)... Jeihani, M., & Ardeshiri, A. (2013). Maryland State Highway Administration. Jiang, X. (2012). Kamruzzaman, M. M. (2013).. Kanellaidis, G. (1996). 122 Khattak, A. J. (1997). Kockelman, K., & Jianming, M. (2010). Freeway speeds and Speed variation... 68

81 Koutsopoulos, H., Lotan, T., & Yang, O. (1994). Krammes, R. A. (1995).. Lemp, J. D. (2011). Analysis of large truck crash severity using heteroskedastic ordered probit models.. Lu, & Jian. (2003). Criteria for setting speed limits in urban and suburban areas in Florida.. Mannering L. Fred, M. J. (1996). Manual, H. C. (2000). Washington DC. MNDOT. (2012). Ottesen, J. (2000).. Papageorgiou, M. E. (2008).. Parker. (1992).. Pisano, P. A. ( 2008). Smile, K. L.-G. (2009).. Srinivasan. (2007). The Expert System for Setting Speed Limit.. 69

82 Transafety. (1997). Urban. (2003)... USDOT. (2009).. Wilmot, C. a. (1999)... Wu, L. Y. (2014).. Xiaobo, A. a. (2014).,.. 70

83 71

84 72