The focus of this dissertation was to examine wrong-way driving (WWD) events on Florida toll roads and Interstates. The universe of WWD data contains many sources of WWD events or incidents. Most of the previous research focused only on WWD crashes without considering other data such as WWD citations and 911 calls related to WWD incidents. While WWD citations and 911 calls data is abundant, this data has been largely overlooked in other studies. This dissertation provides a novel and holistic approach for evaluating WWD risk, which considers other risk factors such as WWD citations and 911 calls in addition to WWD crashes. WWD crashes are rare because they are less than 3% of all crashes, which makes them difficult to predict and analyze. WWD is very dangerous especially on high-speed limited access facilities. A right way driver on the mainline has very little time to take an action and avoid a wrong-way vehicle since the combined approach speed rates of both vehicles is very high. There is an average of 300 to 400 fatalities every year in the United States due to WWD crashes. There were 386 fatalities in Florida due to WWD crashes from 2007–2011; this ranked Florida third in terms of total WWD fatalities. There are many causes for WWD. The majority of WWD crashes occur during late night hours, and these crashes can be attributed to intoxicated drivers, confused/elderly drivers, and suicidal drivers. However, these are not all of the causes of WWD. In order to understand WWD, it is important to look beyond crash events. This research focused on two major toll road networks in Florida, which were the Central Florida Expressway (CFX) and the Florida Turnpike Enterprise (FTE). Overall, WWD crashes on the FTE system accounted for around 0.45% of all crashes, but accounted for 1.5% of fatal crashes. WWD on FTE shows that 15.2% of these crashes are usually fatal compared to only 2% of all WWD rural freeway crashes are fatal and only 0.7% of urban freeway crashes are fatal. In the citation data, not all wrong way drivers were issued citations. 15% of the WWD citations in the FTE dataset resulted in a crash. While analyzing the citation events, it has been found that they commonly do not result in crashes. However, the mere fact that a driver gets a wrong way driving citation, because he or she failed to correct his driving action before a police officer arrives at the scene, is by itself a risky behavior. The WWD Traffic Management Center (TMC) SunGuide data was explored in depth for the FTE system. 55% of the SunGuide events were never found, 11% were pulled over by Law Enforcement Officers (LEO), and 8% of the events resulted in crashes. 19% of the events were false calls. In 3% of the events, drivers corrected their WW action without an incident or crash. Understanding the relationships between non-crash WWD events (WWD citations and 911 calls) and WWD crash events is essential. The interaction between crash events and non-crash events was explored using six different models developed in this dissertation. Weighted crash risk values, which use all three types of WWD events (crashes, citations, and 911 calls), were created using the developed models from this research and were applied to rank locations in priority for enhanced WWD countermeasures. Model 1, a generalized linear model referred to as GLM 1, was developed from Florida statewide WWD data on limited access routes. GLM 1 was built using a Poisson's function. Non-crash events (citations and 911 events) were modeled to predict WWD crash events while leveraging the statewide count data that was broken down by hour of the day. The results of GLM 1 showed that Broward and Miami-Dade Counties are some of the hottest counties in Florida for WWD, and SR 821 located in these two counties is one of the hottest routes for WWD in Florida. SR 821 ranked highest in terms of WWD crash risk using a statewide developed model in this dissertation. Model 2, which was another generalized linear model (referred to as GLM 2), used an additional time variable to square the hour difference from noon. The form of GLM 2 was similar to GLM 1, but the results of GLM 1 were a little stronger than GLM 2. Another model using Artificial Neural Network (ANN) was developed and compared to GLM 2. It was found that ANN provided a stronger fit of WWD crash predictions compared to GLM 1 and GLM 2. However, when the ANN was used with other non-crash events to produce a crash prediction values outside of its original data set, the ANN model was not very useful for this application because of ANN's nature to overfit its original data set. Model 3, noted as GLM 3, used yearly non-crash data in South Florida to predict an entire route WWD crashes annually. Model 4, also noted as GLM 4, was one of the most useful models created from this body of work and used the same South Florida network as GLM 3. Using non-crash events and route characteristics such as geometric design configurations and traffic volumes at interchanges within the segment, GLM 4 predicts WWD crashes within 7- interchange route segments over a 4 year time period. GLM 4 used a method to aggregate the 7-interchange route segments, which leveraged more data points by overlapping segments to provide a larger data set of WWD crashes. The predicted WWD crashes from GLM 4 were added to the actual WWD crashes to produce a 7-interchange crash risk value. Using this WWD risk assessment method allows for the inclusion of more than just WWD crashes when evaluating and prioritizing sites for implementation of WWD countermeasures. In addition, using segments/corridors to target countermeasures is a smart approach for combating the WWD problem because in many instances, it is difficult to know where the WWD event first started or got initiated, and some of the WW drivers can travel considerable distances before they are either apprehended by law enforcement or end up crashing with the oncoming traffic. Similar to GLM 4, GLM 5 was another route segment model developed using WWD data collected for the Central Florida region's limited access network. The developed GLM 5 used 5-interchange segments to predict crash risk. Both GLM 4 and GLM 5 models were microscopic in the sense that they prioritize candidate interchanges for implementation of WWD countermeasures. In order to go beyond the minimal standards for combating WW, Florida toll road agencies are testing enhanced/flashing "Wrong Way" signs at exit ramps. These flashing devices add more emphasis to the existing "Wrong Way" signs (and or other traffic control devices) at the exit ramps. The CFX's application of the Rectangular Rapid Flashing Beacon (RRFB) for "Wrong Way" signs is an entirely new concept that was applied in Central Florida for the first time. The FTE's application of the MUTCD approved Blinker Sign for "Wrong Way" has been used in other states such as Texas. These countermeasures were examined and briefly studied during their test pilot phases. Partial results are documented in this dissertation but continuous observations and data collection at the pilot test sites and potential expansions of these sites in South and Central Florida (and other parts of the state) are needed for complete and comprehensive evaluation of the effectiveness of these new technologies. The FTE SunGuide TMC WWD event durations were collected for the nearest known interchange from the SunGuide reports. This information was compiled for the entire FTE system of interchanges. These SunGuide WWD event durations show the time spent by the FTE operators while actively combating and responding to various WWD events (never found events, pulled over events, and crashes). A method using the actual time spent responding to WWD, and the estimated duration of response (prior to the introduction of SunGuide) to crashes, citations, and 911 calls was developed to rank the interchanges in order of highest durations to lowest. The method developed in this dissertation showed the top percentiles in terms of durations (in minutes), and was used to cross check with the risk ranking of the WWD risk segment models GLM 4 and GLM 5. However, the SunGuide durations method is unique and robust because it weighs in individual interchanges using one common metric of WWD; i.e., total durations of response to the event at each interchange in the FTE system. Engineered countermeasures are important but these countermeasures are only effective if wrong-way drivers understand what they indicate. The Florida driver WWD survey implemented for this research showed that more than half of the respondents did not understand the meanings of the DO NOT ENTER symbol (only 44% of respondents were correct), and only 49% of respondents understood what wrong-way pavement arrows correctly mean. Over 70% of the 900 random respondents surveyed indicate their preference to RRFBs over the BlinkLink Signs. This is important to consider when expanding the implementation of countermeasures to other sites on the FTE system. The implementation of enhanced Intelligent Transportation System (ITS) countermeasure devices shows that Florida toll road agencies are working effectively towards reducing and correcting WWD events on their toll roads' networks. Reducing the risk of WWD crashes and non-crash events in general contributes significantly to the important goal of saving lives and money.

Graduation Date





Al-Deek, Haitham


Doctor of Philosophy (Ph.D.)


College of Engineering and Computer Science


Civil, Environmental and Construction Engineering

Degree Program

Civil Engineering









Release Date

November 2021

Length of Campus-only Access

5 years

Access Status

Doctoral Dissertation (Campus-only Access)