Rapid communication during extreme events is one of the critical aspects of successful disaster management strategies. Due to their ubiquitous nature, social media platforms offer a unique opportunity for crisis communication. Moreover, social media usage on GPS enabled devices such as smartphones allow us to collect human movement data which can help understanding mobility during a disaster. This study leverages social media (Twitter) data to understand the effectiveness of social media-based communication and the resilience of human mobility during a disaster. This thesis has two major contributions. First, about 52.5 million tweets related to hurricane Sandy are analyzed to assess the effectiveness of social media communication during disasters and identify the contributing factors leading to effective crisis communication strategies. Effectiveness of a social media user is defined as the ratio of attention gained over the number of tweets posted. A model is developed to explain more effective users based on several relevant features. Results indicate that during a disaster event, only few social media users become highly effective in gaining attention. In addition, effectiveness does not depend on the frequency of tweeting activity only; instead it depends on the number of followers and friends, user category, bot score (controlled by a human or a machine), and activity patterns (predictability of activity frequency). Second, to quantify the impacts of an extreme event to human movements, we introduce the concept of mobility resilience which is defined as the ability of a mobility infrastructure system to manage shocks and return to a steady state in response to an extreme event. We present a method to detect extreme events from geo-located movement data and to measure mobility resilience and loss of resilience due to those events. Applying this method, we measure resilience metrics from geo-located social media data for multiple types of disasters occurred all over the world. Quantifying mobility resilience may help us to assess the higher-order socio-economic impacts of extreme events and guide policies towards developing resilient infrastructures as well as a nation's overall disaster resilience.

The effect of low visibility on both crash occurrence and severity is a major concern in the traffic safety field. Different approaches were utilized in this research to analyze the effects of fog on traffic safety and evaluate the effectiveness of different fog countermeasures. First, a "Crash Risk Increase Indicator (CRII)" was proposed to explore the differences of crash risk between fog and clear conditions. A binary logistic regression model was applied to link the increase of crash risk with traffic flow characteristics. Second, a new algorithm was proposed to evaluate the rear-end crash risk under fog conditions. Logistic and negative binomial models were estimated in order to explore the relationship between the potential of rear-end crashes and the reduced visibility together with other traffic parameters. Moreover, the effectiveness of real-time fog warning systems was assessed by quantifying and characterizing drivers' speed adjustments through driving simulator experiments. A hierarchical assessment concept was suggested to explore the drivers' speed adjustment maneuvers. Two linear regression models and one hurdle beta regression model were estimated for the indexes. Also, another driving simulator experiment was conducted to explore the effectiveness of Connected-Vehicles (CV) crash warning systems on the drivers' awareness of the imminent situation ahead to take timely crash avoidance action(s). Finally, a micro-simulation experiment was also conducted to evaluate the safety benefits of a proposed Variable Speed limit (VSL) strategy and CV technologies. The proposed VSL strategy and CV technologies were implemented and tested for a freeway section through the micro-simulation software VISSIM. The results of the above mentioned studies showed the impact of reduced visibility on traffic safety, and the effectiveness of different fog countermeasures.

Since the beginning of signalization of intersections, the management of traffic congestion is one of most critical challenges specifically for the city and urbanized area. Almost all the municipal agencies struggle to manage the perplexities associated with traffic congestion or signal control. The Adaptive Traffic Control System (ATCS), an advanced and major technological component of the Intelligent Transportation Systems (ITS) is considered the most dynamic and real-time traffic management technology and has potential to effectively manage rapidly varying traffic flow relative to the current state-of-the-art traffic management practices. InSync ATCS is deployed in multiple states throughout the US and expanding on a large scale. Although there had been several 'Measure of Effectiveness' studies performed previously, the performance of InSync is not unquestionable especially because the previous studies failed to subject for multiple environments, approaches, and variables. Most studies are accomplished through a single approach using simple/naïve before-after method without any control group/parameter. They also lacked ample statistical analysis, historical, maturation and regression artifacts. An attempt to evaluate the InSync ATCS in varying conditions through multiple approaches was undertaken for the SR-434 and Lake Underhill corridor in Orange County, Florida. A before-after study with an adjacent corridor as control group and volume as a control parameter has been performed where data of multiple variables were collected by three distinct procedures. The average/floating-car method was utilized as a rudimentary data collection process and 'BlueMac' and 'InSync' system database was considered as secondary data sources. Data collected for three times a day for weekdays and weekends before and after the InSync ATCS was deployed. Results show variation in both performance and scale. It proved ineffective in some of the cases, especially for the left turns, total intersection queue/delay and when the intersection volumes approach capacity. The results are verified through appropriate statistical analysis.

The main goal of this dissertation is to have better understanding of design and operation of the Continuous Flow Intersection (CFI) and Diverging Diamond Interchange (DDI) - as well as numerous factors that affect signalized intersection and interchange performance due to increased left-turn demand. The dissertation attempts to assess the need and justification to redesign intersections and interchanges to improve their efficiency. And to that end, an extensive literature review of existing studies was done with the prime aim of perceiving the principles of these innovative designs and determining the methodology to-be-followed, in order to reach the study's core. Accordingly, several DDI and CFI locations were selected as candidate locations, where the designs have already been implemented and the required data - to model calibration and validation - was collected. The micro-simulation software (VISSIM 8.0) was used for simulation, calibration and validation of the existing conditions - through several steps - including signal optimization and driving behavior parameter sensitivity analysis. Subsequently, an experiment was conceived for each design, aiming at examining several factors that affect each design's efficiency. The experiment comprised 180 and 90 different CFI & DDI scenarios and their conventional designs, respectively. Two measures of effectiveness were identified for result analysis: the average delay and capacity. Result analyses were performed to detect switching thresholds (from conventional to innovative designs. In addition, performance comparison studies of the CFI and DDI with their conventional designs were performed. The results and findings will serve as guidelines for decision-makers as to when they should consider switching from conventional to innovative design. Finally, decision support systems were developed to speed up the search for the superior design, in comparison with others.

One main interest in crash frequency modeling is to predict crash counts over a spatial domain of interest (e.g., traffic analysis zones (TAZs)). The macro-level crash prediction models can assist transportation planners with a comprehensive perspective to consider safety in the long-range transportation planning process. Most of the previous studies that have examined traffic crashes at the macro-level are related to high-income countries, whereas there is a lack of similar studies among lower- and middle-income countries where most road traffic deaths (90%) occur. This includes Middle Eastern countries, necessitating a thorough investigation and diagnosis of the issues and factors instigating traffic crashes in the region in order to reduce these serious traffic crashes. Since pedestrians are more vulnerable to traffic crashes compared to other road users, especially in this region, a safety investigation of pedestrian crashes is crucial to improving traffic safety. Riyadh, Saudi Arabia, which is one of the largest Middle East metropolises, is used as an example to reflect the representation of these countries' characteristics, where Saudi Arabia has a rather distinct situation in that it is considered a high-income country, and yet it has the highest rate of traffic fatalities compared to their high-income counterparts. Therefore, in this research, several statistical methods are used to investigate the association between traffic crash frequency and contributing factors of crash data, which are characterized by 1) geographical referencing (i.e., observed at specific locations) or spatially varying over geographic units when modeled; 2) correlation between different response variables (e.g., crash counts by severity or type levels); and 3) temporally correlated. A Bayesian multivariate spatial model is developed for predicting crash counts by severity and type. Therefore, based on the findings of this study, policy makers would be able to suggest appropriate safety countermeasures for each type of crash in each zone.

Ubiquitous uses of social media platforms in smartphones have created an opportunity to gather digital traces of individual activities at a large scale. Traditional travel surveys fall short in collecting longitudinal travel behavior data for a large number of people in a cost effective way, especially for the transient population such as tourists. This study presents an innovating methodological framework, using machine learning and econometric approaches, to gather and analyze location-based social media (LBSM) data to understand individual destination choices. First, using Twitter's search interface, we have collected Twitter posts of nearly 156,000 users for the state of Florida. We have adopted several filtering techniques to create a reliable sample from noisy Twitter data. An ensemble classification technique is proposed to classify tourists and residents from user coordinates. The performance of the proposed classifier has been validated using manually labeled data and compared against the state-of-the-art classification methods. Second, using different clustering methods, we have analyzed the spatial distributions of destination choices of tourists and residents. The clusters from tourist destinations revealed most popular tourist spots including emerging tourist attractions in Florida. Third, to predict a tourist's next destination type, we have estimated a Conditional Random Field (CRF) model with reasonable accuracy. Fourth, to analyze resident destination choice behavior, this study proposes an extensive data merging operation among the collected Twitter data and different geographic database from state level data libraries. We have estimated a Panel Latent Segmentation Multinomial Logit (PLSMNL) model to find the characteristics affecting individual destination choices. The proposed PLSMNL model is found to better explain the effects of variables on destination choices compared to trip-specific Multinomial Logit Models. The findings of this study show the potential of LBSM data in future transportation and planning studies where collecting individual activity data is expensive.

Fog-related crashes continue to be one of the most serious traffic safety problems in Florida. Based on the historical crash data, we found that single-vehicle crashes have the highest severity among all types of crashes under fog conditions. This study first analyzed the contributing factors of the fog-related single-vehicle crashes' (i.e., off road/rollover/other) severity in Florida from 2011 to 2014 using association rules mining. The results show that lane departure distracted driving, wet road surface, and dark without road light are the main contributing factors to severe fog-related single vehicle crashes. Some suggested countermeasures were also provided to reduce the risk of fog-related single vehicle crashes. Since lane departure is one of the most important contributing factors to the single-vehicle crashes, an advanced warning system for lane departure under connected vehicle system was tested in driving simulation experiments. The system was designed based on the Vehicle-to-Infrastructure (V2I) with the concept of Augmented Reality (AR) using Head-Up Display (HUD). The results show that the warning with sound would reduce the lane departure and speed at curves, which would enhance the safety under fog conditions. In addition, the warning system was more effective for female drivers.

In recent years, there is growing recognition that common unobserved factors that influence crash frequency by one attribute level are also likely to influence crash frequency by other attribute levels. The most common approach employed to address the potential unobserved heterogeneity in safety literature is the development of multivariate crash frequency models. The current study proposes an alternative joint econometric framework to accommodate for the presence of unobserved heterogeneity – referred to as joint negative binomial-multinomial logit fractional split (NB-MNLFS) model. Furthermore, the study undertakes a first of its kind comparison exercise between the most commonly used multivariate model (multivariate random parameter negative binomial model) and the proposed joint approach by generating an equivalent log-likelihood measure. The empirical analysis is based on the zonal level crash count data for different collision types from the state of Florida for the year 2015. The model results highlight the presence of common unobserved effects affecting the two components of the joint model as well as the presence of parameter heterogeneity. The equivalent log-likelihood and goodness of fit measures clearly highlight the superiority of the proposed joint model over the commonly used multivariate approach.

The impact of freight on our transportation system is further accentuated by the fact that trucks consume greater roadway capacity and therefore cause more significant problems including traffic congestion, delay, crashes, air pollution, fuel consumption, and pavement damage. Assessing the actual effects of truck traffic is a growing need to support the ability to safely and efficiently move goods and people in areas where roadway expansion is not the best option. On one hand, trucks need to efficiently serve commerce and industry, while at the same time their activities need not contribute to a decline in the quality or public safety. In the current practice, to the best of the authors' knowledge, there is no framework methodology for real-time management of traffic, specifically on truck routes, to reduce travel duration and avoid truck travel delays due to non-recurring congestion (i.e. traffic incidents) and to estimate impacts on traffic flows, economy, and environment. The objective of this study is to develop a truck routing strategy and to quantify its' impacts on travel time, emissions and consequently assess the effects on the economy and environment. In order to estimate non-recurrent congestion based travel delay and fuel consumption by real-time truck routing simulation models, significant corridors with high truck percentages were selected. Furthermore, tailpipe emissions (on-site) due to traveled distance and idling are estimated via MOVES emissions simulator software. Economic Input Output-Life Cycle Assessment Model is utilized to gather fuel consumption related upstream (off-site) emissions. Simulation results of various scenarios indicated that potential annual value of time savings can reach up to $1.67 million per selected corridor. Consistently, fuel costs and emission values are lower, even though extra miles are traveled on the alternative route. In conclusion, our study confirms that truck routing strategies in incident conditions have high economic and environmental impacts.

As traffic problems on roadways have been increasing, active traffic management systems (ATM) using proactive traffic management concept have been deployed on freeways and arterials. The ATM aims to integrate and automate various traffic control strategies such as variable speed limits, queue warning, and ramp metering through a decision support system (DSS). Over the past decade, there have been many efforts to integrate freeways and arterials for the efficient operation of roadway networks. It has been required that these systems should prove their effectiveness in terms of travel time reliability. Therefore, this study aims to develop a new concept of a decision support system integrating variable speed limits, queue warning, and ramp metering on the basis of travel time reliability of freeways and arterials. Regarding the data preparation, in addition to collecting multiple data sources such as traffic data, crash data and so on, the types of traffic data sources that can be applied for the analysis of travel time reliability were investigated. Although there are many kinds of real-time traffic data from third-party traffic data providers, it was confirmed that these data cannot represent true travel time reliability through the comparative analysis of measures of travel time reliability. Related to weather data, it was proven that nationwide land-based weather stations could be applicable. Since travel time reliability can be measured by using long-term periods for more than six months, it is necessary to develop models to estimate travel time reliability through real-time traffic data and event-related data. Among various matrix to measure travel time reliability, the standard deviation of travel time rate [minute/mile] representing travel time variability was chosen because it can represent travel time variability of both link and network level. Several models were developed to estimate the standard deviation of travel time rate through average travel time rate, the number of lanes, speed limits, and the amount of rainfall. Finally, a DSS using a model predictive control method to integrate multiple traffic control measures was developed and evaluated. As a representative model predictive control, METANET model was chosen, which can include variable speed limit, queue warning, and ramp metering, separately or combined. The developed DSS identified a proper response plan by comparing travel time reliability among multiple combinations of current and new response values of strategies. In the end, it was found that the DSS provided the reduction of travel time and improvement of its reliability for travelers through the recommended response plans.