A visualization and optimization of the impact of a severe weather disruption to an air transportation network

Highlights

• Air Transportation networks contain Big Data that must be interpreted.

• The data provides insight into the connectivity and capacity of flight routes.

• Visualizations of Big Data facilitate decision-making during disruptions.

• Disrupted airline flights must be rescheduled as soon as possible.

Abstract

The task of visualizing data becomes more challenging as the size and complexity of the data increases. Specifically, for Air transportation networks, large data sets or Big Data contain temporal and spatial information that can facilitate decision-making during disruptions. This research explores the interpretation and visualization of time-dependent flight and weather data. We focus on visualizations that characterize the impact of severe weather disruptions to the Air transportation network with respect to traffic flow capacity and route connectivity. We include a binary integer programming optimization model to propose scheduling options that minimize delays for the disrupted flights. This study analyzes airline and weather Big Data during a severe weather event and highlights air traffic management options.

Introduction

An Air transportation network is typically monitored with a sophisticated and coordinated route management system (Kohl, Larsen, Larsen, Ross, & Tiourine, 2007) that contains large volumes of time-dependent data and data sets or Big Data. For example, one day of air traffic over France contains approximately 20,000 trajectories with thousands of data points (Hunter, Martinussen, Wiggins, & O’Hare, 2011).

When a severe weather disruption impacts the Air transportation network, decision-makers typically generate scientific visualizations of time-dependent Big Data from various sources to allow underlying trends and distributions to be revealed. When Hurricane Katrina 2005 impacted more than one million people and flooded major parts of New Orleans, LA, the disruption highlighted the need for timely and meaningful visualizations that coalesce large volumes of time-varying data from a range of data sources (Venkataraman, Benger, Long, Jeong, & Renambot, 2006). Consequently, researchers combined time-varying wind speed, temperature, pressure and storm surge data to generate imagery of the path of Hurricane Katrina 2005 (Benger et al., 2006). Air transportation officials recognize that graphical displays are a critical tool for analyzing time-dependent Big Data (Becker, Eick, & Wilks, 1995) and they use the visualizations to discover solutions (Olshannikova, Ometov, Koucheryavy, & Olsson, 2015) for flight schedules recovery.

We begin with some background on Big Data and types of disruptions in Air transportation networks, and the objectives of this research. In Section 2, we discuss the literature review. In Section 3, we outline the research methodology and the experimental design of the optimization model. Section 4 presents the visualizations and discusses the results. Section 5 offers some brief concluding remarks and future extensions.

In general, Big Data means a collection of information that is high volume, velocity and variety (Holland, Thornton, & Naude, 2020). The data are on such a scale as to require an algorithmic operation to reduce the complexity of the format and to make analysis easier (Favaretto, DeClercq, Schneble, & Elger, 2020). We live in a data-driven society where large volumes of data are constantly being generated and expectations are high regarding the analysis of this data (Hurter, Conversy, Gianazza, & Telea, 2014).

The United States (US) Air transportation industry has access to and collects large volumes of consumer information and time-dependent data for flight schedules and aircraft. For example, they collect departure days and times, departure and arrival airports, number of seats on the carrier and number of flight cancellations. The airlines may use this data to track and analyze travel behavior for the purposes of expanding marketing efforts and increasing customer loyalty (Exastax, 2017). By leveraging the information obtained from collecting Big Data, airlines can make better strategic decisions that make them more competitive and customer friendly. Although researchers do not agree on a formal definition of Big Data due to the shifting and evolving concepts and computing technology (Favaretto et al., 2020), for the purposes of our study, Big Data consists of high volume, velocity and variety (Holland et al., 2020).

A disruption is an unexpected interruption to normal operations. When the Air transportation network is disrupted, operations must be restored to normal, expeditiously and efficiently. A disruption to the Air transportation network may be caused by various influences such as nature, man-made, an epidemic, or airport capacity constraints. We evaluate disruptions specifically caused by nature. Disruptions caused by nature are severe weather conditions resulting from hurricanes, tornadoes, strong wind or rain, and heavy snowfall.

The motivation for studying the impact of severe weather is influenced by the Bureau of Transportation Statistics, National Aviation System (BTS NAS) assessment of disruptions to Air transportation operations. The BTS NAS report for the 10-year period, 2007–2018, shows that weather disruptions account for the highest percentage of disruptions to Air transportation in the US. We focus on the year 2016 because it is the most recent occurrence of a high percentage of weather related disruptions (USDOT, 2016). Fig. 1 shows the BTS NAS assessment of all interruptions to Air transportation networks in the US for the calendar year 2016.

The largest contributor of disruptions to Air transportation networks in 2016 is weather at 54% and is shown in the separated slice. The second largest type of disruption is volume at 35% which indicates a disruption to airspace capacity and flight route connectivity. The remaining disruptions (equipment, closed runway and other) are significantly less of a contributor to Air transportation networks disruptions in 2016. The BTS NAS assessment for the calendar year 2016 is of particular interest for this research because in 2016, between September 28 and October 9, a severe weather event called Hurricane Matthew 2016 disrupted Air transportation networks when it made landfall in the Caribbean Islands and along the east coast of the US from Florida to Maryland. If Air Transportation officials decide to close the airport because of a severe weather disruption, current flight schedules are canceled, and future flight schedules are interrupted and possibly delayed. For example, the impact of Hurricane Matthew 2016 forced Orlando International Airport to close on October 6–7, 2016 and cancel more than 200 flights on October 6, 2016 (Ruiter, 2016). When the airport resumed operations on October 8, 2016, there continued to be a considerable number of delayed flights due to the cancellations that occurred when the airport closed.

This research seeks to provide relevant and effective visualizations of time-dependent Big Data without sacrificing the context or detail in the data. We present an analysis that enables Air transportation officials to measure and identify trends relating to traffic flow capacity and connectivity, and a model to propose scheduling options during a severe weather disruption. The research is motivated by the following questions.

• To what extent does exploratory data analysis (EDA) and scientific visualizations assist decision-makers with developing recovery actions following a severe weather event? While a variety of approaches exist for analyzing Big Data, graphical displays or visualizations are a critical tool for analysts (Becker et al., 1995). Visualizations permit the user to set aside assumptions and preconceived notions about the data and allow the data to reveal its underlying structure and distribution (Approach, 2014). There is a need for timely and meaningful visualizations that allow the user to smoothly navigate between individual details and wide-spread patterns (Klein, van der Zwan, & Telea, 2014). This question explores the identification of hidden patterns and anomalies in the data; the consequences and trends associated with the traffic flow of arrivals and departures at a hub; and, the benefits to the user in terms of enhanced understanding and clarity.

• What is the impact of a hurricane to the airport traffic flow capacity and the connectivity of flight routes? Visualizing flight and weather data leads to an understanding of the correlation between flight congestion (capacity constraints) and weather patterns (Klein et al., 2014). The assessment of capacity constraints is used to determine the ability of the network to facilitate traffic demand (flow) during disruptions (Klein et al., 2014). The connectivity of an airport network is a fundamental concern to airline decision-makers and officials due to the potentially severe consequences to airport operations when flights do not depart or arrive on time or schedule (Muriel-Villegas, Alvarez-Uribe, Patiño-Rodríguez, & Villegas, 2016). This question evaluates the effects of a hurricane to traffic flow capacity and the connectivity of flight routes in an Air transportation network.

• To what extent does using an optimization model improve flight schedules recovery and reduce delays following a severe weather disruption caused by a hurricane? Generating recovery solutions is a complex task and performing a sensitivity analysis is essential to the process (Zhang, Henry Lau, & Yu, 2015a). Disruptions to flight schedules have operational and economic consequences, therefore, it is essential that decision-makers identify quick and reliable recovery solutions (Eggenberg, Salani, & Bierlaire, 2010). This question examines the robustness of the optimization model, the feasibility of rescheduling flights in the first iteration of the model, the total delay associated with the new flight schedules and how quickly the new flight schedules are generated.