O’Hare Airport: An Investigation of the Relationship between Airline and Road Traffic By Dustin Parker An Undergraduate Thesis Submitted in Partial Fulfillment for the Requirements of Bachelor of Arts in Geography and Earth Science Advised by Dr. Wenjie Sun Carthage College Kenosha, WI April, 2013 Abstract Geography Information Systems (GIS) is used highly in transportation in order to improve traffic flow for the future. Chicago O’Hare is one of the busiest airports in the world. Thousands of airlines fly in and out of the airport monthly. With so many arrivals and departures of airlines comes an ample supply of automotive traffic. Often there are traffic jams in Chicago and near O’Hare. Some of these traffic jams might be related to airline traffic at O’Hare airport. There might be a correlation between the number of airline arrivals and departures with the traffic count near the airport. There could be a time lag in the correlation as well. The time lag might depend on if the flight is arriving or departing. The transportation industry could benefit from information like this. If there is a correlation it would be possible to determine how much traffic would be traveling through due to the airport. This would make estimating appropriate highway sizes easier for locations near airports. 2 Table of Contents Table of Contents Abstract ........................................................................................................................................................ 2 List of Figures ................................................................................................................................................ 4 List of Tables ................................................................................................................................................. 4 Introduction and Literature Review ............................................................................................................ 5 Problem Statement .................................................................................................................................... 10 Methodology .............................................................................................................................................. 11 Data Acquisition ...................................................................................................................................... 11 Correlation Analysis................................................................................................................................. 13 Traffic Count Sample Points .................................................................................................................... 14 Results......................................................................................................................................................... 18 Airline Departures and Traffic Count Comparison .................................................................................. 18 Airline Arrivals and Traffic Count Comparison ........................................................................................ 22 Correlation Analysis ................................................................................................................................ 26 Airline Departures and Short Distance Traffic .................................................................................... 26 Airline Departures and Long Distance Traffic ..................................................................................... 27 Airline Arrivals and Short Distance Traffic .......................................................................................... 28 Airline Arrivals and Long Distance Traffic ........................................................................................... 29 Conclusion .................................................................................................................................................. 30 Discussion................................................................................................................................................ 30 Outlook and Future Direction ................................................................................................................. 34 Acknowledgments ...................................................................................................................................... 36 Appendix..................................................................................................................................................... 36 References .................................................................................................................................................. 41 3 List of Figures Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 15 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 Figure 1 - Location of all 6 sample points ................................................................................................... 16 List of Tables Table 1 - Airline Departures and Short Distance Traffic Correlations......................................................... 27 Table 2 - Airline Departures and Long Distance Traffic Correlations .......................................................... 28 Table 3 - Airline Arrivals and Short Distance Traffic Correlations ............................................................... 29 Table 4 - Airline Arrivals and Long Distance Traffic Correlations ................................................................ 30 4 Introduction and Literature Review The goal of this thesis is to look at the correlation between traffic levels on highways at varying distances from Chicago O’Hare airport and numbers of airline arrivals and departures at O’Hare International Airport. Hourly highway traffic data will be taken along with hourly airline arrivals and departures for a total of six days worth of data. Multiple methods will be used to obtain an accurate result. Not many articles or research have been published attempting to find a correlation between traffic levels and airline arrival and departure volumes at airports. Currently there are very few if any books that concentrate exclusively on geographic information systems and transportation (GIS-T) (Goodchild, 2000). However, the methods that will be used to find this correlation in my thesis have been used extensively. The literature review will cover similar methods that will be needed to properly analyze the data for this thesis. Some articles relating to transportation geography provide important insights. Different trends of uses for transportation geography include using traffic data for statistical purposes, finding specific challenges of the subject and how to combat them in the future. By looking at these works of literature on the subject it is possible to see specific methods that could be used to find a correlation between road and airline traffic and in some cases how to improve them. 5 Chicago O’Hare was named in 1945 after Lieutenant Commander Edward O’Hare who was from Chicago, Illinois and fought in World War II. It wasn’t until 1955 that O’Hare official opened for commercial flights. The airport grew very quickly to become one of the busiest in the world. By around 1970 there were about 40 million travelers passing through the airport a year. Currently Chicago O’Hare is still one of the busiest airports in the world and the second busiest in the United States (Lewis, 2011). It is hard to determine when exactly the term geographic information systems was first used, but GIS-based ideas emerged in the early 1960s. Goodchild (2000, Pg 1) explains that a group of graduate students in Quantitative Geography at the University of Washington in the late 1950s had a major influence on the start of GIS. One student, Duane Marble, possibly had the largest impact. She created a form of GIS in order to study transportation in Chicago. Although this form of GIS she made was basic compared to current forms, it can be considered one of the first times GIS was used. Using GIS-T and analyzing data from it requires many different techniques and approaches. Goodchild explains how current forms of GIS-T consist of three different “views” including “the map view,” “the navigational view,” and “the behavioral view,” (Goodchild, Pg 2). The map view uses two dimensional views, often converting them from one dimension. An example would be a street address that needs to be converted from one dimension to two dimensional coordinates so it can be placed on a map (Goodchild, pg 2). In the 1970s a system of links and nodes was used to represent the street network. Although this system was easy to check for mistakes (since the links must be closed) it had multiple disadvantages. One was that 6 it was difficult to create intersections. Goodchild explains how the map view can be limited since real features need to be represented as one dimensional spaces or centerlines (Goodchild, pg 3). The navigation view is based off the “link / node system”. The “link / node system” is essential to the navigation view since navigation requires following a set path (Goodchild, pg 4). This thesis will use all three views. The behavioral view deals with only discrete objects such as vehicles, airplanes, and trains. This view will obviously help when analyzing airline and road traffic in this thesis. The navigation view will also be used since traffic flow to and from Chicago O’Hare follows certain path just like a link / node system. The final product of this thesis will use a map view to display the data. Transportation will keep growing in the future; therefore it is important to find improved methods for the future. Rodrigue explains that growing demand from increasing population, reduction of costs as technology improves, and expansion of infrastructures will cause the importance of transportation to grow. GIS-T will be very useful in helping to find these methods. By finding a correlation between air and road traffic this will allow better judgment regarding how to design highways near Chicago O’Hare and other airports for improved traffic flow. This information could be very useful for highway designers and the future of the transportation system. GIS-T research and data is currently fairly scarce, but will continue to grow in the future according to many researchers such as Rodrigue and Goodchild have explained. Airline departure and arrival times at airports can have patterns by themselves. In order to see a correlation between airline and automotive traffic it is important to take into account 7 the fluctuations airline traffic can have alone. Li Shan-Mei (2012) looks at only airline arrival and departure traffic for multiple United States airports and airlines. This article is very helpful since it provides information on how airline arrivals and departure patterns at Chicago O’Hare could look like and potentially what can be expected when recording data. Li Shan-Mei found that there are a number of reasons for airline fluctuations at certain airports (Li Shan-Mei, Pg 1). The time scale that is used to measure data has a significant effect. At earlier hours in the day after midnight and 0600 there is little airline traffic. This means that the data collected in this time range might skew the results if only these times are used. Environmental influences can have a major effect on airline traffic as well. Li Shan-Mei explains how severe and nonsevere weather can influence airline traffic flow patterns (Li Shan-Mei, Pg 5). In order to accurately measure the correlation of airline and automotive traffic the time scale and environmental issues will have to be taken into account. Lewis (2011) also covers only airline traffic. Her study uses monthly data in order to analyze the effects of wind at O’Hare airport. Obviously this data will be less specific than intended for finding a correlation between airline and automotive traffic. Both Li Shan-Mei’s and Lewis’ articles will be used to see what can be expected in fluctuations in airline traffic. Lewis starts out by going through the history of the airport and the uses of the runway. Her goal is to show the effective use of the runway in relation to the wind. Each runway is mapped using GIS to show how it is used (airline arrivals and departures). Airline traffic is measured each month for seven different years (Lewis, pg 39-41). The data used is not very specific but covers a large time period. Lewis’s study can be helpful in determining when to record data of 8 airline traffic. Some months may be different from others in terms of airline traffic. It is important to record data that is not influenced by outside factors (wind / weather, time of day). Lewis and Li Shan-Mei’s studies provide valuable information to keep in mind in order to find a certain time period to record data. Mei found that weather can affect airlines differently depending on the month. For example the weather in June affects flights more than in April (Li Shan Mei, Pg 5). The affects of weather and other outside influences need to be considered when trying to find a correlation between airline and road traffic. Both Lewis and Li Shan-Mei’s articles can be expanded upon by using their data and comparing it to hourly road traffic since they only analyze airline traffic. Automotive traffic near O’Hare, just like airline traffic, is influenced by many events. This is important to note in this thesis since unusual events that influence data should be avoided. It is important to look at a time period that is similar to the average pattern throughout any year at O’Hare. For example around Christmas traffic patterns around O’Hare may change drastically over average yearly conditions. Airline traffic, severe weather, holidays and other outside influences should be avoided when collecting data in order to eliminate as many outside influences as possible. The less outside influences there are the easier it will be to see what the correlation is between road and air traffic. Overall the main contributions that the literature has made are techniques and ways to use data in transportation geography. The articles explain the advantages and disadvantages of using certain methods along with explaining some possible new methods. Obviously relating road and air traffic is a field that can be expanded on since none of the articles found do so. 9 The articles found either relate to road or air traffic, but never both. Since the intention of this thesis is to look at road and air traffic this might help uncovered new territory in GIS-T. One of the main challenges to transportation and geography is finding data for GIS. Currently the amount of data is limited, but it is growing. Finding data in the future is bound to become easier since studies on the subject will increase over time according to current trends. These all can be used in finding a correlation between road traffic levels near O’Hare and airline arrival and departure times. Problem Statement The purpose of this project is to find whether or not there is a correlation between hourly highway traffic volume near O’Hare and the number of airline arrivals and departures over a 24-hour period. To be more specific, four sets of alternative (H1) and null (H0) hypotheses will be tested. #1 o H1: Road traffic levels will increase after airline arrivals o H0: Road traffic levels will not increase #2 o H1: Road traffic levels will increase before airline departures o H0: Road traffic levels will not increase #3 o H1: The further away the longer the time lag is between road and air traffic o H0: Time lag will not increase with distance away from the airport 10 #4 o H1: The closer to the airport the stronger the correlation will be between road and air traffic o H0: The correlation between road and air traffic will not decrease with distance from the airport Methodology Data Acquisition This thesis will be looking at road and air traffic count at Chicago O’Hare airport. The arrival and departure times for each flight at Chicago O’Hare will be used. Traffic will be recorded for a number of specific points located on highways around O’Hare airport. Once all the data is collected, the air and road traffic count will be recorded in a time series and mapped in ArcGIS. A correlation coefficient analysis and a significance testing will be run in order to tell if there is a statistically significant correlation between the time series of air and road traffic and if yes how strong it is. The main data source that will be used for traffic count is acquired from the Illinois Department of Transportation (IDOT). Traffic levels will be recorded for 24 hours each day and will be recorded on six different days. This data will then be stored in Excel™ so it can be compared to air traffic. A total of six traffic points will be used that are located on highways around Chicago O’Hare. Four of the points will be located at a distance under an hour drive from the airport. The other two will be located at a distance about an hour or more drive from the airport. 11 The points that are under an hour drive (short distance) to the airport should see the most correlation with air traffic than the points further than an hour away (long distance). The reason the traffic points are separated by under an hour and over an hour from the airport is because the traffic data is recorded by hour. If a long distance point is under an hour away from the airport its data might overlap with that of the short distance point. There will be many more influences other than O’Hare for the long distance points since they are located further from the airport. However, the long distance points serve a very important purpose of acting as a control for this research. By comparing the correlations of the long distance points and the short distance points it will be easier to see how the airport influences the traffic. Airline traffic from O’Hare will be obtained from Research and Innovative Technologies Administration (RITA). The scheduled times for each arrival and departure flight will be used. Using actual flight times would not make sense since people travel to the airport expecting to leave at the scheduled time. People determine what time they leave based on the scheduled time of their flight. Each airline arrival and departure count at O’Hare will be recorded 24 hours a day. There are many airlines at O’Hare, but only the major (most flights per day) airlines will be used. A total of eleven airline companies will be used in this thesis. These airlines include: American Airlines, American Eagle, Atlantic South East Airlines, Continental Airlines, Delta Airlines, ExpressJet Airlines, JetBlue Airways, Mesa Airlines, SkyWest Airlines, United Airlines and US Airlines. The number of arrivals and departures of each airline will need to be recorded and transferred to an Excel™ table. 12 Correlation Analysis In order to determine if there is a correlation between air and road traffic, the correlation coefficient (r) between airline arrivals and departures and road traffic was calculated. Separate correlation tests were conducted for the points located in the “short distance” category and the “long distance” category. The correlation coefficient for each data set is tested using a two tailed t-test. The t-test gives a certain t value. There are critical values that the t value must attain in order for there to be enough statistical evidence to show a significant correlation. Given the alpha value and the degrees of freedom the critical value of the data set can be determined. If the t value falls above the critical value there is a 99% probability it did not occur by chance. An r value can be between 1 and -1. The closer it is to 1 the higher degree of a positive correlation. The closer it is to -1 the more negative the correlation. A few different correlation coefficients will be calculated for each point in order to find the strongest r value. Multiple shifts in time for traffic will be used in order to compare the r values. When comparing traffic with airline arrivals, traffic times will be shifted ahead a certain number of hours. For example a one hour shift ahead would be traffic levels at 15:00 – 16:00 compared to airline arrivals at 14:00 – 15:00. When comparing traffic with airline departures, traffic times will be shifted back a certain number of hours. An example of a one hour shift back would be traffic levels at 13:00 – 14:00 compared to airline departures at 14:00 – 15:00. It is important to test these shifts in time since airline arrivals and departures at a certain time might affect traffic level at a different time. For example, passengers for flights that are departing will need to arrive at the airport well ahead of their flight’s departure. By testing 13 different shifts in time it will be possible to see a correlation of how early passengers have to arrive. Traffic level could be affected by the airline departure an hour, two hours or some other time before it takes off. By using these different time shifts when testing r, it is possible to determine whether or not to reject null hypothesis one and null hypothesis two. By comparing the r values of the long and short points, it will be possible to determine whether or not to reject null hypothesis three and null hypothesis four. Traffic Count Sample Points Each point that was chosen for a traffic count is shown in Figure 1 below and the Appendix in more detail. There were a total of six traffic count points on different days located a certain distance from the airport. It is important to note that the traffic data from some sample points was recorded on different days than other points. All the traffic count data was recorded within the same year, but some points have data from a different month. The traffic data was limited to only certain days; therefore, all the recorded traffic data are not within the same month. If traffic data for several days occurring in the same week / month were recorded this could have increased accuracy. Obtaining traffic data that is closer together in time likely increases accuracy because there are probable seasonal and weekly fluctuations of traffic levels throughout the year. The fluctuations can occur from weather, time of the season, events, week day vs. weekend, holidays, etc. By having data that is recorded on the same day of the week or month, it reduces the chances of having more outside influences (other than airline arrivals and departures) affecting the traffic level. Unfortunately, due to how limited the data was, some points were recorded in different months. 14 Four of the points were located less than 15 miles from the airport. Two other traffic count points were chosen that were further than 39 miles from the airport. The distance from the airport refers to the driving distance the cars would have to travel on the highways. Traffic count points located closer to the airport should have less additional influences beyond airline departures and arrivals than traffic points located further away. The data from these points were organized into a “short distance” vs. a “long distance” group in testing the correlation coefficient (r) between the highway traffic and airline departures and arrivals. 15 Figure 1 - Location of all 6 sample points Figure 1 on the previous page shows all six traffic sample points. The red airplane shown on the map is the location of Chicago O’Hare airport. Each traffic point is marked with a 16 yellow dot and labeled with the distance it is from the airport. For a more detailed map of where each traffic point is located, refer to the appendix. The point that is located .6 miles from the entrance of O’Hare Airport most likely contains only cars traveling to and from the airport. It is located on a section of highway that is in the O’Hare area. There are no other paths that cars at this point travel to other than from or to the airport. This can be considered the most reliable point since it has no other outside influences. Traffic for this point was recorded on 6/14/2011. The point located 3.2 miles from the airport entrance is on Kennedy Expressway. This is the second closest traffic count location. The traffic data for this point was recorded on 6/09/2011. Traffic for the point located 9.4 miles away on Kennedy Expressway was recorded on 8/01/2011. The final point in the “close to the airport” category is located 12.6 miles from the airport on Dwight D. Eisenhower Expressway. This traffic data was recorded on 6/27/2011. The 40 mile point on the map is in the “long distance” category and is located on I-94/US-41 north of the airport. Its data was recorded on 4/18/2011. This point is probably around an hour drive from the airport. If it were possible, a point further than this would have been chosen, but this was the furthest point that could be found located on I-94/US-41. This point might be close to overlapping with the short distance category, but due to the limitations in data this point needs to be used for a long distance point. The second “long distance” point is located 79 miles south of the airport (as shown on the map) on I-57. This point is located well over an hour drive from the airport. The data for this point is unlikely to overlap with the close distance points data. The data for this point was recorded on 6/22/2011. 17 The traffic data recorded at each point is split into to and from the airport. For example points that are located east of the airport are split into westbound traffic and eastbound traffic. Westbound traffic would be traveling toward the airport and eastbound traffic would be traveling away from the airport. Traffic traveling toward the airport will be compared to departure flights since passengers will be traveling toward the airport for their flights. Traffic that is traveling away from the airport will be compared to airline arrivals since passengers will be leaving the airport from their flights. 18 Results Airline Departures and Traffic Count Comparison Below are multiple graphs (Figures 2-7) comparing the airline departures each day with the traffic count at each highway sample point. The graphs below display the hourly traffic count and airline departures displayed on one graph on the same day. The graphs show hourly data of each individual day from 01:00 (1 am) to 24:00 (midnight). The traffic data in the graphs are not shifted. In other words, traffic that occurred at 07:00-0:800 is compared to airline arrivals that happened at 07:00 - 08:00. By displaying the traffic count and airline departures in one graph it is easier to see a possible correlation between the two and the corresponding time lag between the two. 4000 80 3500 70 3000 60 2500 50 2000 40 1500 30 1000 20 500 10 0 0 1 3 5 7 9 Airline Departures Per Hour Traffic Count Per Hour Airline Departures and 0.6 Mile Point (6/14) Traffic Count (.6 Mile Point) Airlines Departures 11 13 15 17 19 21 23 Time of Day (Hours) Figure 2 - Airline Departures compared to the traffic count on 6/14/2011 for the point located 0.6 miles away on the Kennedy Expressway 19 7000 80 6000 70 5000 60 50 4000 40 3000 30 2000 20 1000 10 0 0 1 3 5 7 Airline Departures Per Hour Traffic Count Per Hour Airline Departures and 3.2 Mile Point (6/09) Traffic Count (3.2 Mile Point) Airlines Departures 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 3 – Airline departures compared to the traffic count on 6/09/2011 for the point located 3.2 miles away Traffic Count Per Hour 10000 80 70 8000 60 50 6000 40 4000 30 20 2000 10 0 0 1 3 5 7 Airline Departures Per Hour Airline Departures and 9.4 Mile Point (8/01) Traffic Count (9.4 Mile Point) Airlines Departures 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 4 - Airline departures compared to the traffic count on 8/01/2011 for the point located 9.4 miles away 20 7000 80 6000 70 5000 60 50 4000 40 3000 30 2000 20 1000 10 0 0 1 3 5 7 Airline Departures Per Hour Traffic Count Per Hour Airline Departures and 12.6 Mile Point (6/27) Traffic Count (12.6 Mile Point) Airlines Departures 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 5 - Airline departures compared to the traffic count on 6/27/2011 for the point located 12.6 miles away Traffic Count Per Hour 3500 90 80 70 60 50 40 30 20 10 0 3000 2500 2000 1500 1000 500 0 1 3 5 7 Airline Departures Per Hour Airline Departures and 40 Mile Point (4/18) Traffic Count (40 Mile Point) Airlines Departures 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 6 - Airline departures compared to the traffic count on 4/18/2011 for the point located 40 miles away 21 800 80 700 70 600 60 500 50 400 40 300 30 200 20 100 10 0 0 1 3 5 7 9 Airline Departures Per Hour Traffic Count Per Hour Airline Departures and 79 Mile Point (6/22) Traffic Count (79 Mile Point) Airlines Departures 11 13 15 17 19 21 23 Time of Day (Hours) Figure 7 - Airline departures compared to the traffic count on 6/22/2011 for the point located 79 miles away 22 Airline Arrivals and Traffic Count Comparison Below are multiple graphs (Figures 8-13) comparing the airline arrivals each day with the traffic count at each sample point. Just like the airline departure graphs the hourly traffic data for the graphs below are not shifted. By displaying the traffic data and airline arrivals right next to each other it provides a better perspective on the correlation and possible time lag. 3500 80 3000 70 2500 60 50 2000 40 1500 30 1000 20 500 10 0 0 1 3 5 7 Airline Arrivals Per Hour Traffic Count Per Hour Airline Arrivals and 0.6 Mile Point (6/14) Traffic Count (.6 Mile Point) Airlines Arrivals 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 8 - Airline Arrivals compared to the traffic count on 6/14/2011 for the point located .6 miles away 23 7000 80 6000 70 5000 60 50 4000 40 3000 30 2000 20 1000 10 0 Airline Arrivals Per Hour Traffic Count Per Hour Airline Arrivals and 3.2 Mile Point (6/09) Traffic Count (3.2 Mile Point) Airlines Arrivals 0 1 3 5 7 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 9 - Airline Arrivals compared to the traffic count on 6/09/2011 for the point located 3.2 miles away 8000 80 7000 70 6000 60 5000 50 4000 40 3000 30 2000 20 1000 10 0 Airline Arrivals Per Hour Traffic Count Per Hour Airline Arrivals and 9.4 Mile Point (8/01) Traffic Count (9.4 Mile Point) Airlines Arrivals 0 1 3 5 7 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 10 - Airline Arrivals compared to the traffic count on 8/01/2011 for the point located 9.4 miles away 24 Airline Arrivals and 12.6 Mile Point (6/27) 80 70 5000 60 4000 50 3000 40 30 2000 20 1000 10 0 Airline Arrivals Per Hour Traffic Count Per Hour 6000 Traffic Count (12.6 Mile Point) Airlines Arrivals 0 1 3 5 7 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 11 - Airline Arrivals compared to the traffic count on 6/27/2011 for the point located 12.6 miles away 4000 80 3500 70 3000 60 2500 50 2000 40 1500 30 1000 20 500 10 0 Airline Arrivals Per Hour Traffic Count Per Hour Airlines Arrivals and 40 Mile Point (4/18) Traffic Count (40 Mile Point) Airlines Arrivals 0 1 3 5 7 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 12 - Airline Arrivals compared to the traffic count on 4/18/2011 for the point located 40 miles away 25 800 80 700 70 600 60 500 50 400 40 300 30 200 20 100 10 0 Airline Arrivals Per Hour Traffic Count Per Hour Airline Arrivals and 79 Mile Point (6/22) Traffic Count (79 Mile Point) Airlines Arrivals 0 1 3 5 7 9 11 13 15 17 19 21 23 Time of Day (Hours) Figure 13 - Airline Arrivals compared to the traffic count on 6/22/2011 for the point located 79 miles away 26 Correlation Analysis Airline Departures and Short Distance Traffic The strongest r value for short distance traffic points compared to airline departures were when the time for traffic was shifted back one hour and compared to airline departures. The r value highlighted in red in table 1 below has the strongest correlation. An r value of .751 was found. The r value was then used in a t-test resulting in a t value of 11.031. The critical value for the t-test is 2.640 (as shown in table 1). Since the t-value is greater than the critical value there is ample evidence of a correlation between the data for airline departures and traffic near (short distance points) the airport. As seen in the table below there is still a correlation for the other values. However, the strongest and most statistically significant correlation is when the time for the traffic was shifted back one hour. Table 1 Correlation 0.298111838 0.519575585 0.688266089 0.751108287 0.702653923 Time Shift t Value Critical Value Shifted 4 hours back 3.028 2.64 Shifted 3 hours back 5.896 2.64 Shifted 2 hours back 9.198 2.64 Shifted 1 hour back 11.031 2.64 not shifted 9.574 2.64 27 Airline Departures and Long Distance Traffic Table 2 below shows the r values that were calculated for long distance traffic points with airline departures. The strongest r value was about .525 when the time was shifted one hour back for traffic data. After running a t-test, a t value of 4.186 was found for this strongest r value. Since this falls above the critical value of 2.680 this indicates there is a statistical significant correlation between long distance traffic and airline departures. The t value of 4.186 is not as strong as the t value of 11.031 that was found for the short distance points. This means that the short distance points have stronger and statistical more significant correlation with the number of airline departures. Also there was a correlation when the time was not shifted, shifted 2 and 3 hours back. These correlations were not as strong and significant as when the time was shifted 1 hour back. Table 2 Correlation -0.08452106 0.084719838 0.240464203 0.378802084 0.495252036 0.525245289 0.448276451 Time Shift Shifted 6 hours back Shifted 5 hours back Shifted 4 hours back Shifted 3 hours back Shifted 2 hours back Shifted 1 hour back not shifted t Value Critical Value -0.575 2.68 0.577 2.68 1.68 2.68 2.776 2.68 3.866 2.68 4.186 2.68 3.401 2.68 28 Airline Arrivals and Short Distance Traffic Table 3 below shows the r values that were calculated for short distance traffic points with airline arrivals. The strongest r value was about .650 (highlighted in red) when the time was not shifted for traffic data. The critical value is 2.64 for this data set. A t value of 8.287 was calculated. It can be concluded that there is a statistical significant correlation between short distance traffic points and airline arrivals because the t value is greater than the critical value. A correlation was found for when the time was shifted 1, 2, and 3 hours ahead. The correlations for these were not as strong as the correlation when the time was not shifted. Table 3 Correlation Time Shift t Value Critical Value 0.140618457 Shifted 4 hours ahead 1.377 2.64 0.300181239 Shifted 3 hours ahead 3.051 2.64 0.45151347 Shifted 2 hours ahead 4.906 2.64 0.564560077 Shifted 1 hour ahead 6.632 2.64 0.649737596 not shifted 8.287 2.64 29 Airline Arrivals and Long Distance Traffic Table 4 below shows the r values for long distance traffic points and airline arrivals. The greatest r value is about .513 (highlighted in red). A t value of 4.058 was calculated. Since the t value is above the critical level of 2.68 there is a correlation between the data. A t value of 4.058 is not as strong as the t value of 8.287 from the short distance traffic points. This means that the correlation between long distance traffic points and airline arrivals is weaker and less statistical significant than short distance traffic and airline arrivals. When the time was shifted 1 hour for the long distance points they showed an even weaker and less statistical significant correlation. Table 4 Correlation -0.032177503 0.083592201 0.173265718 0.224506934 0.33081925 0.443231066 0.51341309 Time Shift Shifted 6 hours ahead Shifted 5 hours ahead Shifted 4 hours ahead Shifted 3 hours ahead Shifted 2 hours ahead Shifted 1 hour ahead not shifted t Value -0.218 0.569 1.193 1.563 2.378 3.354 4.058 Critical Level 2.68 2.68 2.68 2.68 2.68 2.68 2.68 30 Conclusion Discussion The results suggest a correlation between road and airline traffic. The correlation depends on the distance the traffic point is from the airport and whether airlines are arriving or departing. For airline arrivals and short distance traffic, the strongest correlation with a t value of 8.287 was found when there was no time shift. Although there was still a correlation when times for road traffic were shifted one, two and three hours ahead, the strongest correlation was when there was no shift. Since no shift in time had the strongest correlation it is used in testing hypotheses number 1. For hypothesis #1, H1 was that road traffic level would increase after airline arrivals while H0 was that road traffic level would not increase. Since the most significant and strongest correlation between traffic and airline arrivals was positive, H0 is rejected and the alternative hypothesis (H1) is accepted. According to the results, the strongest correlation is found within the same hour (no time shift/lag). A likely explanation is that passengers are leaving the airport and passing by the short distance traffic count points within an hour. Since traffic levels are measured hourly it is possible that the passengers could be getting to the short distance points before the next hour of traffic is collected. Airline arrivals and long distance traffic showed the strongest correlation with a t value of 4.058 when there was no time shift. When the time was shifted one hour ahead there was a correlation, but not as significant and strong as no shift in time. For hypothesis #1, H0 is 31 rejected and H1 is accepted since the strongest correlation is positive. However, the strongest correlation is still found within the same hour (no time shift/lag). There might be a variety of reasons why this occurred. These points were located far from the airport and it would be highly unlikely that passengers passed by these points less than an hour after their plane arrived since these points are located about an hour or more from the airport. It is possible that the 40 mile point could be less than an hour away, but this would require passengers to leave from the airport almost immediately following their airline arrival. This is highly unlikely since most passengers have to wait for baggage and walk a certain distance through the airport. However, the results show the strongest correlation of traffic and airline arrivals at the long distance points is within the same hour that the planes arrive. This could be due to other outside influences since these points are located 40 and 79 miles from the airport. There are many other factors that can affect the road traffic count of these long distance points from the airport. The more space there is between a traffic count point and the airport the greater the possibilities that there will be other influences. The strongest correlation at no time shift for these points must have occurred from other outside influences between the traffic points and the airport. For hypothesis #3, H1 was that the further away the traffic point, the longer the time lag is between road and air traffic, while H0 was that the time lag would not increase with distance away from the airport. From the results H1 is rejected and H0 is accepted. There is not enough evidence to show that the further away the longer the time lag is between road traffic and the number of airline arrivals. Airline departures and short distance traffic showed the strongest correlation with a t value of 11.031 when the time was shifted back one hour. In addition, there was a correlation 32 when the time was not shifted back and shifted two, three, and four hours back. Since one hour showed the strongest correlation it was chosen to test the hypotheses. For hypothesis #2, H1 was that road traffic levels would increase before airline departures and H0 was that road traffic levels would not increase. According to the results, the traffic level showed the strongest and positive correlation one hour before the airlines departed. So H1 was accepted and H0 was rejected. The data shows a clear traffic increase an hour before airlines depart. This is most likely due to passengers trying to catch their flight on time. Obviously if a passenger arrives after a plane departs or at the time it departs they will miss their flight so these results make sense in the passenger perspective. The potential reason for a correlation with other shifts is passengers arriving earlier than an hour before their flight. The correlation decreases with time most likely because fewer passengers arrive more than an hour earlier. A correlation was also found for when the time was not shifted. This could be from passengers arriving within an hour of their flight departing, but the correlation is not as strong. Airline departures and long distance traffic showed the strongest correlation with a t value of 4.186 when the time for traffic was shifted back one hour. Correlations could still be seen for no shift in time, two, and three hours back in time. Since the strongest correlation is when the traffic time was shifted back one hour, hypothesis #2 H1 was accepted and H0 was rejected. There is possible explanation for this short time lag between the traffic levels at long distance points and airline departures. Since the time was shifted one hour back this compares traffic that occurs between one and two hours before the airlines depart. This would most likely give passengers enough time to travel from the long distance traffic points to the airport in time to catch their flight. It is important to note that the strongest correlation for airline 33 departures and long distance traffic has the same time shift (one hour back) as the airline departures and short distance traffic. The traffic levels for the long distance points might have a correlation with the airline departures closer to two hours before the airlines depart. However, if the correlation is slightly under two hours it would fall in the one hour shifted back category. This is a possible explanation why the strongest correlation for long and short points occurred when the time was shifted 1 hour back. From these results hypothesis #3 H 1 is rejected and H0 is accepted. The time lag did not appear to increase with distance from the airport, at least not at the temporal resolution of one hour. Without additional data at the distant traffic points between one and two hours before airline departures, there is not enough evidence to support H1. For the last hypothesis (#4), H1 is that the closer the traffic points are to the airport the stronger the correlation will be between road and air traffic; while H0 is that the correlation between road and air traffic will not decrease with distance from the airport. When airline arrival and departure times were compared with traffic levels, their respective correlations decreased with increasing distance. Road traffic levels also had more significant correlations with air traffic when using close traffic points over long distance traffic points. This supports H1 and therefore H1 is accepted and H0 is rejected. The closer the traffic points were to the airport the stronger correlation there was between road and air traffic. The probable reason for the correlation increase was because there was less outside influences other than airline traffic at the short distance traffic points than the long distance traffic points. 34 Outlook and Future Directions This study and the data it provides could be very helpful in helping highway and airport designers and planners. Highway designers might have a better understanding of how traffic is affected around airports and therefore they will have an easier time deciding the type and size of highways of the future. The study also provides a general idea of when people start to arrive for their flights and when they leave. By knowing when people are leaving and arriving at the airports it can be possible to ascertain how to avoid high traffic that is due to airline departures and arrivals. Knowing how to avoid high traffic periods could be very useful for many passengers and could possibly improve traffic flow since some might try to travel when it is less congested according to this study. Further, the methods used in this study could be used to predict traffic in near real time, and streamed on the Internet for trip planning. This service could potentially improve traffic surges as people attempt to avoid predictable peaks. There are a few different ways that this study can be improved in the future. The traffic data source that was used (Illinois Department of Transportation) was limited in the number of days and locations road traffic was recorded. If the number of days and locations of traffic counts could be increased, this would increase the accuracy of the study. Due to the limitations of the data it was impossible to obtain traffic data that was on the same day, or the same day of the week. The closer in time the traffic data is recorded the more accurate the study would be. Another way this study could be improved is if traffic data was recorded more often than every hour. The accuracy would be improved if traffic data could be broken down into half hour 35 blocks, 20 minute blocks, etc. Once more data becomes available in the future for GIS and transportation these improvements might become possible. 36 Acknowledgements I would like to thank my thesis advisor Dr. Wenjie Sun for her enormous help in guiding me through my thesis. Without the help of Dr. Sun, completing my thesis would have been very difficult. Secondly I would like to thank Dr. Kurt Piepenburg for his assistance in guiding me through my thesis. I would also like to thank my family for their constant support and assistance. Also I would like to thank the entire Geography and Earth Science department for their guidance and providing academic knowledge in the courses I have taken here at Carthage College. Appendix 37 38 39 40 41 References “Processing traffic data for statistical use; The Norwegian methods” by Roar Norvik. Norwegian Public Roads Administration. Department of Technology. Centre for Road and Traffic Technology, Trondheim. Using a GIS-Based Approach and Wind Rose to Determine Runway Effectiveness and Study the Impacts of O’Hare Chicago International Airport by Patrici A L. Danyelle Lewis “The Geography of Transportation Systems” by Jean-Paul Rodrigue, Claude Comtois and Brian Slack (2009), New York: Routledge. http://people.hofstra.edu/geotrans/eng/content.html “Fluctuations in airport arrival and departure traffic: A network analysis” by Li Shan-Mei, Xu Xiao-Hao, and Meng Ling-Hang (2012). School of Computer Science and Technology, Tianjin University, Tianjin 300072, China. College of Air Traffic Management, Civil Aviation University of China, Tianjin 300300, China. Research and Innovative Technologies Administration http://www.bts.gov/programs/airline_information/ Illinois Department of Transportation Traffic Count Database System (TCDS) http://www.ms2soft.com/tcds/tsearch.asp?loc=Idot&mo “The ESRI Guide to GIS Analysis Volume 2: Spatial Measures & Statistics” by Andy Mitchell (2005), ESRI. 42