National Academies Press: OpenBook

Toolkit for Establishing Airport Catchment Areas (2023)

Chapter: Tools for Catchment Area Analysis

« Previous: Preliminary Analysis
Page 21
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 21
Page 22
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 22
Page 23
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 23
Page 24
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 24
Page 25
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 25
Page 26
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 26
Page 27
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 27
Page 28
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 28
Page 29
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 29
Page 30
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 30
Page 31
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 31
Page 32
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 32
Page 33
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 33
Page 34
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 34
Page 35
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 35
Page 36
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 36
Page 37
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 37
Page 38
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 38
Page 39
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 39
Page 40
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 40
Page 41
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 41
Page 42
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 42
Page 43
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 43
Page 44
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 44
Page 45
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 45
Page 46
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 46
Page 47
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 47
Page 48
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 48
Page 49
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 49
Page 50
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 50
Page 51
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 51
Page 52
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 52
Page 53
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 53
Page 54
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 54
Page 55
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 55
Page 56
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 56
Page 57
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 57
Page 58
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 58
Page 59
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 59
Page 60
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 60
Page 61
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 61
Page 62
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 62
Page 63
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 63
Page 64
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 64
Page 65
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 65
Page 66
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 66
Page 67
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 67
Page 68
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 68
Page 69
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 69
Page 70
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 70
Page 71
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 71
Page 72
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 72
Page 73
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 73
Page 74
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 74
Page 75
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 75
Page 76
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 76
Page 77
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 77
Page 78
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 78
Page 79
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 79
Page 80
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 80
Page 81
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 81
Page 82
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 82
Page 83
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 83
Page 84
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 84
Page 85
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 85
Page 86
Suggested Citation:"Tools for Catchment Area Analysis." National Academies of Sciences, Engineering, and Medicine. 2023. Toolkit for Establishing Airport Catchment Areas. Washington, DC: The National Academies Press. doi: 10.17226/27424.
×
Page 86

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

21 Tools for Catchment Area Analysis In this toolkit, six analytical tools are provided to help users with different needs and resources to infer airport catchment areas using different approaches and data sources. Users are recommended to read the Overview of Analytical Tools section first to select tools that suit their specific requirements. Various examples based on actual airports are provided to help explain analysis procedures. Users are recommended to select more than one tool to cross-validate the analysis results. Overview of Analytical Tools This section provides a summary of the strengths and weaknesses of each tool. To help potential users determine the most appropriate tools that suit their purposes, we suggest a few scenarios to deploy these tools. Travel Utility Analysis • Strengths – Widely applicable to airports of any capacities, hub status, or resources – Does NOT require additional data collection – The analysis is route/destination/market specific – Applicable for both existing and proposed new air services • Weaknesses – Users need to input the value of travel time savings (VTTS) for the ground access component and flight component of the trip – Dependent on several key assumptions  Travelers make rational decisions  Travelers are informed and aware of alternative airport options • Recommended Application Scenarios – Potential market analysis for new air services – Sensitivity analysis of key parameters in shaping airport catchment area  Parking rates, airfares, flying time, VTTS, etc. Passenger Survey • Strengths – Relatively easy data collection if relevant questions  Can be embedded in routine air passenger surveys, or  Can be included on the splash page of Wi-Fi hotspots – Can be used to analyze catchment areas for both outbound travelers and inbound visitors – ZIP codes level data are easier to plot in Microsoft Excel

22 • Weaknesses – The quality of data can be questionable as there is no good mechanism to validate responses – Survey responses may be distorted by participation bias: Only data from passengers who are willing to participate in the survey will be collected – Due to privacy concerns, location data can typically only be collected at the ZIP codes level or higher • Recommended Application Scenarios – Quick, low-cost establishment of existing catchment area – Repeat passenger surveys periodically to gauge the change in passengers’ opinions and compositions Mobile Location Data Analysis • Strengths – Sufficient data counts for large airports – Historical and live data are available – Data available for subject airports and competing airports • Weaknesses – Data quality for small airports can be problematic – Potential selection bias: Only travelers with mobile devices can be recorded – Mobile location data do not provide sufficient information about trips and travelers – Acquisition costs can be high, depending on the providers and their pricing models – Not applicable for non-existent air services • Recommended Application Scenarios – Ad hoc catchment area analysis – Traffic leakage analysis Parked Vehicle Analysis • Strengths – Revealed and validated data: Outbound passengers park at airport parking facilities – Existing access to data sources: Vehicles are parked at airport premises – Data can be easily supplemented if airports use pre-booking systems or collect billing ZIP codes at payment kiosks • Weaknesses – Data only represent passengers who drive and park – Access to state registration records is required for detailed results if data are collected from license plates – Vehicle registration associated with a license plate may not represent passengers’ current residential address – Airports can only access vehicles parking at their own facilities, therefore this tool cannot be used for traffic leakage analysis

23 • Recommended Application Scenarios – Appropriate for airports that conduct routine (e.g., overnight) parking inventories that collect license plate data – Suitable for airports that can get access to state vehicle registration records or airports that collect billing addresses of credit cards used to pay for parking or parking reservations. Billing Data Analysis • Strengths – Excellent representativeness of the travel population and the sample rate can be relatively accurately estimated – Data are relatively consistent and are validated (actual purchase) – Data are available for both the subject airports and competing airports • Weaknesses – Certain passengers’ billing addresses may not match their current residential addresses – Results may be biased: Certain carriers can be over-represented or under-represented – Data acquisition cost is high • Recommended Application Scenarios – The airport is mainly served by airlines that are fairly represented in billing data – Analysis of traffic leakage to competing airports User-Generated Content (UGC) Analysis • Strengths – Data may be inexpensive to acquire (application dependent) – Data of any business entities, including competing airports, are available – Supports real-time data acquisition • Weaknesses – Certain SNS applications may limit free data acquisition – Users’ residential addresses may be difficult to infer – Participation bias: Only SNS users contribute their date • Recommended Application Scenarios – Customer feedback monitoring – Service disruption alert

24 Travel Utility Analysis This tool analyzes, from travelers’ perspectives, the itinerary (a combination of airline choice and airport choice) that can deliver the highest utility (i.e., the lowest total cost). The tool accounts for both the economic expense and value of time into this analysis. This tool provides a “base case optimal” catchment area that could be realized if travelers (a) chose their airport based on overall travel time and cost and (b) were fully aware of the travel time and cost associated with each of their potential travel options. The tool is presented in a web-based interactive form so that users can test different scenarios and perform sensitivity analysis to understand the effect of input parameters on the boundary of airport catchment areas. Below are links to demonstrations of the Travel Utility Analysis tool using CAK, ONT, and OAJ airports. Case 1 – Akron-Canton Airport (CAK), OH (https://airtransport.shinyapps.io/CAK_catchment/) Case 2 – Ontario International Airport (ONT), CA (https://airtransport.shinyapps.io/ONT_Catchment/) Case 3 – Albert J. Ellis Airport (OAJ), NC (https://airtransport.shinyapps.io/OAJ_Catchment/) Usage This tool is intended to:  Test the sensitivity of the input parameters on the coverage of airport catchment areas,  Explore the potential market of proposed new air services, and  Assess the competitiveness of existing air services. Data Requirements This tool uses the following data:  Driving distance and driving time from each geographic entity to the subject airport and nearby competing airports. Such data are the results from Spatial Analysis.  The Internal Revenue Service (IRS)-issued standard mileage rates used to calculate the deductible costs of operating an automobile for business. During January 1 – June 30, 2022, this rate was 58.5 cents per mile.  Daily rates for comparable parking options at the subject airport and nearby competing airports, for instance, economy parking or close-in parking.  Airfares and total travel time (including layovers) from the subject airport and nearby competing airports to various destinations of interest. Such destinations can be historical, current, or proposed. Software Requirements The demos included in this tool kit are built using R, a programming language for statistical computing and graphics. This tool can also be built with other programming languages, such as Python or JavaScript. Once built, the tool can be accessed using any web browser.

25 Theoretical Background The theory behind the Travel Utility Analysis tool is based on Gao (2020a). In the study focusing on the catchment areas of Indianapolis International Airport (IND), Gao (2020) proposes that the overall travel utility comprises ground access cost, parking cost, and flying cost for travelers who choose to drive to access commercial flights. Therefore, a comparison can be made among different itineraries involving different flights and origin airports to determine which itinerary delivers the highest travel utility or, alternatively, the maximum cost savings. Once the optimal itinerary is determined for each geographic entity (county, ZIP code, or census tract), the user can reversely identify if a particular geographic entity belongs to the catchment area of the subject airport or not. The travel utility of an itinerary is measured by the monetary expense and time expense. The monetary expense is simply the sum of various expense items, including mileage-based driving experience, parking expenses, and airfares. The time value is converted from total travel time, using the appropriate value of travel time savings (VTTS). Gao (2020) uses the following rates to calculate time expenses for the ground access segment and flight segment (See Table 4). Table 4.The 2018 Value of Travel Time Savings (VTTS) ($/hour) Mode of Travel Purpose of Travel Personal Business Intercity (Ground access) $20.65 $27.64 Air & high-speed rail (flight) $39.24 $63.20 Source: Gao (2020a) User Interface The web-based Travel Utility Analysis tool allows users to change the values of input data and view the corresponding catchment area map update instantaneously. The input section is organized into five tabs, titled Info, Trip, Airfare, Flying Time, and Parking. See Figure 6 – Figure 10 for details of each tab of the Travel Utility Analysis tool designed for CAK. The Info tab shows how the travel utility is calculated. For this tool, we consider the overall travel utility is composed of three components, which are ground access cost, flight cost, and parking cost. See Figure 6 for the screenshot of the Info Tab for CAK airport.

26 Figure 6. The Info tab of the Travel Utility Analysis tool for Akron-Canton Airport (CAK) The Trip tab collects basic information of travelers, including total duration of the trip, the party/group size, VTTS (ground), VTTS (air), IRS standard mileage rate for business (IRS, 2022), one-way or round- trip, and how the group values their time values. User of this tool can adjust each input variable to reflect the market situation they are facing. See Figure 7 for more information. Figure 7. The Trip tab of the Travel Utility Analysis tool for Akron-Canton Airport (CAK)

27 In the Airfares tab, users are asked to provide airfares for flight itineraries departing from different airports. This is to simulate the circumstance of analyzing destination-specific catchment areas. Users can also use average airfares of a period to simulate more generic situations. If no flight options are available from a certain airport, users are advised to leave the airfare of that airport as $9,999. This arbitrary expensive airfare is to deliberately generate travel options that are infeasible and uncompetitive so that residents are forced to consider other airports. See Figure 8 for more information. Figure 8. The Airfares tab of the Travel Utility Analysis tool for Akron-Canton Airport (CAK) The Flying Time tab collects the total flying time associated with flight itineraries departing from different airports. The flying time includes not only the block time between departure and arrive, but also connecting time between flights. This is to simulate the effect of itinerary durations on travelers’ choices. If no flight options are available from a certain airport, users are advised to leave the flying time of that airport as 99. This arbitrary long flying time is to deliberately generate travel options that are infeasible and uncompetitive so that residents are forced to consider other airports. See Figure 9 for more information. Figure 9. The Flying Time tab of the Travel Utility Analysis tool for Akron-Canton Airport (CAK)

28 The last input tab, the Parking tab, collects data about daily parking rates at airport parking facilities. To simplify the analysis, this tool only asks for the daily rate for the economy parking option. Similarly, if no flight options are available from a certain airport, users are advised to leave the parking rate of that airport as $99. This arbitrary expensive parking rate is to deliberately generate travel options that are infeasible and uncompetitive so that residents are forced to consider other airports. See Figure 10 for more information. Figure 10. The Parking tab of the Travel Utility Analysis tool for Akron-Canton Airport (CAK) Output Based on the input condition, the Travel Utility Analysis tool will display a live map reflecting the potential catchment area of the subject airport, which includes all the geographic units where the subject airport has an utility advantage over other competing airports. If needed, this tool can also display additional data about the airport catchment area, such as the number of geographic units and total population included in the catchment area. See Figure 11 for a sample catchment map of CAK generated by the Travel Utility Analysis tool.

29 Figure 11. Sample catchment area map of Akron-Canton Airport generated by the Travel Utility Analysis tool Recommendations The Travel Utility Analysis tool requires users’ inputs for airfares, flying time, and parking rates. Therefore, it will be more appropriate for airports to analyze destination–, or route–, specific catchment areas. This will enable airports to estimation their potential market for existing or proposed air services. If airports wish to produce “neutral” catchment areas that do not depend on specific destinations or routes but only rely on spatial relationships between airports and all potential geographic units, then users are recommend putting in identical airfares, flying time, and parking rates for both the subject airport and competing airports.

30 Passenger Surveys Airports conduct passenger surveys often to gauge users’ opinions in regard to their experience at airports. The Passenger Survey tool establishes an airport’s catchment area using information (a) collected as part of surveys of airport passengers or (b) voluntarily provided by passengers as part of other airport functions. In general, the tool uses zip code data provided by passengers to create a map of responses that practitioners can use to establish an estimated catchment area for a given airport. The purpose of the tool is to quickly identify an airport’s “primary catchment area,” which is the area in which the airport will typically focus its analytics and marketing resources to generate additional traffic. An airport may attract traffic from outside of this primary catchment area, but such traffic volumes are typically very low and investments in growing traffic in these areas would likely have very limited returns. Usage The goal of this tool is to establish the catchment area using low-cost data collection methods targeting passenger origins within the region or using information already provided by passengers as part of other airport interactions. Requirements Use of this tool requires: • A data collection source (such as a prior or upcoming passenger survey) that includes responses to questions about passenger origin or destination within the region, such as zip code. • Survey results available in an electronic format readable by mapping software. • Mapping software that can display the results in a visual format, including the absolute or relative number of responses provided for each location in the region. Methodology Utilizing passenger survey data to establish airport catchment areas normally encompasses four steps, which are as follows: • Data collection • Data review and analysis • Results mapping, and • Catchment area determination Data Collection Data collection for this tool relies on data collected as part of other airport survey efforts (such as a passenger hold room survey), data collected as part of other airport commercial activities, or data collected specifically to establish passenger origins or destinations within a region.

31 Passenger Hold Room Surveys Airports routinely conduct passenger hold room surveys to collect information on a wide range of topics, including passenger demographics, access mode choice, and use of concessions or other facilities. These surveys are typically conducted by interviewers who record the responses or use tablets where interviewees input their responses. Similar surveys are also used to gauge customer satisfaction with the airport and with specific aspects of the airport (ACI World’s Airport Service Quality, or ASQ, program is an example of such a survey). In these surveys, questions may address the passenger’s point of origin (for local residents) or destination (for visitors) within the region, often by asking the passenger to provide the appropriate zip code. This information can then be used to determine the primary geographic area from with the airport is drawing its passengers. As this information is provided voluntarily by passengers, it is subject to errors including, but not limited to: • Unfamiliarity with the zip code of their origin or destination • Unwillingness to provide the information (leading to empty or willfully incorrect entries) • Data entry errors by the interviewer or the passenger (depending on the survey method used) In addition, the phrasing of the question may lead to inaccurate results (as described below in “Results Mapping”). Such surveys typically cost several thousand dollars (for the conduct of the survey and subsequent analysis) to collect several thousand usable responses. However, the incremental cost of including questions related to passenger origin/destination within the region is minimal. Wi-Fi Surveys If an airport does not have recent survey data or the prior survey did not include questions on passenger origin or destination within the region, the airport’s Wi-Fi system can be used as a vehicle to conduct a Example: 2019 Passenger Survey, Los Angeles International Airport. Los Angeles World Airports (LAWA) conducts periodic surveys of passengers flying out of Los Angeles International Airport (LAX). Most recently, these surveys were conducted in 2015 and 2019. The survey included approximately 90 questions (though not all respondents would need to answer each question as some were dependent on responses to prior questions) of which several were related to their origin and destination within the region. Such questions included: • What is the ZIP code of the place you came from prior to arriving at LA Airport today? • What part of Southern California did you depart from prior to coming to LA Airport today? • What local attraction(s) [did you visit]? • Do you currently live in the Southern California area? • What is your home zip code? Based on the questions asked, this survey provided over 8,000 data points that could be used to estimate the airport’s catchment area and was used as the basis for the description of the tool, as provided later in this section.

32 brief, focused survey. As described in ACRP Research Report 235: Guidebook for Conducting Airport User Surveys and Other Customer Research (Franz et al., 2021), such surveys can be conducted as part of the process of connecting to the airport’s public Wi-Fi system. During the connection process, the customer can be presented with the survey questions (e.g., “Please answer the following questions to connect…”), after which they are provided access. Typically, such surveys are very short (i.e., three or fewer questions) as longer surveys tend to have lower response rates and can create frustration. As this method relies on information provided voluntarily by passengers, it is subject to the same potential errors as is a passenger hold room survey. Such surveys typically cost less than $1,000 and can collect several thousand usable responses depending on the length of time the survey is in use. Other Data Sources Airports may have access to other data sources that provide passenger home ZIP codes. Such sources include: • Frequent parker programs. As part of such programs, customers are asked for their home addresses. • Parking reservations systems. As customers reserve parking with the airport and pay for the reservation, they provide their zip code as part of the credit card verification process. Further discussion of using parking reservations data is provided as part of the Parked Vehicle Analysis tool. • Other loyalty programs. Programs such as Thanksagain! allow passengers to earn rewards points as they spend money at participating airports for terminal concessions. Participating airports may be able to obtain the passenger address data for passengers using the program at their airports. Data Review and Analysis As described above, collected data should focus on address information, such as zip codes (5 or 9 digits) provided by residents or some other geographic identifier, such as a known destination within a region (e.g., Disneyland), provided by visitors. Based on capabilities of commonly available mapping software, such as Example: 2019 Mode Share Survey, Gerald R. Ford International Airport (Grand Rapids, Michigan) As part of a ground transportation study, Gerald R. Ford International Airport conducted a brief survey using login screens for the airport’s public Wi-Fi system. While the primary purpose of the survey was to understand passenger access mode share, the survey asked for each customer’s home zip code. Questions included: • How did you travel to Grand Rapids Airport today? • Please provide your home zip code. • If you would like to be contacted regarding special offers, news, and events at the Airport, please provide your email address. While the home zip code question resulted in responses from around the world, the responses could be filtered to identify passengers from the airport’s region versus visitors. Over seven days, the survey resulted in over 7,000 usable responses.

33 the 3D Maps tool provided as part of Microsoft Excel, zip codes provide a reliable source of data for displaying passenger origins within a region. If data is provided in other forms, such as named destinations, such destinations could be converted to zip codes for purposes of mapping. While there are several software tools available to convert address data to maps, the following steps assume the use of Microsoft Excel as it is a commonly used program within the industry. Using address or location information, apply the following steps to the data. 1. Import or copy collected information into a table, 2. As necessary, convert information, such as identified regional destinations, to zip codes, 3. Identify data that contain a number of characters that is not 5, 6, or 9, 4. If your catchment area potentially includes zip codes beginning with a zero (e.g., Connecticut, Massachusetts, Maine, New Jersey, Rhode Island, Vermont), Canadian addresses, or Mexican addresses, additional considerations are provided below, and 5. Remove data that are not 5 digits, 6 characters (for Canadian postal codes), or 9 digits, as these likely represent incorrect information provided by customers or international locations that cannot be part of a U.S. airport’s catchment area. Considerations for catchment areas potentially include zip codes beginning with a zero, Canadian addresses, or Mexican addresses. • U.S. zip codes beginning with a zero. Excel will likely display such zip codes as 4-digit numbers (e.g., 09999 will display as 9999). Using a standard Excel number format, “zip code,” provides the leading zeros. • Canadian postal codes. Such codes will likely display as two three-character groups (e.g., A9A 9A9) or one six-character group (A9A9A9). These codes can be retained as-is if your mapping software can properly place them (as does the 3D Map in Excel). Otherwise, their locations may need to be identified using other means. • Mexican postal codes. Mexican postal codes are 5 digits, as are U.S. zip codes. If your catchment area potentially includes addresses within Mexican states close to the border (Baja California, Sonora, Chihuahua, Coahuila, Nuevo Leon, and Tamaulipas), Mexican postal codes may appear in the data set. These codes can be retained as-is if your mapping software can properly place them (as of June 2022, the 3D Map in Excel does not). Otherwise, their locations may need to be identified using other means.

34 Results Mapping To display the results, import the resulting data into a mapping program, such as 3D Maps in Microsoft Excel, which can display the locations of the data provided as well as the quantity (or relative quantity) of responses from a given location. Figure 12 depicts the zip codes provided by airline passengers as part of the 2019 Passenger Survey, Los Angeles International Airport (Unison Consulting 2019). Zip codes were provided in response to the question, “What is the ZIP code of the place you came from prior to arriving at LA Airport today?”. As shown, within the total number of responses (over 8,000), passengers indicated their originating points were located across a large area, including locations requiring a drive of over 6 hours and locations near other large-hub airports (e.g., San Diego and Las Vegas – McCarran International airports). Furthermore, most of the zip codes identified in the survey had very few responses. Figure 12. Map Display of Zip Code Responses, 2019 LAX Passenger Survey

35 Figure 13 shows the same data but focuses on Southern California. Aside from three locations with very high numbers of responses (e.g., the area immediately adjacent to LAX, downtown Los Angeles, and Disneyland), responses appear to have been spread relatively evenly throughout a certain region. Filtering responses to remove zip codes with low-response numbers (due to distance and/or a low population base) can focus the map results to better identify an airport’s primary catchment area. Figure 14 and Figure 15 depict the same survey data, but only show zip codes with responses exceeding 0.1% of the total sample size (i.e., more than 8 responses for a single zip code). As shown, the exclusion of low-response zip codes focuses the results on a much tighter area that limits more- distant results to areas with a high population (e.g., Santa Barbara, Bakersfield, and Palm Springs). Figure 13. Map Display of Zip Code Responses, Southern California Area, 2019 LAX Passenger Survey The high number of responses for the area immediately surrounding the airport highlights the importance of question phrasing in surveys. Some passengers with early morning flights may choose to drive to the airport the night before and stay at a nearby hotel. The question phrasing (“What is the ZIP code of the place you came from prior to arriving at LA Airport today?”) could lead some of those passengers to indicate their hotel as opposed to their original origin within the region.

36 Figure 14. Map Display of Zip Code Responses Comprising >0.1% of All Responses, 2019 LAX Passenger Survey

37 Figure 15. Map Display of Zip Code Responses Comprising >0.1% of All Responses, Southern California Area 2019 LAX Passenger Survey Catchment Area Determination Based on the map shown in Figure 15, professional judgment can be used to estimate the airport’s primary catchment area, as shown in Figure 16. In this example, the determination of the primary catchment area reflects an understanding of general population density within the region, population characteristics, and availability of service at other airports. For example, while zip codes near Palm Springs and Bakersfield may have similar numbers of responses, the Palm Springs area (a) has a lower population, (b) is a leisure destination, and (c) has substantially more air service at a nearby airport (Palm Springs International Airport served approximately 1.3 million enplanements in 2019 while Bakersfield’s Meadows Field served approximately 130,000). Therefore, it could be concluded that Bakersfield residents are more likely to consider LAX as a primary option (along with other airports, such as those in Burbank, Ontario, Orange County, and Fresno) than would residents of Palm Springs.

38 Figure 16. Potential catchment area identified using passenger survey data Considerations  Cost. This tool relies predominately on data already collected as part of other efforts. If such data are not readily available, they can be collected through efficient, low-cost methods such as a Wi-Fi survey. Therefore, the cost required to use this tool is likely limited to the time required to analyze and map the collected data.  Precision. As this tool relies on zip codes, the resulting catchment area boundary will typically reflect the boundaries of zip codes.  Potential errors. As described above, this tool is subject to errors (both intentional and unintentional) associated with the collection of information voluntarily provided by passengers.  Other limitations. This tool uses information available only regarding the subject airport. As such, while it can help determine the airport’s catchment area, it cannot determine the extent to which passengers within that area choose other airports. Customer Survey at Akron-Canton Airport (CAK) During the development of this toolkit, a voluntary customer survey has been setup at Akron-Canton Airport from mid-June to mid-July to collect the passengers’ origins when passengers tried to access Wi- Fi hotspots at the airport. A total of 91 responses have been collected during four weeks of data collection, of which 55 respondents were from Ohio and 36 were from other states. We plot the ZIP codes of local and nearby passengers in Figure 17. Please be noted this figure is only for reference. Due to the limited number of responses, we do not have sufficient data to plot an insightful catchment area map for CAK.

39 Figure 17. Origins of passengers who responded to the customer survey at CAK during June – July 2022 Sources for Additional Information Users who plan to conduct airport customer surveys can refer to the following ACRP reports for additional information:  ACRP Report 26: Guidebook for Conducting Airport User Surveys, by Biggs et al. (2009)  ACRP Research Report 235: Guidebook for Conducting Airport User Surveys and Other Customer Research, by Franz et al. (2021)

40 Mobile Location Data Analysis Mobile location data, also known as places data in certain contexts, refer to the data that are derived from applications installed on smartphones and other mobile devices and can provide information with regard to users’ locations. With smartphones increasingly becoming daily essentials for many people today, data collected from such devices can provide more insights than conventional data sources. The Mobile Location Data Analysis tool is designed to infer airport catchment areas through location data shared by smartphone users and supplied by dedicated data providers. This tool makes use of mobile data collected and shared by commercial data vendors to establish the spatial distribution of travelers. A detailed procedure is provided to document key steps that are necessary to select, clean, analyze and visualize data. Usage Potential users can adopt this tool to achieve the following objectives: • Establish airport catchment areas by mapping the origins of airport visitors, • Analyze monthly and seasonal traffic variations (subject to consistent sampling requirements), and • Estimate traffic leakage from the airport catchment area to competing airports. Data Requirements This tool uses the following data: • Mobile location data, acquired from selected data providers. Refer to Data Selection/Acquisition for more information. • Driving distance and driving time from each geographic entity included to the subject airport and nearby competing airports. Such data are the results of Spatial Analysis. • Geometry information of relevant geographic entities. Such information is usually stored in the format of shapefile and can be accessed from federal and state agencies’ websites. • Monthly airport enplanement data, available from the Bureau of Transportation Statistics. Software Requirements Mobile location data, acquired from dedicated data providers, are usually stored in the form of comma- separated values (CSV) or JavaScript Object Notation (JSON) files. Such data are more efficient to be processed using software such as Tableau, Python, or RStudio. Our demonstration of the Mobile Location Data Analysis tool uses Python to process raw location data and RStudio to map the airport catchment areas. Users of this toolkit can choose other programming languages to process and map mobile location data. Analytical Procedure The procedure of mapping airport catchment areas using mobile location data is composed of four major steps, which are Data Selection, Data Processing, Data Assessment, and Visualization (See Figure 18). In this tool, we use Akron-Canton Airport, OH (CAK) to demonstrate this process.

41 Figure 18. Analytical procedure of the Mobile Location Data Analysis tool Data Selection/Acquisition With the data collected from mobile devices gaining increasing popularity, multiple providers are offering location data and analytics. For instance, as of Summer 2022, SafeGraph, Near, Foursquare, Placer.AI, Tamoco, Unacast, Veraset, and Gravy Analytics provide mobile location data. Users of the Mobile Location Data Analysis tool should contact multiple data providers to find an optimal solution that meets their individual requirements. When contacting mobile location data providers, users are recommended to inquire the following issues:  Available time frame: The collection of mobile location data started only a few years ago. Some providers may not have historical data meeting users’ specific requirements.  Sampling/Reporting frequency: Some data providers only report mobile location data in an aggregated approach, for instance, monthly or quarterly; Some providers may report data daily, or even hourly.  Geographic entities: Mobile location data may be available at different geographic entity levels, including counties, ZIP codes, census tracts, and census blocks (refer to Geographic Entities for more information).  Pricing model: Data providers may price their projects based on queries or subscriptions. Some providers may also provide other value-add analytic services. In this toolkit, we use data provided by SafeGraph for demonstration. Data Selection •Identify potential mobile location data providers •Contact data providers for data features and pricing •Select data/data providers Processing •Specify Point of Interest (POI) and time frame for data acquisition •Determine appropriate geographic entity level •Obtain geometry information of relevant geographic entities Assessment •Access airport enplanement data from the Bureau of Transportation Statsitics •Calculate montly sampling rate for the specified time frame Visualizatoin •Produce airport catchment areas by mapping airport visitors' origins •Analyze traffic leakage to nearby competing airports

42 Data Processing To enable catchment area analysis, raw data from SafeGraph need to be processed using the following procedures:  Understand data structure – In this tool, we use Monthly Patterns data from SafeGraph. Raw data are stored in .csv format. Instructions provided by SafeGraph (2022) provide information on the structure and variables of the SafeGraph Monthly Patterns data. – Our demonstration uses data from July 2020 – June 2021, which were provided in separate .csv files with each file representing information of a month. Within each .csv file, data are organized by geographic locations. Each line in the Monthly Patterns data represents a geographic location, or a Point of Interest (POI). Users need to use a combination of “placekey,” “parent_placekey,” and “location_name” information to locate lines that represent airports of interest. Refer to SafeGraph instructions for details (SafeGraph, 2022).  Extract visitor count information from Monthly Patterns data – For airport catchment area analysis, the most valuable data are travelers’ origins and the count of each origin. Such data are stored in the column “vistor_home_cbgs” in the Monthly Patterns data. The format of data is JSON, and the reporting geographic entity is census block group (refer to Geographic Entities for more information). – See Figure 19 for notes provided for “vistor_home_cbgs” by SafeGraph. Numbers followed census blocks represent the number of visitors from those blocks. For instance, 603 residents in census block 360610112021 visited the point of interest (in this case, the airport) during a month. – We use the Python package “JSON” to decode raw JSON data and convert “vistor_home_cbgs” records into tabular data frames for further analysis. In the converted data frames, each line now represents the record of a census block group. Figure 19. Notes of “visitor_home_cbgs” in SafeGraph Monthly Patterns data  Aggregate monthly records into annual totals – Depending on the number of visitors each month, certain airports may not have sufficient records to generate a meaningful catchment plot using only mobile location data for a month. And also due to inconsistent sampling rates of the monthly data, users of the mobile location data should consider aggregating monthly reported data into annual totals so that the analysis results can be more representative.

43 – In the case of Akron-Canton Airport (CAK) and its nearby competing airports (CLE, CMH, and PIT), we aggregate visitors’ counts in “visitor_home_cbgs” of the Monthly Patterns data to generate the total visitors’ counts during July 2020 – June 2021 for each reporting geographic entity (census block group in this case). • Aggerate census-block level data into selected geographic entities – Depending on the selected geographic entities, mobile location data reported at the census-block- group level may need to be further aggregated. In our demonstration using Akron-Canton Airport, OH (CAK), we select census tracts for data analysis and plotting. – In SafeGraph Monthly Patterns data, or essentially any data that contain geographic entities, Geographic Identifiers (GEOIDs) are used to denote different geographic entities. Federal Information Processing Series (FIPS) codes are used by data/tools of this toolkit as GEOIDs. Depending on the level of geography, FIPS may have different digits. For instance, at the state level, FIPS codes have two digits. And US county-level FIPS (with state information) have five digits. At the census-block- group level, which is used by the raw SafeGraph data, FIPS codes have 12 digits. – Based on the FIPS codes of census block groups, visitors’ count numbers can be aggregated to the census tract level by adding up all the census block groups in a census tract. Once this step of aggregation is completed, the format of visitors’ count data includes two key components, census tracts denoted by FIPS and a corresponding number of visitors from each census tract to an airport of interest. Assessing data sampling rate As available information with regard to the quality of mobile location data is still limited when this toolkit is drafted (mid-2022), users are recommended to assess the sampling rate before conducting formal analysis so that analysis results based on mobile location data can be interpreted with appropriate perspectives. To calculate the sampling rate of mobile location data for Akron-Canton Airport (CAK) and its nearby competing airports (CLE – Cleveland Hopkins International Airport, CMH – John Glenn Columbus International Airport, and PIT – Pittsburgh International Airport), we • Obtain Monthly Patterns data of four airports (CAK, CLE, CMH, & PIT) for the period during July 2020 – June 2021, • Aggregate monthly visitor counts of all included census block groups for each airport, and • Use monthly enplanement data of all US and foreign carriers of each airport, provided by the Bureau of Transportation Statistics (2022b), as denominators and mobile data count sum as numerators to calculate the sampling rate of each airport for each month. See Figure 20 for the sampling rates of SafeGraph Mobile Location data for CAK and its competing airports. It needs to be noted that the collection of mobile data inevitably captures the visits of both customers and employees. Depending on the data providers, some mobile datasets may have already filtered out data counts of employees using certain algorithms, while other datasets may require users to set their own filtering criteria to exclude employees’ records. For instance, if a mobile device visits an airport more than ten times (or other number that is deemed appropriate by the airport) a month, this device will be considered as being likely associated with an employee working at the airport.

44 Figure 20. Sampling rates of SafeGraph Mobile Location data for CAK and its competing airports (July 2020 – June 2021) As seen in Figure 20, sampling rates of mobile location data for four airports of interest in the case of catchment area analysis of CAK are not all consistent in July 2020 – June 2021. While the sample rate for Cleveland Hopkins International Airport is satisfactory, three other airports reported two months of missing records. This would prevent us from performing variation analysis month-to-month or generating catchment area plots using monthly data. Users of mobile location data could consider requesting sampling rates from data providers when they are assessing different data options. If data of inconsistent sample rates are to be used, such data should be aggregated to generate annual total visitors counts for subsequent catchment area analysis. Results Visualization With the visitors’ origins and counts data from SafeGraph Monthly Patterns data aggregated at the census tracts and annual level, we then plot the data on maps to visualize catchment areas of the subject airport and its competing airports. See Figure 21 for the map of the catchment area of Akron-Canton Airport, OH (CAK) based on the mobile location data provided by SafeGraph during July 2020 – June 2021.

45 Figure 21. Catchment areas of Akron-Canton Airport (CAK) based on SafeGraph Mobile Location data, July 2020 – June 2021

46 Parked Vehicle Analysis Driving and parking is a popular way for commuting between home and airports in the United States, especially for the vast majority of areas where public transit services to airports are very limited and inconvenient. Considering all vehicles legally driven on public roads are registered, geographic information linked to vehicle registrations or payment information used to reserve and pay for parking at airport parking facilities could be a potential source of information to determine the origin of vehicles/drivers, thus helping infer airport catchment areas. Usage The Parked Vehicle Analysis tool is designed to extract geographic information from vehicles parked at the parking facilities operated by airports. Depending on the state legislation with regard to vehicle registration and how airports collect registration and payment information of parked vehicles, this tool can be used to: • Determine the state origins of the parked vehicles, if the airport is pulling passengers from more than one state, • Find out the county with which the parked vehicle is registered, if the license plates of the registered state provide such information, and • Use the billing address ZIP codes of the payment methods for parking reservations to infer the catchment areas of airports. Meanwhile, it needs to be pointed out that the analysis results based on parked vehicles should be viewed with caution. Vehicle registrations and billing addresses of credit cards used for parking reservation may not match travelers’ current residential addresses. In addition, if the data collection process does not or cannot differentiate rental cars from travelers’ vehicles, that might distort the analysis results. Data Requirements Unlike other analytical tools included in this toolkit that can be universally adopted by any airport if budgets allow, the Parked Vehicle Analysis tool may only be applicable to certain airports due to its data requirements. The following requirements should be met by the Parking Vehicle Analysis tool to achieve the best analytical results: • License plate numbers, • County names or codes (if displayed on license plates), • State vehicle registration records (access required), and • Billing addresses of payment methods used for parking reservations if the airport is operating a parking reservation system. System Requirements The Parked Vehicle Analysis tool primarily uses two information sources, which are license plates and billing information provided by airport visitors. Prior to this tool, users should follow the decision tree in Figure 22 to determine if this tool is appropriate for their intended use.

47 Figure 22. Decision Tree for the Parked Vehicle Analysis tool License Plate Analysis Scenarios The License Plate Analysis extracts geographic information from license plates of vehicles parked at the airport parking facilities. Users are recommended to refer to the decision tree in Figure 22 to determine which of the following scenarios fits their circumstances: • Scenario 1 – The airport has access to vehicle registration records from state governments. If the airport can access vehicle registration records from state governments, then the airport can request to use license plate numbers to match such registration records to retrieve the home ZIP codes of vehicle owners. Using the assumption that most vehicles’ registration addresses are current, a map of vehicles’ origins at the ZIP code level can be plotted using one of the recommended visualization tools. Airports can use their own parking management system to capture the license plate numbers when vehicles enter and exit parking facilities or use the License Plate Recognition Tool developed for this toolkit to scan and recognize license plates.

48 • Scenario 2 – The airport does NOT have access to vehicle registration records from state governments. – Most of the vehicles in the airport parking facilities are from states that display county information on their license states. In this scenario, airports are unable to locate the origin of parked vehicles down to the ZIP code level. However, with the county information displayed on license plates, airports can still trace the origin of counties for parked vehicles. Such information is not suitable to produce detailed catchment area analysis results for airports, but it can be used to cross-validate results from other analytical tools. If parking management systems currently used by airports are not capable of capturing county information from license plates, airports are recommended to either adopt the License Plate Recognition Tool developed for this toolkit or manually sample vehicles from their parking facilities. • Scenario 3 – The airport does NOT have access to vehicle registration records from state governments. – Most of the vehicles in the airport parking facilities are NOT from states that display county information on their license states. – The airport primarily attracts passengers from more than one state. In this scenario, airports have limited information from license plates to conduct catchment area analysis. They can only rely on the registration state of vehicles to estimate the origin of states for passengers who drive to access commercial aviation. This information can be used to cross-validate results from other analytical tools. Airports can use their own parking management system, manually sample vehicles, or adopt the License Plate Recognition Tool developed for this toolkit to capture state the origins of vehicles. • Scenario 4 – The airport operates a parking booking/reservation system, or – The airport collects travelers’ billing ZIP codes when they pay for parking If airports are collecting travelers’ billing ZIP codes through parking booking or payment systems, such ZIP codes can be used to infer their origins based on the assumption that the majority of travelers have the same ZIP codes for billing and residential addresses and their information is accurate. • Other Scenarios – If an airport does not fit into any of the above scenarios This suggests that the airport does not have necessary information or does not meet the requirements of using the License Plate Analysis tool. The airport should consider other tools for catchment area analysis. License Plate Recognition Tool The License Plate recognition tool is a solution using videos captured through mobile cameras as input and exploiting deep-learning (DL) and computer vision techniques to retrieve geographic information down to the county level. This tool is developed primarily for airports without existing capabilities of capturing license plate information from their parking management systems.

49 The deployment of the License Plate Recognition tool involves multiple steps. See Figure 23 for an illustration of the deployment pipeline. Source: Purdue University Figure 23. Overview of the deployment pipeline of the License Plate Recognition Tool  Step 1: Video capture Action cameras mounted to a mobile platform can be used to capture videos of vehicles parked at the airport parking facilities. To ensure the quality of videos for subsequent processing, we have the following recommendations: – Use HD video cameras that are capable of recording videos at a minimum resolution of 1080p (1920x1080 displayed pixels) and 30 frames per second. – Avoid handholding. Mount the camera to a platform at a fixed height and move at a constant speed while shooting videos. – Capture videos when the lighting condition is good. Considering use external illumination for parking garages with poor lighting conditions. – Keep the length of each video clip under 2 minutes to enable fast data transferring and processing.

50 • Step 2: Image Extraction This step extracts images containing license plates from raw videos captured in Step 1. To do so, we first convert raw videos into a list of frames using the “VideoCapture” package in OpenCV (2022). We then adopt the open-source automatic license plate recognition library (OpenALPR) (Rekor Systems, 2022) to detect regions of the image with license plates and then crop images with plates. The cropped license plate images are subsequently used as data samples in the testing dataset. • Step 3.1: State Classification We develop a deep convolutional neural networks (DCNN) algorithm for state classification. In the initial development stage, we consider multiple DCNN structures proposed in the literature that showed high- performance results for natural image classification tasks. These DCNN structures include ResNet, AlexNet, VGG, SqueezNet, DenseNet, and Inception. Among these DCNN structures, ResNet outperforms other structures for our specific state classification task (V. Feng, 2017). Because of this, we determined to leverage the RestNet-50 DCNN structure for this task. The RestNet-50 DCNN model is trained using Python Pytorch and Torch-Vision software libraries. Torch-Vision allows loading pre-constructed RestNet-50 model structure. Specifically, RestNet-50 model is first trained using the training dataset for the classification of different states. For this task, additional linear and Softmax output layers are added for the initial pre-constructed ResNet-50 model structure. In the training procedure, we adopt the Adam optimization method to minimize the cross-entropy loss between the categorical 52 prediction vector and ground-truth labels formulated while constructing the dataset. The training is carried out till the loss converges. To further improve the performance of the ResNet-50 DCNN model, we finetune the trained model by leveraging an additional training dataset extracted from observation of actual vehicles in the target airport and considered to have more correlated data distribution with the intended testing dataset. Additionally, we also integrate the prior knowledge about the target airport to improve the performance. We deployed our License Plate Recognition tool at Louisville Muhammad Ali International Airport, KY (SDF) during development. SDF was selected as the testing airport for two primary reasons: 1). SDF airport is on the outskirt of Louisville, KY, which connects immediately to Indiana, thus attracting passengers from both states. 2). License plates of Indiana and Kentucky display county information. By leveraging the prior knowledge, we reformulate the state classification problem of SDF airport into a three-class classification problem (Class 1: Kentucky, Class 2: Indiana, and Class 3: neither Kentucky nor Indiana) instead of a 52-class classification problem. To realize the reformulation, we resize the final linear layer of the ResNet-50 DCNN model to the required new category size, initialize the weights of the final layer randomly, and initialize the rest of the layers in the DCNN model with the weights previously trained for the classification of 52 states. As shown in Figure 23, the trained RestNet-50 DCNN model can be deployed on a decision-making pipeline to real-time detection of state based on capturing video frames. After a license plate is detected on a frame in the preprocessing stage, the crop license plate part of the image will be given as the input for the ResNet-50-based classifier, which provides the state classification results for the detected license plate.

51 • Step 3.2: County Classification The result of state classification obtained from Step 3.1 is used as a reference to select the proper county classifier and the detected plate region obtained via exploiting OpenALPR (Rekor Systems, 2022) as the input of the county recognition model. The county recognition model includes a text detection model called CRAFT (Baek et al., 2019) and post-processing model based on edit distance that will be introduced in subsequent steps. This step still uses Louisville Muhammad Ali International Airport, KY (SDF) for algorithm development and testing. Based on its location, we target to design two county classifiers for Indiana and Kentucky, respectively. Both classifiers are realized by leveraging the open-sourced optical character recognition (OCR) models (Morales, 2022). Additionally, the extracted tests via the OCR models are further processed via our post-processing model, where the post-processing model considers the edit distance to match the county name, which is closest to the recognized texts at a particular region on the plate based on the given state. For example, Indiana displays a two-digit county code located at the bottom right of the plate, and Kentucky prints county names at the central bottom of their license plates. For some detected car plates, the county name is covered by the license plate frame. In this situation, it is difficult to detect the covered or semi-covered texts. Therefore, our county recognition tool outputs unfound for this situation. • Step 3.3: Plate Number Identification As illustrated in Figure 23, this tool identifies the plate numbers by exploiting OpenALPR (Rekor Systems, 2022). The functionality of each stage of the pipeline can be summarized as eight main steps: 1) Detection: finding potential license plate regions, 2) Binarization: converting the plate region image into black and white, 3) Character Analysis: finding character-sized "blobs" in the plate region, 4) Plate Edges: finding the edges/shape of the license plate, 5) Deskew: transforming the perspective to a straight-on view based on the ideal license plate size, 6) Character Segmentation: isolating and cleaning up the characters so that they can be processed individually, 7) OCR: analyzing each character image and providing multiple possible letters and associated confidences, and 8) Post Processing: creating a top n list of plate possibilities based on OCR confidences. Additionally, it also performs a regular expression to match against the region templates if requested. In this task, we utilize OpenALRP for license plate detection and plate number identification on each captured video frame. Plate detection is also realized as a preprocessing stage for the state classification as stated in Step 3.1. As illustrated in Figure 24, some frames can capture multiple license plates. In this work, we only consider the captured license plate with the highest confidence of a frame to simplify the implementation and improve the trustworthiness of the outputs.

52  Step 3.4: Post-Processing Design The post-processing model is critical in our proposed deployment for a couple of reasons: – As seen in Figure 24, a video stream can capture the same license plate in multiple frames and the number of frames associated with the same license plate can vary. Therefore, the distribution of captured license plates in the video stream and the actual observations do not always match. To address this issue, a post-processing model is required to remove the detected overlapping license plates. – For the same license plate, some captured frames enable better prediction results, especially on plate number identification tasks. This is because of the change in capturing angle and lighting conditions. The essential principle of our post-processing model is to improve the accuracy of the final prediction outputs by clustering all the detected data samples belonging to the same license plate and then achieving a highly trustworthy detection output according to the majority-based decision within each cluster. Based on this essential principle, we design the proposed post-processing model consisting of two main steps: Figure 24. Example of frames included in captured video stream – Cluster the plate number prediction As shown in Figure 23, plate number prediction tuples (plate number, confidence) are buffered and clustered, such that each cluster contains predictions corresponding to the same plate. The clustering is performed considering the sequential nature of the predictions and a similarity measure. That is, since we analyze the video frames sequentially, predictions belonging to the same license plate are already clustered together in the prediction sequence. The remaining challenge is that the boundaries of each cluster are still unknown. To address this challenge, we exploit similarity measures on the previous and new predictions for identifying the boundaries. As stated previously, the video stream can capture two or more license plates in the same frame, as shown in Figure 24. There can be an overlap between clusters at the boundary. Therefore, when deciding on a cluster of a new prediction, the similarity of it for previous two clusters is considered. If it is sufficiently different from the previous two clusters, a new cluster is created.

53 Additionally, the edit distance of strings, formally known as Levenshtein distance (Hofmann, 2019), is used as a similarity measure. Levenshtein distance between two words is defined as the minimum number of single-character edits (insertions, deletions or substitutions) required to change one word into the other. If the edit distance is less than half the length of the new prediction compared to the previous cluster, the new prediction is considered to belong to the same cluster. To filter out the prediction that potentially does not identify all the characters on the license plate, additional filtering based on string length is also designed for each cluster. Furthermore, considering that some predictions identify shadows on license plate edges as characters, predictions with string lengths of more than 2 less compared to the maximum string length on each cluster are also filtered out. – Select the best plate number prediction for each cluster To achieve a highly trustworthy prediction output for each cluster, the cluster is further filtered by the confidence level. In our current design, five predictions with the highest confidence within a cluster are considered for further processing. Within these five predictions, if a repeating plate number pattern is observed, this plate number will be identified as the final prediction with high confidence. If two plate numbers are observed with repeating patterns, both are given as output with a low-confidence indicator. Additionally, if no repeating patterns are observed, all five predictors are given as output with a low- confidence indicator. If a low-confidence indicator is presented, a human operator may need to visually verify and identify the correct license plate number. • Step 4: Results Interpretation As mentioned above, we use Louisville Muhammad Ali International Airport, KY (SDF), to develop and test the License Plate Analysis tool. This section provides the results of deploying the tool at SDF. – State Classification  The testing dataset consists of 1,385 license plate images, some of which include the same license plate captured at different angles as the camera moves. The overall classification accuracy of the testing dataset is 93.63%. This accuracy can be further improved if human operators are able to identify the prediction results with lower confidence. For example, if we consider the confidence level higher than 0.98 as high confidence, 91% of the predictions will belong to the high-confidence category, and the human operator only needs to verify the left 9% of testing data that have a lower confidence level. The confusion matrix of high-confidence predictions is presented in  Table 5. In this case, the classification accuracy of high-confidence predictions is 98.88%. If we consider that the low-confidence predictions will get correctly identified by a human operator, the accuracy of the human-assisted system can reach 98.99%. Table 5. Confusion Matrix for High-confidence State Classification Predictions Actual Indiana Actual Kentucky Actual other states Predicted Indiana 282 4 0 Predicted Kentucky 0 893 6 Predicted Other States 0 4 71

54 – County Classification  In our post-processing model, we use edit distance to match the detected text results with the closest county name or county number, the matched results will be modified to the closest known county name or number. Thus, if a county name is detected, the matched output will be 100% correct. The only exception that occurred is that one image has a detected county name that is equally close to two different counties. In this case, it is difficult to identify the correct county name, and thus we drop this result as the wrong detection. As shown in Table 6, the frame-based detection accuracies of Indiana and Kentucky counties are 89.43% and 95.23%, respectively. Table 6. Detection Accuracy on County Identification State Total images Numbers of images with county name Numbers of correctly detected images Numbers of incorrectly detected images Numbers of undetected images Indiana 312 123 110 0 13 Kentucky 245 147 140 1 6  Some examples are presented in Figure 25. As illustrated, most plates with undetected county are resulted by the situations where close to half of the county-related information is covered by the frame or the information is very difficult to recognize even by human eyes. Figure 25. Examples of plate recognition outputs

55 – License Plate Number Recognition  Clustering-based number identification is also performed for the video samples. The results are shown in Table 7. As shown in Table III, the accuracy of detecting all characters correctly after our post-processing model is 73.27%. The accuracy of detecting license plate numbers with a maximum edit distance of 1, which is maximum of one character difference, is 85.02% for these captured videos. Table 7. License Plate Number Identification Results Video index Number of plates Not detected One character missing Clip 1 80 6 10 Clip 2 39 7 1 Clip 3 33 4 6 Clip 4 95 20 12 Considering many airports have parking management systems capable of taking pictures of license plates with good angles and lighting, it is reasonable to leverage the dataset generated from this system to enhance the plate number identification accuracy. – Mapping  Using the identified state and county information from license plates, we are able to plot a map of passengers’ origins for SDF airport at the county level (Figure 26). As expected, SDF airport draws most traffic from northern Kentucky counties, including Jefferson County, where SDF is located, Hardin County, Oldham County, Anderson County, and Warren County. However, data representing southern Indiana counties were very limited in the sample we collected on May 16, 2022. The scarcity and discontinuity of license plate data from different counties in Figure 26 reveal the limitation of using the License Plate Analysis tool without access to vehicles’ registration records from state government agencies.

56 Note: License plate sample date is May 16, 2022 Figure 26. Catchment area map of SDF based on license plates of parked vehicles

57 Parking Reservation Analysis Tool A growing number of US airports are allowing passengers to book their parking spaces in advance. By doing so, passengers will have guaranteed parking spots before arriving at airports, and airports are able to better predict demand for their parking facilities. The reservation/booking system with payment options also allows data collection for further business analytics. For instance, using the assumption that most passengers’ billing addresses are also their residential addresses, we will be able to map passengers’ origins based on the ZIP codes of their billing addresses. We tested the approach of generating catchment area maps using parking reservation data supplied by Ontario International Airport (ONT). The data contain only two variables, which are passengers’ ZIP codes and booking arrival dates. From January 12 – May 27, 2022, ONT received 26,882 parking reservations. By aggregating these reservation data to the ZIP level, we are able to plot the summary data on a map representing the origins of passengers who booked the parking spaces. See Figure 27 for the results of the mapping. Figure 27. Origins of Parking Reservations for ONT during January 12 – May 27, 2022

58 Billing Data Analysis A common method for traditional catchment and leakage analysis has historically been ticket billing data from Airline Reporting Corporation (ARC) or Marketing Information Data Tapes (MIDT). Both sources have various factors that need to be weighed by an airport when deciding on the data source to define their catchment area. The analysis below will review airport use cases for billing data using ARC and MIDT sources. Additionally, we develop a standard process map for developing catchment area analysis based on billing zip code data. Usage This tool is used by airports to • Develop a catchment area using traditional billing zip code data sources. This is often the first step for leakage studies. • Define airport utilization rates by zip code at a market level. Data Requirements This tool uses the following data • Billing data based on ARC’s Market Locator dataset or a MIDT-based point-of-sale database • Point-of-origin data based on an industry demand set that has been adjusted to match actual passenger volumes Software Requirements An advanced Microsoft Excel user can complete the entire process in Excel for most smaller airports. A large data processor such as Microsoft SQL or Python is recommended to complete the process for airports located in multi-airport catchments or larger airports. Theoretical Background: Billing Data Catchment analysis typically assumes a person’s primary residence is the best location to utilize when determining what airport/s they are likely to patron, since their journeys that involve air transportation are likely to begin and end at home. While air travel journeys sometimes commence from or end at a location other than home – for example, an office – we presume it is rare for an air travel journey to both begin and end at a location other than home. To determine where individual lives, airlines, airports, and consultants have historically examined billing data. Most passenger reservations include information on the billing address for the passenger/s on that reservation. There is a high correlation – particularly for non-business travelers -- between their billing zip code and where the customer lives.

59 However, billing data itself is not without limitations that may cause distortions or data interpretation difficulties. For example, not all credit card billing addresses are tied to where an individual lives. For many business travelers, corporate credit cards are issued to the employee and are used for travel-related expenses, including airfare -- but these credit cards will more frequently have billing information that is tied to the employee’s corporate office, which may or may not be in the same metro area as the employee. This limitation should be considered for all catchment studies where billing zip code data is the primary source for defining an airport’s catchment area. An employee may live in the Oklahoma City metropolitan area, but their corporate travel credit card may be tied to the employer’s corporate address in Chicago. In this example, the traveler’s billing zip code would be located in the Chicago area rather than Oklahoma City. As will be demonstrated later in our analysis, smaller airports that are highly dependent on corporate or military traffic could face difficulties utilizing billings data to determine their catchment regions, since a disproportionate number of their departing passengers may have billing zip codes that lie outside the originating airport catchment region. • MIDT Data Marketing Information Data Tapes (MIDT) data is collected from bookings via a Global Distribution System (GDS). These GDSs are portals for traditional and online travel agencies to book airline tickets on behalf of their clients. These travel agencies can be small local agencies, large corporate agencies (e.g., American Express Global Business Travel, BCD, CWT) or large online retailers (Travelocity, Expedia, etc.). MIDT is based on bookings activity, not tickets. There is a difference between booking with an airline and ticketing – booking activity holds the inventory, but the fare is not paid until the booking is ticketed. Historically, travel agencies could book a reservation with an airline but delay the ticketing/purchase process. Today, most fare products have tight ticketing time limits that prevent inventory from being held without ticketing. There are exceptions in certain international and/or specialty product markets (e.g., cruise reservations, military and other negotiated rate bookings) where ticket ticketing time limits do not apply, and therefore bookings data may overstate actual passenger activity. MIDT data is packaged and sold by the largest GDSs, including Sabre, Amadeus, and Travelport. These three GDSs make up a vast majority of global GDS bookings. GDS providers share summary level booking information for bookings with each other. This information sharing allows all MIDT providers to estimate global airline demand. However, the information shared is summary information only. Metadata such as point-of-sale or ticketing agency information is not shared between the competitors. Therefore, when purchasing MIDT data for metadata research -- such as agency point-of-sale, point-of-origin, or length-of- stay analysis -- it is important to understand which GDS provider has a strong representation in the geographic area being researched. Generally, in the United States, Sabre, Amadeus, and Travelport all have strong agency representation and broad geographic coverage. This analysis utilizes Sabre’s Market Intelligence Global Demand Data (GDD) for analysis.

60 • ARC Data The Airline Reporting Corporation (ARC) is an airline-owned, privately held company that settles payments between participating airlines and United States-based travel agents. ARC Data comes from tickets sold and ticketed through ARC’s airline settlement plan (ASP) clearinghouse. After a travel agency books and tickets a reservation for a customer, the travel agent enters the customer’s credit card and billing information into the reservation. This data is then passed to ARC via ASP. The ARC settlement process covers travel agents in the United States, including larger online travel agents (OTAs) such as Expedia or Travelocity. Unlike MIDT, ARC data collects customers billing zip codes rather than the travel agents’ zip codes. ARC billing data is not a complete set of all billing zip codes processed thru their settlement process. According to ARC, only 20% of the tickets processed and settled through ARC are matched with the customers billing zip code. • Strengths and Weaknesses of ARC Billing Data ARC has the ability to offer additional insights that cannot be gleaned from publicly reported data sources, such as the Department of Transportation’s DB1B data. ARC offers insights into point-of-sale information rather than just point-of-origin. These are two separate concepts. Point-of-sale is the address zip code associated with the reservation. Point-of-origin is the first airport on an entire itinerary, including outbound/return or multiple operations segments. Point-of-origin data is available in the DOT’s DB1B database. Effectively, point-of-origin data allows a researcher to understand whether traffic on a given O&D pair is disproportionately driven by one airport origin versus the other. For example, a market with a strong inbound leisure component, such as Arkon- Canton to Las Vegas, is likely to have a higher Arkon-Canton airport point-of-origin than the Las Vegas airport. Point-of-origin data is important for an airport to assess how important their local catchment is in driving total passenger volumes. All else equal, an airport with a high inbound travel component may be less likely to benefit from catchment enhancement efforts than an airport with a lower inbound travel component since a higher share of total passengers on aircraft to/from the city are merely visiting the local community. One of ARC’s primary strengths is it captures the billing zip code for the credit card used to purchase an airline ticket. By using the credit card’s billing zip code, catchment research can more closely mirror where an individual resides. Additionally, ARC data does not require the travel agency associated with a ticket to be located in a specific billing zip code to identify bookings associated with that zip code. While ARC offers insights into the passenger’s point-of-origin, they do have specific limitations. Both ARC captures passenger data from travel agency-dependent booking channels. According to a 2016 IATA airline distribution analysis by Atmosphere Research Group, around 50% of airline reservations come via travel agency channels. While a 50% sample rate seems large, a few considerations need to be factored in when developing a catchment study. First, not all travel agency reservations or tickets will have relevant (or any) point-of-sale information included with the record. As demonstrated previously, MIDT data’s point-of-sale concentrates around travel agency locations rather than where an individual lives or works. Only 20% of ARC records capture a billing zip code for ARC data. This means ARC only has a 10% sampling rate for overall bookings (Figure 28).

61 Sample sizes are not the only consideration when using ARC data. Customer segmentation between travel agency bookings and direct distribution bookings (airlines’ websites and mobile applications) could be materially different from one another. While the different segments may not cause an issue on an aggregate, at a route level, different customer segments could drive material differences in point-of-sale or booking trends. Consider a carrier such as Southwest Airlines. Southwest Airlines distributes most of its tickets via direct distribution channels. In 2021, Southwest reported that 86% of its revenue originated thru its direct distribution channels. This means the ARC and MIDT sampling rate for Southwest and other similar carriers would be low. Southwest’s third-party distribution strategy targets corporate travel customers and travel management companies. These distribution streams focus on more inelastic business travelers who may have different travel patterns than more elastic, leisure-focused travelers. Often carriers do not distribute their tickets in the GDSs at all. Certain ultra-low-cost carriers (ULCC), such as Allegiant, only distribute their tickets via direct distribution channels. This means Allegiant Air’s point-of-sale data is not included in ARC datasets, and assumptions must be made regarding their point-of- sale mix. Other ULCC carriers, such as Spirit and Frontier Airlines, fully distribute their tickets to travel agencies through GDSs, however, they do not participate in the ARC settlement process. This means Spirit’s and Frontier’s bookings will have no data available from ARC. In either of these cases, a catchment study has to factor in both of these limitations. One of ARC’s most significant limitations is the validity of the location data. While both data sources focus on the travel agency distribution channels, ARC only provides ticket settlement services for United States travel agencies. This means that while ARC has participating international carrier and billing zip code data in the dataset, the data is only reliable for United States point-of-sale markets. ARC’s effectiveness with international leakage studies would only be relevant for destination markets, such as Cancun or Montego Bay. Finally, it is essential to note that while billing data may provide additional insights into where a customer originates, billing zip and home zip codes may be materially different. This is critical for smaller markets that are heavily dependent upon managed corporate or military travel and is demonstrated in the catchment study process with Albert J Ellis Airport (OAJ) in Jacksonville, North Carolina. Ellis Airport estimates that approximately 50% of its traffic is military-related travel. Much of this travel is managed and booked outside the OAJ catchment area, meaning the billing zip codes associated with military travel could reside outside the OAJ catchment area. This can be demonstrated in the carrier/market point-of-origin to ARC ratios, which will be discussed in- depth in the process section. As the sampling rate for a catchment area decreases relative to its point-of- origin traffic, the carrier/market point-of-origin ratio increases. For OAJ, all of its top business and military markets have significantly higher ratios than observed with CAK’s ratios.

62 Figure 28. Carrier/Market Industry POO to ARC Ratios OAJ vs. CAK  Strengths and Weaknesses of MIDT Billing Data MIDT shares many of the same strengths and weaknesses found with ARC: dependency on GDS reservations; carriers with various distribution strategies; differences in point-of-sale vs. where an individual resides. There is, however, one critical difference between ARC and MIDT. While ARC’s point-of-sale information depends on a traveler’s credit card billing zip code, MIDT depends on the location of the travel agency booking the reservation. Since MIDT data captures travel agency details, the reservation’s point- of-sale can be different from where the traveler lives. To demonstrate this limitation, we can query MIDT’s point-of-sale information to understand which zip codes drive the most reservations for travel within the United States (Figure 29). 0 5 10 15 20 25 30 35 40 AACLTOAJ AAOAJSAN DLATLOAJ AAOAJORD AACAKCLT UACAKORD AACAKDCA AACAKMIA OAJ Ratio CAK Ratio PO O In du st ry to A R C R at io

63 Source: Sabre Intelligence, Point-of-Sale, 2021 Travel Figure 29. Share of CAK 2021 MIDT travel agency by zip code When we examine the zip codes with the largest number of CAK point-of-origin reservations, we see that Bellevue, Washington generates approximately 47% of agency bookings for travel. Bellevue is home to the largest online travel agency in the United States, Expedia Group. All reservations made through the Expedia Group’s distribution channels will show as point-of-sale as Bellevue. This concentration of reservations significantly limits the MIDT’s reliability for granular catchment and leakage research. Other concentrations of reservations correlate to other large online travel agencies or travel management companies: Norwalk, CT is home to Bookings Holdings; New York City is Travel Leaders Group; Milwaukee is home to Adelman Travel. In contrast, ARC data captures the billing zip code for the credit card used to purchase an airline ticket. By using the credit card’s billing zip code, catchment research can more closely mirror where an individual resides. Additionally, ARC data does not require the travel agency associated with a ticket to be located in a specific billing zip code to identify bookings associated with that zip code. Figure 30 examines Akron-Canton’s airport utilization using a 75-mile catchment area at a zip code level. ARC data shows a more encompassing view of the catchment’s zip code level utilization on the left. On the right, MIDT is processed using the same methodology as ARC, allocating CAK point-of-origin bookings to MIDT captured travel agency zip codes within a 75-mile radius. Note all MIDT travel agency zip codes located outside the 75-mile study area are excluded. This excludes those bookings reporting zip codes relating to online travel agency headquarters locations. 0% 10% 20% 30% 40% 50% Bellevue, WA Norwalk, CT New York City, NY Milwaukee, WI Cambridge, MA North Olmsted, OH Sh ar e of A ge nc y B oo ki ng s

64 Figure 30. Share of 2019 MIDT Travel Agency by Zip Code Despite the limitations, for traditional catchment and leakage studies, ARC is preferred over MIDT due to the granularity provided by ARC billing zip codes. ARC data has a long and trusted history in providing foundation data for leakage and catchment studies. That said, for border airports located near a competitive international airport, MIDT provides high-level point-of-sale data that may be useful for international leakage studies. Process Flow At a high level, the catchment area analysis uses billing zip code data, typically from ARC, to allocate the originating customers back to their assumed home zip code. First, two data sources must be secured for the analysis: access to an industry demand dataset that has been adjusted to match actual passenger volumes and ARC billing data. Industry demand datasets can be licensed from various industry data intelligence providers. Which industry demand dataset an airport should use depends on the passengers the airport is studying. The Department of Transportation’s DB1B data is likely to be the preferred dataset for smaller, domestically focused airports. DB1B is a 10% sample of flown passengers ticketed by a U.S. carrier. For airports that already have industry demand data, the dataset is likely based upon DB1B. There are some drawbacks to using DB1B that are relevant depending on the catchment analysis to be performed. First, the industry data is only based on U.S. ticketing carriers. If a foreign carrier serves the study airport or a nearby competitive airport, DB1B will not capture the foreign carrier’s traffic. Second, when using DB1B, there are restrictions on accessing and sharing the international traffic data derived from DB1B. If international traffic analysis is a significant component for the airport, DB1B may not be the preferred dataset. Third, the DB1B data is a random 10% sample, and therefore may not have desired granularity, especially in smaller markets. CAK Booking Density Based Upon ARC CAK Booking Density Based Upon MIDT

65 There are other industry demand datasets available besides DB1B. MIDT-based demand sets are often used for larger international airport analyses. MIDT-based demand sets are licensed in partnership with one of the major GDS providers. MIDT-based demand sets are based upon bookings reservation data (booked through all GDSs) with adjustments to develop a global demand set. In many markets, this results in a total market capture of up to 50% (or more) of total passengers. However, while the demand set is global and not subject to international restrictions, this data may be subject to sampling biases. MIDT is based upon GDS reservations, so market demand for carriers with little or no GDS distribution can be skewed – this particularly impacts low-cost and ultra-low-cost carriers. GDS providers often attempt to correct for these sampling biases, but the accuracy of these correction methodologies is uncertain. Additionally, fare data in the MIDT demand set is based on the fare booked only through the providing GDS rather than a global sample set. Depending on the market and GDS representation in the relevant geographic region being studied, fare information in the dataset could be materially higher or lower than the actual. Nevertheless, for catchment studies with a significant amount of international traffic, MIDT-based demand sets are typically the preferred industry demand set. To purchase ARC data, an airport can request Market Locator (billing zip code) data on the ARC website (ARC, 2022). The typical turnaround for the request is two to three weeks. Once both data sources are secured, the airport can begin pulling point-of-origin data from their industry demand provider. When pulling the data, the airport must include the following metrics: point-of-origin airport, origin and destination market, marketing carrier, and the number of passengers for the study period. If the catchment analysis is for a smaller airport, the following steps can be completed in Microsoft Excel. For larger airports or airports in multi-airport catchments, SQL or Python could be required. Both industry and ARC datasets can now be loaded into Excel. The first step in processing the Excel data is developing POO Industry to ARC Ratios for every marketing carrier/route combination. Generally, it is easiest to develop a PivotTable in Excel on the ARC billing data filtering on the studied point-of-origin airport. Industry point-of-origin data should be joined into the spreadsheet (Column C, Figure 31). The airport should divide point-of-origin industry data by ARC passenger data (Column D, Figure 31). This will develop a list of ratios for all ARC marketing carrier/market combinations. Note: Non-ARC participating carriers, such as Spirit Airlines in CAK’s case, will not be included in the ARC ratios. Allocating Spirit’s passengers will be handled in another step.

66 Figure 31. Developing POO Industry to ARC Ratio The airport needs to provide a list of zip codes they believe to be in their extended catchment area. This radius should be set to extend somewhat beyond the likely catchment range for the study airport, to ensure the geographic limits of the study airport’s actual catchment are accurately captured. For airports in urban regions with competing airports nearby, generally, the radius will be smaller. A larger study radius will be required for airports that are surrounded by rural communities with few alternative commercial airports. In the case of multi-airport cities and regions, the catchment study area should normally extend at least to the nearest competitive airports since this allows the catchment study to capture the study airport’s actual capture in geographic areas where passengers may choose between competing airports. Finally, the radius should consider that depending on transportation networks and the location of competing airports, the actual catchment of the study airport may be substantially larger in certain directions than others. In the case of CAK, for example, its likely geographic capture in the north/northwest directions is likely to be lower than toward the South/Southeast since Cleveland Hopkins airport is located 40 miles to the northwest of CAK. Our demonstration defines CAK’s study area to be a 75-mile radius (straight-line distance) for CAK, as this permits catchment study to include zip codes surrounding nearby CLE and PIT airports. After pulling the zip codes surrounding the airport, the next step of the analysis is for the airport to make a list for every airline/market and zip code combination. This process takes a considerable amount of computer resources and is the limiting factor between using Excel and a large data processor such as Python or SQL. In CAK’s instance, 571 zip codes were joined on 1,114 airline-market combinations (636,094 rows of data).

67 After all zip code and airline-market combinations have been entered into the spreadsheet, the airport should sum the number of ARC-reported bookings for the zip code and airline-market combinations. The airport should then join POO Industry to ARC Ratio developed in the previous step and multiply the ratio by the number of reported ARC bookings. This will give the airport assumed zip code level usage for all markets and reporting carriers. To maintain data integrity, an airport should not import all ARC ratios into the dataset. As ARC and industry demand sample sizes begin to get smaller, variability in the data increases, and the reliability of the calculated ratios declines. When to discontinue importing ARC ratios is subjective. For our CAK study, we discontinued importing ratios after the ARC sample size was less than 0.2% of all ARC-reported records. This rule excluded 24% of all ARC bookings relating to the CAK study area, which would then need to be allocated using our secondary and tertiary methods. As another quality assurance, an airport may decide not to import ARC ratios for airline-market combinations, which ARC reports less than 5-10% of overall airline-market industry demand (ratio larger than 10-20). For CAK’s catchment study, we eliminated all airline-market combinations with a ratio larger than 15x. This quality assurance process required 3% of ARC bookings to be allocated through other methods. Figure 32. Allocating Reported POO Passengers at Carrier/Market/Zip Code Level Next, the airport must allocate bookings for non-ARC reporting carriers and airline-market combinations that were not allocated previously. This allocation process is completed in two steps, first allocating non- ARC reporting carriers and then all remaining demand. When we allocate non-reporting carriers, most of the allocation will come based on LCC and ULCC bookings. To best model where the passengers are coming from, we allocate LCC/ULCC passengers based on origin and destination market, mimicking the booking patterns found in our primary methodology. This methodology assumes that leisure travelers likely come from the same zip code as reported by other carriers. However, the methodology is not without its disadvantages. It is possible that LCC/ULCC carriers may attract a different customer segment that may not reside in the same geographic areas found by reporting (network) carriers.

68 To allocate non-reporting carriers, the airport must sum the ARC adjust total (Column H, Figure 32) by market and zip code. It is best practice to do this with an Excel Pivot Table to reduce computer resources. The PivotTable should present the data as “% total of row” to generate the market’s share by zip code. Figure 33. Zip Code Share by Market The airport should then return to the previous worksheet, import the POO industry reported market size (Column I, Figure 34), and join the market-zip code level share (Column J, Figure 34). The POO industry reported market size can then be multiplied by the zip code level share to allocate non-reported passengers in top markets. The airport can then add ARC Adjusted Total (not shown, Column H) to Non-Reported ARC Passengers (Column K, Figure 34) to make “Allocated Passengers” (Column L, Figure 34). Figure 34. ZIP Code Share by Market

69 The airport will need to allocate the remaining balance of POO demand. These remaining markets are smaller O&D markets with smaller sample sizes, and thus can only be allocated based on the zip code utilization observed in the primary and secondary allocation methodology. As such, we allocate the balance of passengers by “Allocated Passengers” (Column L, Figure 34) zip code share. To do this, the airport should sum all of the “Allocated Passengers” by zip code and divide by the total allocated passengers. The overall zip code share should then be loaded back into Column N (Figure 35), “Overall Zip Code Share,” and multiplied by the POO industry reported market size (Column I, Figure 35). The “Remaining Allocated Passengers” then should be summed with “Allocated Passengers” to give the “CAK Total Market Size” (Column P, Figure 35). The airport should then do quality checks to ensure that “Total Market Size” is the same size as the reported point-of-origin market size. Figure 35. Allocated Remaining Markets Finally, the airport should repeat this process for nearby airports, assuming the airport has ARC data for the airports. If the ARC data is incomplete for nearby airports, the airport can use carrier/market ratios, substituting the original airport for the nearby airport. The airport should then sum all passengers at a zip code level to define zip code level usage for the studied airport.

70 Figure 36. Catchment Area Process Flow Chart Output After utilization by zip code is developed, the data should be mapped to visually demonstrate the catchment area. The map in Figure 37 examines how much traffic from each zip code is utilizing CAK vs. other competitive airports. This allows us to start defining our study airport’s catchment area. For CAK, the highest zip code utilized areas are between Akron and Canton, south to Cambridge, and west toward Wooster.

71 Figure 37. CAK Utilization by Zip Code Next, we utilize zip code utilization rates to provide a basis for an airport to best define its catchment. For CAK we may use a zip code utilization rate of 50% to define its primary catchment area, 25% to define its secondary catchment area, and 10% to define its opportunity catchment area. The primary catchment helps define the area where an airport’s most loyal customers live. An airport’s secondary catchment is where customers are likely to shift between two or more airports depending upon the market conditions (schedule, price, etc.). Finally, an airport’s opportunity catchment are areas where it is unlikely to pull from unless they offer a unique advantage over a competitive airport. See Figure 38 for detail.

72 Figure 38. CAK Primary and Secondary Catchment Areas Post-Analysis Notes It is also important to note that an airport’s catchment is not static and is dependent, in part, on the competitive environment of surrounding airports. In Figure 39, we have three views for CAK’s catchment: CAK and CLE have nonstop service, only CLE has nonstop service, and neither has nonstop service. As we can see, if CLE has nonstop service, much of CAK’s primary catchment fades away. For many airports, this indicates leakage rates rather than a contraction in the catchment area. CAK’s reduction in its primary catchment when only CLE has nonstop service demonstrates the need for the airport to market directly to these customers to prioritize their use at CAK over CLE. There are top markets, however, that the airport cannot market itself to recapturing much of the secondary catchment area. These markets include top northeast business markets, such as New York, or shorter-haul hubs like Detroit. Without competitive nonstop service, passengers will likely prioritize CLE over CAK. In a smaller passenger volume but important bucket, CAK’s primary and secondary catchment areas greatly expanded when neither CAK nor CLE has access to nonstop service. This finding hints that passengers prefer to fly from CAK vs. CLE when all things are equal.

73 Figure 39. CAK Catchment Area by CLE Competitive Environment

74 User-Generated Content (UGC) Analysis Social Network Service (SNS) applications such as Facebook, Instagram, and Twitter 1 are popular platforms for travelers to share their traveling experiences in the form of text, images, videos, and audio. Such data shared by users are referred to as User-Generated Content (UGC). UGC is a promising source of stated preference data as it is primarily unsolicited and thus representing genuine opinions and preferences of random users. Usage The User-Generated Content (UGC) Analysis tool is designed and trialed to infer travelers’ origins based on location data they share on SNS applications. Potential users of this analytical tool are able to: • Collect UGC data shared by users on various SNS applications with regard to their business entities (such as airports), • Identify users’ SNS accounts that post relevant information about their business entities, and • Infer frequent locations from the previous sharing of the identified users. This toolkit uses Twitter to demonstrate the process of mining and analyzing UGC data to infer users’ origins. Data Requirements In order to deploy the User-Generated Content (UGC) Analysis tool, airport users need access to the following: • Authorization of mining UGC data in the selected SNS application(s), or • UGC data shared in the selected SNS application(s) Some SNS applications, such as Twitter, allow users to access UGC data shared on their platforms for academic research purposes. Users can apply for such access to mine relevant data. Users need to be aware that such access is often restricted to non-commercial use. Software Requirements Mining and analyzing UGC data require special tools. Various programming languages are capable of handling this task. In this toolkit, we use Python to demonstrate the process of mining data from Twitter. Analytical Procedure The procedure of mapping the origins of airport visitors using Twitter data is composed of four steps, which are Access Application, Data Collection, User Tracing, and Visualization. See Figure 40 for the analytical procedure of the UGC analysis tool.

75 Figure 40. Analytical procedure of the UGC Analysis tool Access Application The Twitter Application Programming Interface (API) is required to programmatically retrieve and analyze Twitter data. Twitter currently offers three access levels for their API v2. Details of these levels are provided in Figure 41. Users are recommended to select appropriate access levels based on their anticipated project requirements. While access level Essential is free to any Twitter user, it has less functionality than the advanced levels. Two advanced levels, Elevated and Academic Research, require application and approval from Twitter. Interested users need to submit their requests through Twitter’s Developer Platform. In our demonstration, we applied and were subsequently approved for Academic Research API. Access Application •Apply for Twitter's Academic Research access Data Collection •Identify a list of airports (or other business entities) to be analyzed •Retrieve live tweets that tag users at those selected airports User Tracing •Identify users that have shared their locations at the selected airports •Trace previous locations shared by these users •Infer the mostly likely home locations of users Visualization •Visualize origins of users visiting airports on maps •Analyze traffic leakage to nearby competing airports (with sufficient data)

76 Source: (Twitter, 2022b) Figure 41. Three access levels of Twitter API v2 Data Collection With Twitter’s API v2, users now have the required access to retrieve historical and live tweets. In order to test the amount of UGC data generated daily for airports of different categories, we selected five airports representing different regions and airport categories to collect live tweets shared from these venues. The five airports are:  BOS – Boston Logan International Airport – Primary – Large Hub – New England Region: Boston, Massachusetts  DAL – Dallas Love Field – Primary – Medium Hub – Southwest Region: Dallas, Texas  ONT - Ontario International Airport – Primary – Medium Hub – Western-Pacific Region: Ontario, California  CAK - Akron-Canton Airport – Primary – Small Hub – Great Lakes Region: North Canton, Ohio  OAJ – Jacksonville Albert J Ellis Airport – Primary – Nonhub – Southern Region: Richlands, North Carolina

77 Using Python, we set specific rules to track and collect live tweets that are associated with the geographic location data (or “geo-tagged”) from any of the five selected airports. We understand Twitter users may not necessarily geo-tag the primary place_id of an airport but rather tag a specific location within the airport. To better capture these geo-tagged tweets, we use the reverse geocode and geo search function of Twitter’s API v2 to find and include places that may be associated with the primary place_id of an airport, such as airport terminals (Twitter, 2022a). To ensure the quality of data collected, we screen features of data, including follower/following count, account creation date, username, etc., to validate that collected tweets are from real travelers rather than bots, which are not uncommon in SNS applications. We run the Tweet collection app from February 1 – May 31, 2022, to collect any tweets that geo-tagged one of the five selected airports. A total number of 509 valid tweets have been collected from four airports. Considering geo-tagged tweets representing CAK, OAJ and ONT are very limited, our subsequent analysis will focus only on BOS and DAL. See Table 8 for the summary of tweets collected. Table 8. Summary of Geo-tagged Tweets Collected during February 1 – May 31, 2022 Airport Number of geo-tagged Tweets collected BOS 308 CAK 4 DAL 142 ONT 55 User Tracing The collected tweets that meet our geo-tag requirements contain a set of parameters called fields. In this project, we collect tweet fields, user fields and place fields. Two other available fields, media fields and poll fields, are not required by this project. Refer to the Twitter Develop Platform (2022) for a complete description of fields. Using author_id, we are able to identify individual Twitter users who posted the collected tweets during our data collection period. We then access historical tweets posted by the identified Twitter users to retrieve their previously shared locations. Another source of information that can be used to infer a Twitter user’s origin is the location provided in the user’s profile. See Figure 42 for the tracing process. Figure 42. Twitter users' location tracing process

78 However, both information sources have their limitations. As of previously shared locations, such data are very rare. Twitter has gradually evolved into a platform where people tend to express opinions about public issues, and fewer users are sharing personal experiences on Twitter. In addition, the majority of users are less likely to share their locations, especially near their residences, due to privacy concerns. In terms of user profiles, Twitter does not have any requirements for users to provide genuine addresses or current addresses. With the collected location data, further cleaning is necessary to filter out incomplete records and faked records. An example of such records includes those who only use state or county in the location data. Such records are too broad to make any meaningful inferences. Some tweets mention the specific city or county names but do on provide state affiliations. When there are duplicate cities or counties in more than one state, the detailed locations of the Tweet users cannot be confirmed. Such records will be removed from the collected data before plotting users’ origins. Visualization After data cleaning, we are able to plot locations of users who tweeted at BOS or DAL from February 1 – May 31, 2022. For each airport, we include two figures representing identified Twitter users’ likely locations across the US and near the airport. The first figure can be viewed as the origins of inbound visitors, and the second one is more likely to represent the potential locations of outbound passengers. See Figure 43 and Figure 44 for Twitter users’ origin map for BOS and DAL, respectively.

79 Figure 43. Location of Twitter users who tweeted at BOS during February – May 2022

80 Figure 44. Location of Twitter users who tweeted at DAL during February – May 2022

81 Cross-Validation Airports are recommended to pick more than one tool from all the analytical tools included in this toolkit to conduct their catchment area analysis so that results produced by different tools can be cross validated to ensure the validity of the analysis. During the development of these tools, we used a few airports to demonstrate the analytic process. This section presents analysis results of these airports side by side for cross-validation purposes. Akron-Canton Airport (CAK) Our Travel Utility Analysis tool predicts a competition-neutral catchment map for CAK, in which we assume all inputs are the same for CAK and its nearby competing airports. As seen in Figure 45, CAK is most competitive in the Canton region, and its attraction extends to the south along the interstate highway I-77. On the other hand, the influence of CAK to its north fade quickly due to the strong market pulling effect of the Cleveland Hopkins International Airport (CLE). Note:  This map is generated by the Travel Utility Analysis tool using the following inputs: – Identical airfare, trip time, and parking cost at CAK, CLE, CMH, and PIT – Single traveler, 5-day round-trip, $21 VTTS-Ground, $39 VTTS-Air, $0.585/mile IRS mileage rate  The airport catchment area is color coded with yellow, orange and red. Figure 45. CAK catchment areas generated by the Travel Utility Analysis tool The prediction is confirmed by airport visitors’ mobile location data (Figure 46) and outbound passengers’ billing data (Figure 47). Though there are some sparse data from areas that are distance from the airport, most of the records indicate that the core market of CAK is indeed concentrated in the Akron- Canton area and extends south along I-77.

82 Figure 46. CAK catchment area maps generated by the Mobile Location Data Analysis tool Figure 47. ZIP-code level catchment areas of CAK generated by the Billing Data Analysis tool

83 Ontario International Airport (ONT) Greater Los Angeles is a competitive market for airports. The Ontario International Airport (ONT), located on the east side of the region, mainly attracts travelers from Ontario, Pomona, Rancho Cucamonga, and San Bernadino, as predicted by the Travel Utility Analysis tool and shown in Figure 48. Note:  This map is generated by the Travel Utility Analysis tool using the following inputs: – Identical airfare, trip time, and parking cost at ONT, BUR, LAX, LGB, PSP & SNA. – Single traveler, 5-day round-trip, $21 VTTS-Ground, $39 VTTS-Air, $0.585/mile IRS mileage rate.  The airport catchment area is color coded with yellow, orange and red. Figure 48. ONT catchment areas generated by the Travel Utility Analysis tool Our analysis is confirmed by airport visitors’ mobile location data (Figure 49) and outbound passengers’ billing data (Figure 50). The majority of ONT’s customers are from nearby regions and northeast areas of Ontario, though there are also some isolated records from the Greater LA. As we have previously identified in the Preliminary Analyses, airlines operate some unique services from and to ONT. These flights are popular among travelers who wish to avoid departing from LAX due to the traffic condition of the region.

84 Figure 49. ONT catchment area maps generated by the Mobile Location Data Analysis tool Figure 50. ZIP-code level catchment areas of ONT generated by the Billing Data Analysis tool

85 Albert J Ellis Airport (OAJ) Albert J Ellis Airport (OAJ), located near Jacksonville, NC, faces competition from several airports in the region. Among these airports, Raleigh-Durham International Airport (RDU) poses the strongest attraction to travelers living within the catchment areas of OAJ. The Travel Utility Analysis predicts the catchment areas of OAJ extend approximately 25 to 30 miles from the airport (Figure 51). The advantage of OAJ over its competing airports is modest, even in its core market, as measured by the travel utility advantage. This suggests the airport is under constant threat of leaking traffic to nearby airports. Note:  This map is generated by the Travel Utility Analysis tool using the following inputs: – Identical airfare, trip time, and parking cost at OAJ, EWN, FAY, ILM, & RDU. – Single traveler, 5-day round-trip, $21 VTTS-Ground, $39 VTTS-Air, $0.585/mile IRS mileage rate.  The airport catchment area is color coded with yellow, orange and red. Figure 51. OAJ catchment areas generated by the Travel Utility Analysis tool The travel utility analysis results are confirmed by airport visitors’ mobile location data (Figure 52) and outbound passengers’ billing data (Figure 53). Mobile location data show very limited visits from travelers living outside of the Jacksonville region. And the OAJ utilization rates from billing data also suggest the primary catchment area living in or near Jacksonville. The influence of RDU and Fayetteville Regional Airport (FAY) is strong.

86 Figure 52. OAJ catchment area maps generated by the Mobile Location Data Analysis tool Figure 53. ZIP-code level catchment areas of OAJ generated by the Billing Data Analysis tool

Next: Post-Analysis Recommendations »
Toolkit for Establishing Airport Catchment Areas Get This Book
×
 Toolkit for Establishing Airport Catchment Areas
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The catchment area of an airport encompasses areas where passengers are more likely to use the subject airport, even when there are other airport options in the vicinity.

ACRP Web-Only Document 56: Toolkit for Establishing Airport Catchment Areas, from TRB's Airport Cooperative Research Program, comprises various analytical tools, such as the Travel Utility Analysis tool, that enable airport industry practitioners to calculate the likely responses of travelers to different market and operational inputs, thus forecasting potential catchment areas for airports.

Supplemental to the report are three case studies: Case 1: Akron-Canton Airport (CAK), OH; Case 2: Ontario International Airport (ONT), CA; and Case 3: Albert J. Ellis Airport (OAJ), NC.

READ FREE ONLINE

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!