Abstract
Accurate airport operations counts are important for determining appropriate funding allocations for airport development and improvement. Fewer than 270 of the 2,950 non-primary airports in the United States, however, have air traffic control (ATC) personnel who are available to count airport operations. Existing counting methods such as automatic acoustic counters (AAC) are not viable long-term solutions because of the expense and inconvenience of deploying the devices on a large scale. This paper validates a cost-effective counting technology based on a technique that uses signal strength obtained from aircraft transponders to register the occurrence of aircraft operations at non-towered airports. Over 50 million transponder records were collected from two different versions of the system, which were installed at Purdue University Airport (KLAF), Terre Haute Regional Airport (KHUF), and Indianapolis Executive Airport (KTYQ), all in Indiana. The operations counts calculated from these records were compared with those obtained from the Federal Aviation Authority (FAA)’s Air Traffic Activity Data System (ATADS) database, which contains official operations data reported by airports with ATC towers. The Version I device utilized a Raspberry Pi platform and produced error rates ranging from −10.2% to +7.6%. The Version II device consisted of the pre-production commercialized system and resulted in error rates ranging from −4.9% to −1.4%. The test results suggest that this pre-production implementation of the transponder signal-counting technology is an accurate and cost-effective way to count non-towered airport operations. Improvement and testing of this technology is being undertaken, and field deployments are ongoing at additional airports.
Accurate operations counts are critical for determining funding allocation for national airports. Such counts are also important for facilitating a thorough understanding of the national airspace system. The Federal Aviation Administration (FAA) annually spends approximately $1.0 billion in Airport Improvement Program (AIP) funding at 2,950 small commercial and general aviation airports. At airports with air traffic control (ATC) towers, aircraft operations are recorded by air traffic controllers manually; however, fewer than 270 of those non-primary airports have ATC personnel ( 1 ). Operations counts are therefore more difficult to obtain at non-towered airports due to an absence of full-time personnel. An accurate counting technology is reasonably expected to estimate airport operations with an absolute error rate of less than 10% based on data aggregated over 60 days or more. Consequently, several estimation methods and counting technologies have been developed in an effort to derive accurate total aircraft operations counts.
Three methods of estimating annual operations at airports were summarized by Muia and Johnson in Airport Cooperative Research Program (ACRP) Report 129 ( 1 ). These estimation methods included multiplying based aircraft by an estimated number of operations per based aircraft (OPBA), applying a ratio of FAA instrument flight plans to total operations (IFPTO), and using a sample extrapolation method to estimate an annual operations count. Since no official personnel are responsible for counting aircraft operations at all times during which the non-towered airports are open, OPBA and IFPTO methods were not recommended for estimating annual operations at these airports because of a lack of consistent OPBA and IFPTO figures at small, non-towered airports nationally. An extrapolation of sample data to estimate annual operations counts was, however, recommended ( 1 ). Existing counting technologies that are used to collect sample aircraft operations at airports were summarized by Ford and Muia ( 2 , 1 ). These technologies included automated acoustic counters (AAC), sound-level meter acoustic counters (SMAC), security/trail cameras (S/TC), video image detection (VID), and Automatic Dependent Surveillance-Broadcast (ADS-B) transponder receiver technology. According to test results provided by Bahler et al. in a highway traffic counting environment, the accuracy of AAC was impacted by low temperatures and other factors, so that the relative accuracy is roughly 15% ( 3 ). Using S/TC to count aircraft is labor intensive because of the need for manual tallying of images. VID is the most expensive option for counting aircraft operations; Muia and Johnson reported a lease cost of $36,000 for two cameras and one ADS-B receiver over the seven-month period of their evaluation of the technology ( 1 ).
McNamara et al. developed a technology that can be used for counting operations using extended Mode S aircraft transponder signals, which contain global positioning system (GPS)-derived aircraft position information and can be received with a 1090 MHz software-defined radio platform in conjunction with a single-board reduced instruction set computer and Linux operating system. This technology is cost-effective (less than $100 per unit for an experimental model) compared with acoustic counters (around $4,800 per unit) or VID, so it can be deployed on a large scale over long period of time ( 4 ).
In the United States in the calendar year 2014, approximately 72% of aircraft in the general aviation fleet were equipped with Mode C transponders, 9% were equipped with Mode S short squitter (SS) units, and 7% were equipped with Mode S extended squitter (ES) units (Figure 1) ( 5 ).
General aviation (2014) transponder capabilities in the national airspace system ( 5 ).
Mode A and Mode C transponders operate in identification-only pulse amplitude-modulated modes that transmit a four-digit octal code to the ground-based interrogating station. Mode C includes barometric altitude information; Mode A does not ( 5 ). Aircraft equipped with a Mode C transponder are interrogated by secondary surveillance antennas, which have a rotational period of approximately 4.8 seconds. Mode S signals differ from both Mode A and Mode C in that they are pulse position-modulated and contain altitude information and a 24-bit data stream that is a combination of parity information and an International Civil Aviation Organization (ICAO)-issued code used to identify the aircraft. In the United States, there is a unique, one-to-one correspondence between this ICAO code and the FAA aircraft registration code. Mode S ES replies (transmitted periodically without interrogation) contain a 56-bit data field used for transmitting both altitude and position information ( 6 ). The ES reply is capable of carrying more data than the basic SS Mode S version. For appropriately equipped aircraft, extended Mode S data is transmitted without interrogation at a nominal rate of one record every 5 seconds when airborne and every 10 seconds when on the ground.
Such transponder signals are easily obtained in a passive manner by inexpensive ground-based receivers, and the data from the signals can be stored to provide the researcher with the ability to derive a rich range of related operational metrics. While Mode S ES data is very useful in determining aircraft position relative to a particular runway, the aircraft fleet penetration of Mode S ES transponders, as noted previously, is only about 7%, despite an FAA requirement that most domestic aircraft be equipped with either Mode S ES or universal access transceiver (UAT) ADS-B transponders by January 1, 2020 ( 5 ). In contrast, because the combined penetration of Mode S SS and Mode C is about 81%, it is important to find opportunities to utilize this data (neither of which contain GPS-derived aircraft position information) to estimate airport operations counts.
Mott developed an aircraft distance estimation method based on Mode S SS and Mode C transponder signal strength by employing a self-calibrating adaptive digital filter ( 7 ). A methodology to extend the Mode S ES operations counting technology to include the use of the Mode S SS and Mode C data, thereby including a large portion of the general aviation fleet in the samples available for counting, was developed by Mott et al. ( 6 , 8 , 9 ).
Objective
The motivation for this research was the validation of a methodology for counting non-towered airport operations developed by Mott and Bullock, and the testing of the pre-production prototype of the receiver/data collection device, consisting of a low-cost receiver, self-calibrating signal-processing algorithm, and an estimation technique providing greater accuracy than that associated with traditional acoustic counters (Figure 2).
Block diagram of pre-production prototype (Version II).
Count Registration Process
The researchers developed a technique to use Mode C, Mode S SS, and Mode S ES signals to count aircraft operations, as described in ( 5 – 9 ).
Altitude information can be obtained from most received transponder signals, with the exception of Mode A, which, as noted, is less common. Because Mode C and Mode S SS responses do not contain position or heading information, aircraft position must be estimated from the strength of the received transponder signal. Once an appropriate threshold detection level has been determined, the distance for a particular aircraft may be measured. Consecutive transponder records with decreasing distances and altitudes below that of the airport traffic pattern suggest that the related aircraft is executing a landing, while records with increasing distances and altitudes suggest that the associated aircraft is engaged in a take-off.
Mott described the heuristics for recording airport operations using received transponder records ( 5 ). For Mode S ES operations, an air-to-ground or ground-to-air transition between contiguous entries is identified. These entries must be separated by greater than 10 seconds (to eliminate erroneous transitions due to bounced landings) and less than 90 seconds (to account for what is likely a separate operation). If the aircraft is within 35 degrees of the runway heading, as determined from the transponder record, an operation is registered.
Accurate estimation of aircraft distances is necessary for Mode C and Basic Mode S operations, as those two modes do not contain encoded position information. In order to estimate the distance between aircraft and receiver unit, a data vector consisting of eight signal strength values from the transponder receiver is collected and filtered using a combined digital adaptive first order low-pass Butterworth filter and Rayleigh maximum likelihood estimator. The filter coefficient is adjusted using distances computed from the known positions of aircraft equipped with Mode S ES transponders.
For a Mode C operation, consecutive transmissions must be separated by a distance of less than 1.1 nm and times of between 18 and 90 seconds; these conditions allow discrimination between possible multiple Mode C aircraft using the assumption that the maximum airspeed in Class D airspace is not exceeded. If the Mode C aircraft is registered as descending below a 300 Ft horizontal plane above the airport elevation, the operation is recorded.
Basic Mode S operations counts involve ensuring three conditions: the aircraft must be within 2 miles of the receiver, the aircraft must be 300 Ft below the pattern altitude with either consecutive increases or decreases in altitude, and more than 90 seconds must occur since the last operation of the aircraft.
Experimental Data Collection Infrastructure
To fully test and validate the estimation methods and counting technology, field deployments of Version I (the initial experimental transponder signal receiver and processing system) and Version II (the pre-production prototype of the commercial version of the signal-counting technology) devices were conducted at three airports over extended periods. During Version I and Version II data collection, more than 50 million transponder records were examined. These deployments were conducted at Purdue University Airport (KLAF), Terre Haute Regional Airport (KHUF), and Indianapolis Executive Airport (KTYQ). Among these airports, KLAF and KHUF have FAA air traffic control towers; KTYQ is a non-towered airport. The data collection sites may be seen in detail in Figure 3.
Data collection sites (with location numbers as callouts): (a) KLAF, (b) KHUF, and (c) KTYQ aerial photo.
The Version I device is the experimental version of the transponder data collection system created by McNamara, Mott, and Bullock ( 8 ). The signal processing platform included a Raspberry Pi single-board computer, a software-defined radio, custom scripts written for the project, and dump1090, an open-source script running on the Raspberry Pi and utilized to capture aircraft transponder signals ( 10 ). An R script was written by the authors to pre-process the signal records and output a .csv data file, as suggested by ( 11 ). Version I devices were deployed at two locations at KLAF for varying deployment windows and near KHUF for an eight-day deployment window. At KLAF Location 1, an indoor antenna was placed in an office window facing southwest, and over 1,000,000 transponder signal records were obtained from this deployment over a period of 60 days. At KLAF Location 2, a pole-mounted omnidirectional antenna was installed on the rooftop of the terminal building at KLAF. Over 15,000,000 transponder records were logged from KLAF Location 2 over 180 days. Finally, approximately 400,000 records were received by a Version I installation at KHUF Location 4 over a span of eight days. The antenna locations for these deployments are displayed in Figure 4. The data from Version I was collected to validate the accuracy of the methods utilized to process the transponder signals and register the operations counts, as developed by Mott ( 5 ). The authors then tested the counting performance of the Version II device.
Closeup views of several antenna deployment locations: (a) KLAF Location 1, (b) KLAF Location 2, (c) KLAF Location 3, and (d) KHUF Location 4.
Version II is a pre-production prototype of the Blueavion f1 device manufactured by Bluemac Transportation Data Systems and released on July 23, 2018. This device was developed from the Version I device and features a low-cost transponder data collection system in a solar-powered, self-contained unit. The device provides considerable flexibility regarding installation locations, a self-calibrating signal-processing algorithm, and a Bayesian estimation technique providing improved accuracy for small sample sizes. The Version II device is shown in Figure 5.
Field deployments of system used at KHUF (Location 5) and KTYQ (Location 6): (a) Pole-mounted Version II device and (b) Stand-alone Version II device.
The Version II device was deployed at KLAF, KHUF, and KTYQ. At KLAF Location 3, a Version II device in conjunction with an indoor antenna was installed in an office window facing southwest on December 1, 2017. Over 39,000,000 transponder records from Version II were logged at this site over a period of 178 days. At KHUF Location 5, a pole-mounted, solar-powered Version II device was located on the rooftop of the terminal building, facing south. About 2,800,000 transponder records were obtained at KHUF Location 5 over a period of 44 days. At KTYQ Location 6, a Version II unit was installed outdoors near a warehouse facing southeast on June 26, 2018. Over 2,900,000 transponder data records were recorded as of July 18, 2018 at this location. These data were used to validate the accuracy of the device, with additional data collected on an ongoing basis to continue the validation process over an extended period of time.
A summary of the collected data records from both Version I and Version II devices is presented in Table 1.
Overview of Collected Records
Total record counts were not calculated for Version I devices operating at KLAF and KHUF.
The KHUF air traffic control tower operates continuously; the KLAF tower does not. Thus, total records at KLAF differ from total records during control tower operating hours, while those at KHUF do not.
Version I data from Location 1, 2, and 4 provides total Mode S records below 2000 Ft, but not necessarily within 5 nm. Version II data provides total Mode S records below 2000 Ft within 5 nm.
Results
An analysis of Version I records from the KLAF Location 1 installation over a 60-day period indicated an overall percentage error between this data and the FAA ATADS data of slightly less than 1%. At KLAF Location 2 over a 30-day period, the model produced a 10.2% undercount compared with the ATADS count, with the absolute percentage error decreasing as the data collection period increased. Additional Version I data from KLAF Location 2 over a 90-day period was tested; the resulting error rate was 1.7%. Finally, at KHUF Location 4 over an 8-day period, the percentage error was approximately 3.4% (Table 2).
Version I Test Results
It is always possible to select particular data points where behavior of individual subsets is not representative of the aggregate case; this was discussed by Mott ( 5 ).
Based upon the validated estimation methods and counting technology, the authors deployed Version II at KLAF, KHUF, and KTYQ to examine the performance of these devices. At KLAF Location 3 over a 179-day period, the model resulted in a 1.4% undercount as compared with the FAA ATADS count. As the data collection period increased from 30 days to 179 days, the error rate decreased significantly from a 4.2% undercount to a 1.4% undercount. The error rate at KHUF (Location 5) over a 44-day period was 4.9%. The count at KTYQ Location 6 was 7,584 over a 22-day period (Table 3); however, because KTYQ is a non-towered airport, the ATADS count at KTYQ was unavailable, so no comparison could be made.
Version II Test Results
Testing break due to updating of the collection units.
Testing break due to shutdown for data retrieval.
ATADS data was unavailable at KTYQ due to lack of an ATC facility.
Note that there is some overlap and disjunction in various data collection periods due to the times at which the different versions of the devices were available for testing. In addition, some gaps in the data itself occurred due to losses of electrical power to field-based units. Regardless, the test results suggest that the pre-production prototype of the transponder signal-counting device is an accurate means of counting operations at non-towered airports.
Further Work
The data collection and validation described herein is continuing at KLAF, KHUF, and KTYQ, and new deployments are being implemented at additional small general aviation airports with fewer annual operations than those examined in this study. Future research will include a refinement of algorithms to further improve the accuracy of the signal processing algorithms and decision heuristics, and an extraction of additional information from collected data, including aircraft type. This information is expected to provide additional insight to airport managers about the fleet mix of aircraft operating at their respective airports.
Conclusion
Current operations counting technology at non-towered airports is generally somewhat inaccurate, labor-intensive, and sensitive to prevailing environmental conditions. The technology described in this article has the capability to provide information about aircraft operations that is either not available from other automated counting devices (e.g., acoustic counters) or not available without a great deal of effort (e.g., visual records from security/trail cameras). This study validated a cost-effective data collection system for non-towered airport operations counting that is more accurate than traditional counting technologies (Table 4). Version I and Version II devices were deployed to validate the accuracy of the aircraft operations estimation technology developed by Mott and Bullock. Over 50 million transponder records from KLAF, KHUF and KTYQ were collected and processed to produce regular operations counts. Over different time periods which ranged from 8 days to 180 days, the accuracy of operations counts from the Version I devices ranged from −10.2% to +7.6% as compared with FAA ATADS counts. The Version I test results suggest that the new method of registering operations counts based on transponder signal data is more accurate than other approaches in use. Over different time periods which ranged from 22 days to 179 days, the differences between ATADS and estimated operations counts from Version II ranged from −4.9% to −1.4%. Results from the Version II device suggest that the pre-production prototype of the transponder signal-counting device is an accurate means to count operations at non-towered airports.
Basic Accuracy and Cost Information for Existing Counting Technologies
All data in this table were retrieved from ACRP Report 129 ( 1 ).
The costs are represented as paid for the equipment tested in ( 1 ), and do not include any installation time (except for the leased VID equipment) or data retrieval time.
The costs decrease to $31,000 without the ADS-B receiver. This is a lease cost and will vary from airport to airport depending on the airport layout.
Footnotes
Acknowledgements
This research was funded by the FAA’s Project to Enhance General Aviation Safety and Sustainability (PEGASAS) Center of Excellence. The pre-production prototype shown in Figure 4 was provided by Bluemac Transportation Data Systems. The authors have no financial interest in Bluemac Transportation Data Systems.
Author Contributions
The authors confirm contribution to the paper as follows—study conception and design: JHM; data collection: BH, SZ; analysis and interpretation of results: JHM, CY; draft manuscript preparation: CY, JHM, BH, SZ, DMB. All authors reviewed the results and approved the final version of the manuscript.
The Standing Committee on Intergovernmental Relations in Aviation (AV010) peer-reviewed this paper (19-00693).