Abstract
Global navigation satellite systems have enabled significant improvements in aeronautical navigation. However, in recent years, a growing number of interference events have been reported by flight crews. In this paper, we first identify such events using crowd-sourced surveillance data collected between February and December 2022 for three different regions: the Baltic states, eastern Europe bordering the Black Sea, and the eastern Mediterranean. Then, we assess the extent and duration of these events to determine their impact on civil aviation. The analysis shows different characteristics, ranging from isolated events to regular large-scale and recurrent disruptions. Next, we identify aircraft types for the affected flights and evaluate flight plan data with respect to navigation equipment in order to identify flights that rely solely on satellite navigation and that might require assistance in the case of a loss of satellite navigation. Finally, we show the impact of radio frequency interference (RFI) on a selected passenger flight by analyzing automatic dependent surveillance-broadcast data as well as avionics data obtained from the airline’s flight data monitoring department for that specific flight and link the observations to the warnings triggered by the aircraft to alert the flight crew while encountering RFI.
1 INTRODUCTION
Currently, air traffic relies heavily on global navigation satellite systems (GNSSs) for accurate and reliable positioning and guidance in most phases of flight, including departure, cruise flight, and arrival. For the approach and landing phases, satellite-based augmentation systems and ground-based augmentation systems (GBASs) provide corrections for GNSS signals, yielding improved accuracy. With additionally provided integrity parameters, residual navigation errors can be safely bounded, and the augmented GNSS position information can be used for a variety of applications, such as en route navigation in airspaces where performance-based navigation (PBN) is specified, navigation in terminal airspaces, and even precision approach guidance. Because GNSSs are currently the primary means for providing integrity for a position solution in aeronautical navigation, GNSSs are generally the only means of navigation for supporting procedures with a required navigation performance.1 Furthermore, GNSSs feed a variety of other aircraft systems, such as the aircraft clock, terrain awareness warning systems, and different surveillance functions of the aircraft, as described in the in-service information provided by Airbus (2019) on potential cockpit effects in the case of GNSS loss. Automatic dependent surveillance-broadcast (ADS-B) technology transmits aircraft position information along with parameters regarding the estimated position uncertainty (EPU) and position source integrity level (SIL). These parameters change when GNSS-based navigation is disturbed or interrupted and can thus serve as indicators for determining regions where aircraft are subject to radio frequency interference (RFI) in the form of jamming of GNSS signals. Unfortunately, there are many areas globally that either were or still are regularly affected by RFI. In the United States, two incidents near the airports of Denver (Cybersecurity & Infrastructure Securtiy Agency (CISA), 2022) in January 2022 and Dallas in October 2022 (Liu et al., 2023) significantly disrupted operations. Since the beginning of the conflict in Ukraine in early 2022, there have been numerous reports and documented instances in which air traffic over Europe experienced a denial of GNSSs by large-scale jamming. These incidents were mainly observed over Finland, the Baltic states, and the Russian exclave Kaliningrad as well as in Romania, Bulgaria, and above the Black Sea, as described by the European Aviation Safety Agency (EASA) in a safety information bulletin (SIB) (European Union Aviation Safety Agency, 2022). Two revised versions of this SIB were published in February 2023 and November 2023, stating that “GNSS jamming and/or spoofing has intensified in recent months” (European Union Aviation Safety Agency, 2023a; European Union Aviation Safety Agency, 2023b) and extending the list of affected regions to also include, among others, the flight information regions (FIRs) of Istanbul and Ankara, as well as Tiblisi, Yerevan, and Baku. In January 2024, the U.S. Federal Aviation Administration issued a safety alert for operators warning airspace users of potential jamming and spoofing of GNSS signals and associated effects, such as unwarranted ground proximity warnings (Federal Aviation Administration, 2024). Although air traffic being subjected to RFI is not a new phenomenon, the extent of observed disturbances warrants a more detailed analysis. Most large commercial aircraft do not solely rely on a GNSS as their exclusive navigation system, but are equipped with high-grade inertial reference systems (IRSs), other radio navigation equipment, or a combination thereof. However, even for those aircraft, a GNSS loss still causes nuisance warnings in the flight deck and thus increases the workload of pilots. Furthermore, a GNSS loss prevents the use of all procedures requiring the GNSS. Moreover, the impact of RFI on GNSSs in smaller and less well-equipped aircraft may be more severe, as they may require radar assistance from air traffic control (ATC). A study by EUROCONTROL estimates that “5% of traffic in these regions could, given current RFI levels, need special assistance” (EUROCONTROL, 2021).
As there is limited information available on the extent and evolution of this problem and the potential implications for aircraft operators and ATC, in this paper, we investigate the following aspects:
First, we start by identifying the RFI hotspots in Europe and the timely evolution of RFI affecting air traffic in those regions for February through December 2022. As the wide appearance of GNSS spoofing (i.e., the transmission of fake signals misleading GNSS receivers to estimate an incorrect location and/or time) only started to occur in the fall of 2023, in this analysis, the term RFI refers to GNSS jamming, not spoofing. The analysis is performed based on transmitted ADS-B parameters.
Next, we analyze the potential need for ATC support of less well-equipped aircraft. Based on an analysis of the distribution of aircraft types, equipage, and effect of jamming on aircraft, this study allows for a more detailed impact assessment on air traffic regarding the number of affected aircraft and the share of affected traffic in a given region.
Then, we analyze the effects of jamming on a specific commercial flight into an affected area. The investigation first focuses on operational aspects for aircraft operators and flight crews and the effects observed in the flight deck, followed by a more technical, in-depth analysis investigating the navigation system performance of the aircraft under the influence of jamming. The impact on the navigation system is then correlated with the observed cockpit effects and ADS-B data.
Finally, we discuss the potential impact of these jamming events on global and regional air navigation strategies regarding the evolution of aeronautical navigation systems.
2 BACKGROUND AND LITERATURE
This section provides an overview of the literature on the detection and localization of GNSS RFI events on the basis of ADS-B data. In addition, it presents background information on the effects of GNSS jamming on aircraft systems and concludes with a paragraph on strategies and roadmaps relying on the availability of GNSSs for future aeronautical navigation and associated procedures.
Following several reports on RFI affecting civil air traffic, the EASA published an SIB on March 17, 2022, and a revision on February 17, 2023, describing potential effects on aircraft and outlining suggestions for addressing the issue. The recommendations included the statement that “Air operators [..] should ensure flight crews and relevant flight operation personnel are aware of possible GNSS jamming and/or possible spoofing” (European Union Aviation Safety Agency, 2022). To do this, as well as to maintain a high level of safety in aviation in general, GNSS jamming incidents must be detected and localized in a timely manner. To this end, various detection methods have been described in the literature. For instance, Scaramuzza et al. (2015) recorded and subsequently analyzed the RFI situation in the Swiss airspace over a period of three years using measuring devices installed on board rescue and military helicopters. However, such large-scale measurement campaigns are resource-intensive and often do not allow for a real-time evaluation of an acute RFI situation. To address these limitations, ADS-B parameters such as the navigation accuracy category for position (NACp) and the navigation integrity category (NIC) are often used in the literature for both the detection and localization of RFI incidents. Figuet et al. (2022) detected RFI-affected aircraft by evaluating how transmitted NACp values change in zones of active jamming. Put simply, the NACp values degrade upon entry and increase when the aircraft leave an area affected by RFI. Darabseh et al. (2019) compared NACp values transmitted via ADS-B from flights affected by known RFI incidents with NACp values broadcast by unaffected flights. Similarly, Lukeš et al. (2020) employed pattern recognition techniques to detect RFI events on the basis of fluctuations in NACp values. RFI incidents can also be localized by means of the NIC parameter in ADS-B messages via mathematical models (Liu et al., 2022) or machine-learning methods (Liu et al., 2021). Finally, one can localize RFI incidents by analyzing gaps, i.e., outages, in ADS-B trajectory data by applying the power-difference-of-arrival method (Jonáš & Vitan, 2019) or convex optimization models (Liu et al., 2020b). A different approach for RFI detection was reported by Murrian et al. (2021). In their work, they used a GNSS receiver mounted on the International Space Station to detect and localize large-scale RFI sources from space. Receiver manufacturers are also addressing the problem of GNSSs being vulnerable to RFI by using different sensor fusion techniques and the two existing antennas to increase resilience against interference, as shown by, e.g., Joseph et al. (2023) and Blois et al. (2023).
Furthermore, jamming incidents can also be detected by pilots in the cockpit, as aircraft affected by RFI often display corresponding advisory and caution messages. However, as Osechas et al. (2021) showed, the effects of jamming may be subtle. When an aircraft starts to be affected by RFI, only a subset of all visible satellites may be affected and become unavailable for navigation. Initially, this might simply degrade the navigation performance but not trigger any alerts in the flight deck. In most cases, it is only after GNSS-based navigation has been lost altogether that the first warnings that appear, related to a failure to transmit the aircraft position via ADS-B (Fol & Felux, 2022). Navigation-related warnings tend to appear only after the aircraft has been subject to RFI for some period of time. However, short periods during which the navigation system may be able to track at least four satellites while otherwise flying through an area with active jamming, as well as differences in the hybridization of different navigation sensors in different aircraft, lead to different effects for different aircraft types. Although the effects differ, they all increase the workload and require attention of the flight crew and adherence to manufacturer recommendations and company standard operating procedures on how to address the warnings.
Apart from the immediate effects visible in ADS-B data and announcements in the flight deck, frequent unavailability of satellite navigation may also have some consequences for the evolution of civil air traffic and the navigation infrastructure necessary for efficient operations in general. In the Global Air Navigation Plan (GANP) of the International Civil Aviation Organization (ICAO), GNSSs are considered a key technology for developing and maintaining an efficient, safe, and interoperable future air navigation system (International Civil Aviation Organization, 2022a). In this context, the availability of GNSSs is considered an integral part for numerous elements of the Aviation System Block Upgrade (ASBU) framework defined on the global technical level of the GANP. Moreover, on the basis of the global strategy outlined in the GANP, the European Air Navigation Plan (International Civil Aviation Organization, 2016) and the European ATM Master Plan (Single European Sky ATM Research, 2022) consider GNSSs as the primary means for implementing PBN across Europe. To this end, the implementation of PBN has already been mandated by Commission Implementing Regulation (EU) 2018/1048, which requires exclusive use of PBN for air traffic management (ATM) and air navigation services by the year 2030, with the exception of CAT-II/III landing systems, such as instrument landing systems (ILSs) and GBAS landing systems. Conventional, land-based navigation systems, most likely in the form of a network of distance measuring equipment (DME) ground stations, will complement and support PBN operations and provide an alternative means for complying with PBN requirements, at least in continental airspace, especially in the case of GNSS outages. However, aviation will most likely become even more dependent on the availability of GNSS signals in the future.
3 DATA AND METHODS
This chapter provides an overview on the methods applied to determine the impact of RFI on civilian air traffic in Europe. In a first step, Section 3.1 explains how the ADS-B data set used in this paper was retrieved and processed. Subsequently, Section 3.2 illustrates how European geographical areas of interest subject to RFI were identified and selected, whereas Section 3.3 outlines the details of how we classified aircraft affected by RFI. Section 3.4 describes the methods used to analyze the data. Finally, Section 3.5 describes the parameters and observations obtained during a scheduled flight on an Airbus aircraft through an area with known RFI.
3.1 Data Set
This paper makes use of crowd-sourced and publicly available ADS-B data. For an observation period of February 1 to December 31, 2022, position and status information messages were collected from the OpenSky Network (OSN) database (Schäfer et al., 2014) processed with the traffic library (Olive, 2019). As such, aircraft position information was obtained from the state_vectors_data4 table, whereas the NACp values transmitted by the aircraft were derived by decoding ADS-B raw messages from the operational_status_data4 table. Subsequently, the resulting data sets were merged via both the icao24 code of the aircraft, which is a unique identifier assigned to an aircraft, and the timestamp of each message. For the analysis in this paper, data for 1,364,170 flights were downloaded and analyzed. It should be noted that because the data originate from a crowd-sourced and ground-based receiver network, coverage may be somewhat limited and the observed flights may thus be a subset of all flights that were actually operating in the area.
3.2 Selection of Areas of Interest in Europe
Because of the large amount of data available, we restricted the analysis to three geographical regions of particular interest explicitly mentioned by EASA in their SIB (European Union Aviation Safety Agency, 2022). Subsequently, we consulted the web page http://www.gpsjam.org/, provided and hosted by Wiseman (2022), to cross-check and define areas of interest (AoIs) for further analysis. This website provides an overview of the daily aggregated number of flights affected by RFI on the basis of ADS-B data. It should be noted, however, that this website is based on crowd-sourced ADS-B data from a terrestrial network. Hence, not all areas in which air traffic is actually affected by RFI are necessarily identified as such because of a lack of coverage of ADS-B ground receivers. In the course of this work, three different AoIs were selected for the more detailed analysis presented in this paper. The first area, referred to as AoI-1, extends from the southern tip of Finland through the Baltic states of Estonia, Latvia, and Lithuania, the Russian exclave of Kaliningrad, and Poland into the northern half of the Czech Republic. In the east–west dimension, AoI-1 reaches approximately from the German–Poland border to Belarus. AoI-2 is situated mainly around Romania, Bulgaria, Moldavia, and the western part of the Black Sea. It also covers part of the Ukrainian airspace; however, no civil traffic was operating in that region during the period investigated. Especially since the closure of the Russian airspace to most European and Asian airlines, safe and efficient flight operations in AoI-2 have been of high interest, as a significant share of the air traffic between Asia and Europe now crosses this sector. Finally, AoI-3 is located in the airspace around Cyprus in the eastern Mediterranean Sea and parts of southern Türkiye. AoI-3 has been known to be affected by RFI for several years and has already been studied in other scientific papers, e.g., as reported by Osechas et al. (2022) and Fol & Felux (2022). The geographical dimensions of the selected AoIs are shown in Figure 1.
AoIs for which ADS-B data were collected
3.3 Detection of Flights Affected by RFI
To detect aircraft subject to GNSS RFI, the method described by Figuet et al. (2022) is adopted in this paper. Based on the NACp value sent via ADS-B messages, this method aims to determine whether an aircraft is jammed. As shown in Table 1, the NACp value provides an indication for the EPU determined by the navigation system of an aircraft. To this end, high NACp values suggest that the positioning accuracy is high, whereas low NACp values indicate the opposite.
NACp and Corresponding EPU Values From EUROCAE (2010)
Darabseh et al. (2019) and Lukeš et al. (2020) suggested that a degradation from NACp values of 8 or above to a value of 0 occurs when an aircraft is affected by Global Positioning System (GPS) jamming. By the same logic, if the NACp value transmitted by an aircraft increases from 0 to a value of 8 or more, an aircraft can be assumed to have left a zone of active RFI, with the GNSS again being used as the primary means of navigation. Consequently, in this paper, aircraft are labeled as being affected by GNSS RFI based on changes in NACp values transmitted within the AoI. We consider an aircraft in one of the AoIs to be jammed (i) if it transmits a NACp value of zero cumulatively for more than 1 min while within an AoI, as an indication that the aircraft is subject to jamming, and (ii) if it transmits a NACp value greater than 7 cumulatively for more than 1 min within the entire period of observation in order to verify proper operation of the equipment when not affected by RFI. By requiring both conditions (i) and (ii) to occur for at least 1 min each, outliers in the data set can be excluded. The value of 1 min was chosen to ensure that a negative impact on navigation solutions due to effects other than RFI would not lead to false classifications as aircraft being subject to RFI. Such a short-term degradation could arise, e.g., from aircraft maneuvering such as that reported by Liu et al. (2023). However, we recognize that the choice of the minimum time for which the NACp value must be zero may lead to some missed detections if the RFI affects an aircraft only for a short period of time. Furthermore, some differences exist between aircraft types and even within different avionics within the same aircraft type regarding the exact behavior during RFI, recovery after leaving a zone of RFI, and the data fed into the ADS-B. For example, Fol & Felux (2022) found that the degradation of NACp is slower and more gradual for the Airbus A350 than for the A321 when flying into an RFI-affected area. In this work, we chose to base the RFI detection on NACp, whereas Liu et al. (2020a) based their method on NIC. While NIC is transmitted in the same message as the position, NACp is transmitted in a different message and may thus have a slight delay until the message is received and decoded. However, the NACp value is still transmitted even when no valid position (or associated NIC) is transmitted via ADS-B because of jamming.
3.4 Analysis of RFI Activities in AoIs
This section describes the evaluations performed to determine whether there are any clear patterns in jamming activity, as well as the impact of jamming on different aircraft. More specifically, we performed (i) an analysis of the number and share of flights affected by RFI as well as the timely evolution of RFI occurrences, (ii) an analysis of aircraft-type-specific influences of RFI activities, and (iii) an analysis of aircraft equipage-specific influences of RFI activities.
Once specific portions of flights are identified as being affected by RFI, the impact of jamming activities on air traffic can be investigated further. Most simply, the cumulative number of flights likely affected by RFI per day over the period of observation is determined. Because the cumulative number of affected flight movements depends on both the geographical extent of the AoI and on the amount of traffic passing through that region, we also considered the relative share of flights affected by jamming on a day-by-day basis. Moreover, to identify potential patterns in the timely evolution of GNSS interruptions, we fine-screened the entire observation period by analyzing the number of affected aircraft in 30-min intervals.
As mentioned above, the reaction of an aircraft to RFI depends heavily on the architecture of the on-board navigation sensors, as well as other factors such as antenna placement on the fuselage and aircraft orientation toward the RFI source. Hence, we investigated which types of aircraft were affected by RFI and how the aircraft types relate to the observed flights showing a degradation in NACp within the selected AoI. The aircraft type of each flight was identified via the OpenSky aircraft database, which maps the unique icao24 code of each aircraft to an aircraft type. This information was used to compare the traffic share of all flights passing through the AoIs and jammed flights with respect to the aircraft type, helping to potentially identify which aircraft types are more robust to RFI.
RFI activities can affect aircraft differently. For example, aircraft that exclusively rely on a GNSS for navigation will most likely be more affected by RFI activity than aircraft that are equipped with a number of different navigation sensors. In the case of RFI, such aircraft may require ATC assistance, thereby increasing the workload of controllers. In the case of multiple aircraft needing assistance simultaneously, this may reduce the capacity of ATC sectors. When filing a flight plan, operators must indicate the navigation equipment they carry on board in accordance with ICAO’s Procedures for Air Navigation Services - Air Traffic Management (International Civil Aviation Organization, 2022b). Among others, operators must indicate whether the aircraft is equipped with DME, GNSS, and/or inertial navigation equipment. For this study, we considered aircraft that indicated GNSS equipage but neither inertial navigation nor DME capabilities, as these aircraft are more likely to lose navigation capability and require assistance in a zone of active RFI.
3.5 Effects of RFI on a Specific Flight
To gain a deeper understanding of the impact of RFI on flight planning, flight operations in general, and the workload of flight crews in particular, one of the authors of this paper accompanied a commercially operated flight to Beirut, one of the locations known to be heavily affected by GNSS interference, with an approach that traverses AoI-3. The observed flight was carried out on an Airbus A321 aircraft in July 2022. Both the ground flight preparations and the flight execution in the cockpit of the aircraft were observed. To document the effects of RFI on the flight under consideration, an event log was created in which all warning messages generated by the aircraft and crew actions relating to GNSS interference were recorded. Additionally, a commercially available u-blox EVK-M8N GNSS evaluation kit capable of tracking GPS, Galileo, and GLONASS was placed in the cockpit. To more precisely assess the effect of RFI on the observed flight, the airline’s flight data monitoring (FDM) department provided the authors with a log containing the following parameters:
IRS 1+2 latitude/longitude (positions of two IRSs2),
GPS 1+2 latitude/longitude (positions of the two GPS receivers),
Radio navigation position (latitude/longitude) from conventional navigation aids, typically multiple DMEs,
Flight management guidance computer 1+2 latitude/longitude (positions determined by the two flight management and guidance computers), and
GPS primary yes/no.
Finally, events observed in the cockpit and in the FDM data set were related to the ADS-B data of the accompanied flight. For analysis of the ADS-B data, the operational status data containing, amongst others, the NACp, SIL, and NIC parameters, as well as the transmitted position messages by the aircraft, were downloaded from the OSN. The aircraft was considered to be under the influence of RFI if the status data either begin to fluctuate between high values and 0 or remain at 0. Moreover, once the aircraft determines that the position source is not sufficiently reliable or does not meet the requirements for ADS-B position transmission, it stops broadcasting the position information and the associated NIC. In this case, however, the transmission of the status data (including NACp and SIL) continues, allowing for further evaluation of these parameters.
4 RESULTS
In the following, the results are presented. Section 4.1 focuses on both the localization of RFI activities in each AoI and the determination of the number of flights affected by jamming. Then, Sections 4.2 and 4.3 illustrate which aircraft types are affected by GNSS jamming and investigate the aircraft navigation equipage. Finally, Section 4.4 presents the observations made in the cockpit during a flight to Beirut that was affected by jamming. This section further presents a post-flight analysis of the recorded navigation data obtained via flight data monitoring.
4.1 Number and Share of Affected Aircraft
For the selected observation period from February 1 to December 31, 2022, data for 780,929 flights were collected for AoI-1 (Baltic states and Kaliningrad area), of which 2,709 were identified as being impacted by RFI. Similarly, data for 432,290 flights were collected for AoI-2 (Romania, Bulgaria area), of which 27,911 were identified as impacted by RFI. Finally, in AoI-3 (Cyprus area), data for 150,951 flights were collected, of which 57,803 were identified as being impacted by RFI. Figure 2 depicts the percentage of flights affected by RFI for the selected observation period and for each of the AoIs in hexagonal bins. For AoI-1, the area most affected by RFI is the region around Kaliningrad, the southern half of Lithuania, and the northeastern part of Poland. Another area that shows increased levels of jamming is located over the Baltic Sea, between Latvia and Sweden. In AoI-2, the most significantly affected area extends from over the Black Sea off the coast of Bulgaria in a northwesterly direction toward central Romania. High levels of jamming are also observed for flights into Moldova. For AoI-3, high levels of jamming activities were observed, especially in the area from southern Türkiye, along the coast of Syria and Lebanon, up to the Israeli border. Throughout AoI-3, a decreasing impact of RFI from east to west can be observed.
Percentage of flights affected by RFI per hexadecimal bin between February and December 2022 a) AoI-1 b) AoI-2 c) AoI-3
When comparing the percentage of flights jammed over the whole observation period, it can be observed that the maximum percentage of flights affected by RFI varies significantly across the three AoIs. Indeed, the maximum share of affected flights is only 1.2% in AoI-1, approximately 14% in AoI-2, and over 70% in AoI-3. To put these numbers into context, Figure 3 shows the total number of flights affected by jamming in each AoI, along with the relative share of flights within the AoI affected for each day. For AoI-1, the largest peak occurred in early March, when approximately 25% of all flights were affected. The maximum share of affected aircraft reached just over 50% in AoI-2 and reached almost 80% in AoI-3. For a more in-depth analysis, Figures 4, 5, and 6 depict the intensity of RFI activities, quantified by both the relative and absolute number of aircraft labeled as being jammed, in 30-min intervals over the entire observation period.
Daily number and percentage of flights affected by RFI for all AoIs
Jamming activity for AoI-1 over 30-min intervals covering the entire observation period
The color of each cell corresponds to the relative and absolute number of flights impacted by RFI within the 30-min interval. a) Relative share of flights subject to RFI b) Absolute number of flights subject to RFI
Jamming activity for AoI-2 over 30-min intervals covering the entire observation period
The color of each cell corresponds to the relative and absolute number of flights impacted by RFI within the 30-min interval. a) Relative share of flights subject to RFI b) Absolute number of flights subject to RFI
Jamming activity for AoI-3 over 30-min intervals covering the entire observation period
The color of each cell corresponds to the relative and absolute number of flights impacted by RFI within the 30-min interval. a) Relative share of flights subject to RFI b) Absolute number of flights subject to RFI
In AoI-1, a period of intensified jamming activities can be clearly identified from March 4 to March 6, 2022. This phase is followed by a 5-day period with less intense interference. Apart from a half-day impact in February, only isolated RFI events are observed in AoI-1 for the remainder of the observation period.
For AoI-2, segregated events were identified until mid-May 2022. A period with a relatively high intensity of RFI activity was observed between May 17 and May 24, followed by the most intense activity from June 2 to June 19. In that period, up to 80 flights per 30-min interval were identified as affected by RFI. Afterward, only mild activity is observed until the end of the observation period.
In AoI-3, the RFI activity is somewhat constant through the entire observation period, with up to 20 flights being affected by RFI within the 30-min intervals.
Figure 7 presents a box plot of the number of simultaneous flights affected by RFI when at least one flight was identified as affected. For AoI-1, slightly less than two flights were affected by RFI at the same time on average, compared with three flights for AoI-3 and six flights for AoI-2. It is important to note that the number of flights affected by RFI can vary depending on the jamming intensity, the amount of traffic, the geometric dimensions of the AoI, and the ADS-B coverage in an AoI.
Box plot of the number of flights affected simultaneously by RFI for each AoI at times when RFI was detected The solid vertical bars indicate the minimum, first quartile, median, third quartile, and maximum number (from left to right), whereas the dashed vertical line represents the mean value.
4.2 Aircraft-Type-Specific Behavior
To analyze potential aircraft-type-specific patterns, we identified and grouped the aircraft affected by jamming according to aircraft families. For each AoI, the blue bars in Figure 8 indicate the proportion of the 10 most frequently observed aircraft families in relation to all traffic observations. Moreover, the red bars in the same figure indicate the share of the aircraft families that were labeled as jammed. With a traffic share between 20% and 40%, aircraft of the Airbus A320 and Boeing B737 families accounted for most traffic in all three AoIs. In AoI-3, the Airbus A320 and Boeing B737 families have similar shares of the total traffic and similar percentages of jammed fights. In contrast, in AoI-1 and AoI-2, the share of jammed flights for the A320 and B737 families is smaller than their share of total traffic. Although for the most part, the traffic share of the total number of flights and the share of jammed flights are similar, in AoI-2, the B777 and B747 families account for significantly larger shares of the jammed flights compared with their share of the total number of flights. However, in AoI-3, this result for the B777 family could not be identified with the same significance. For the B747 family, there were not enough data available for a meaningful analysis. Finally, in AoI-1, the B787 and Embraer E195 families account for a significantly smaller share of the number of jammed flights compared with their share of the total flights. For the B787 aircraft, a similar trend can be observed in AoI-3 but not in AoI-2.
Comparison of aircraft-type share between the total traffic and RFI-impacted flights a) AoI-1 b) AoI-2 c) AoI-3
4.3 Aircraft Equipage to Determine Potential Impact for ATC
Finally, information about the navigation equipment available on board each aircraft was retrieved from flight plan data. Based on the data available, the aircraft type could be identified for approximately 90% of the flights in the data set. Particular attention was given to those aircraft that indicated being equipped with GNSS navigation but not with an IRS or DME. Table 2 shows for each AoI the number and aircraft types of flights impacted by RFI with this combination of equipment according to flight plan data. A total of 5 flights affected by RFI and having GNSS as the only navigation solution were identified in AoI-1, with 13 in AoI-2 and 43 in AoI-3. The types of aircraft indicate that business jets such as Learjet 35 or Embraer Phenom 300 make up a large proportion of these identified flights.
Number of Occurrences of Flights Affected by RFI for Which No Inertial or Terrestrial Navigation Capabilities Were Reported
4.4 Effects on a Specific Flight
This section presents the results of an observation flight on a mainline aircraft, an Airbus A321, to Beirut. The results presented in this section were anonymized according to data protection requirements by the operating airline; however, all relevant data for the analysis were made available.
4.4.1 Flight Preparation and Cockpit Effects
The crew was made aware of potential interference in the eastern Mediterranean area in the pre-flight briefing package, and this issue was briefly discussed. The crew was familiar with the potential effects of a loss of GPS, and no operational consequences were expected. The entire flight was planned through Area Navigation 5 (RNAV5)3 airspace. In the FIRs of Nikosia and Beirut, a GNSS is not mandatory, and conventional navigation aids are available to ensure that the required navigation performance can be met. For landing in Beirut, an ILS approach was planned, requiring no GPS.
During the flight, a number of potential influences of RFI were observed in the cockpit. The first caution message of type NAV ADS-B RPTG 1+2 FAULT was announced to the crew when the airplane traversed Cyprus. This message indicates that the required navigation performance for transmitting the aircraft position via ADS-B was no longer sufficient. According to the announcements on the flight deck, the navigation mode remained in GPS PRIMARY and the navigation accuracy was shown as high, indicating that the aircraft determined that the navigation requirements for the airspace in which it was operating were met. The two cautions were canceled by the flight crew with the emergency cancel button, which silences these cautions for the remainder of the flight. If not canceled, these cautions reappear regularly under the influence of RFI, which is a potential source of distraction for flight crews. In addition, the aircraft clock source was switched from GPS to internal. No further actions were performed by the flight crew. According to the GPS monitoring page on the aircraft’s multifunction control display unit, GPS 1 showed no position. Furthermore, the display indicated that GPS 1 was in AIDED mode, meaning that the GPS 1 position was aided by the aircraft’s inertial sensors and no standalone GPS position could be determined. GPS 2 was observed to switch back and forth between AIDED mode and NAV mode a few times, indicating that GPS 2 was occasionally able to determine a GPS-based position solution, alternating with periods during which inertial navigation was necessary. As the flight progressed, both GPS receivers remained in AIDED mode until shortly before landing with varying numbers of tracked satellites. In total, ADS-B was inoperative for approximately 20 min. However, the indicated navigation accuracy remained high for the whole flight. During the arrival, the message GPS PRIMARY LOST was displayed on the captain’s navigation display, indicating that the aircraft could no longer ensure that the actual navigation performance was sufficient to comply with the required performance in the airspace where the aircraft was operating. Nevertheless, the approach was continued, as GPS was not required for the arrival procedure or the ILS approach. The GPS recovered shortly before landing at approximately 700 ft above mean sea level.
4.4.2 ADS-B Data Analysis
This section shows the results of the RFI analysis specifically for the observer flight into Beirut. The flight path based on the OSN-sourced ADS-B messages is depicted in Figure 9. In this data set, the geographical area where RFI was experienced could be clearly identified from the status data. The time period during which position information was received but the NIC parameter dropped to zero because of potential RFI is marked in red and starts at 09:04 Universal Coordinated Time (UTC). The blue dashed line is an interpolated track for which no position information was contained in the ADS-B message. The GNSS outage ends at 09:28 UTC, when the ADS-B messages again contained position information.
Flight track of the observation flight into Beirut
The solid black line represents normal operation, whereas the red line depicts the flight track for the time period during which NAC, NIC, and SIL dropped to 0 but position information was still transmitted. The blue dashed line is an interpolation for when no position information was available.
Some of the events observed in the cockpit, i.e., the caution messages described above, could be clearly correlated with the reported values for NACp, NIC, and SIL. The corresponding NACp, NIC, and SIL values from the flight are plotted in Figures 10, 11, and 12, respectively. Within these plots, the observed time of the NAV ADS-B RPTG 1+2 FAULT, GPS PRIMARY LOST, and the recovery of the GPS navigation indicated by the GPS PRIMARY events are marked.
NACP value
The first red dashed line indicates the NAV ADS-B RPTG1+2 FAULT, the second dashed red line indicates the GPS PRIMARY LOST message, and the green dashed line indicates the recovery to GPS PRIMARY.
NIC value
The first red dashed line indicates the NAV ADS-B RPTG 1+2 FAULT, the second dashed red line indicates the GPS PRIMARY LOST message, and the green dashed line indicates the recovery to GPS PRIMARY.
SIL value
The first red dashed line indicates the NAV ADS-B RPTG 1+2 FAULT, the second dashed red line indicates the GPS PRIMARY LOST message, and the green dashed line indicates the recovery to GPS PRIMARY.
As indicated in Figure 10, the NAV ADS-B RPTG FAULT message is immediately triggered as soon as the NACp value drops to zero. The same is true for the NIC and SIL values. Furthermore, it was observed that at the beginning of the RFI event, one GPS receiver switched between NAV mode and AIDED mode a few times. In addition, a fluctuation in the NACp value is visible in the data during the same time interval. The GPS PRIMARY LOST message could not be related to any specific event in the ADS-B data. When the aircraft was on approach and descending through approximately 1000 m, the integrity parameters of the ADS-B message all reverted to high levels, consistent with the aircraft leaving RFI influence.
4.4.3 Aircraft Navigation Data and Commercial Receiver
Following the flight, the operating airline provided the recorded position data through their FDM department. Additionally, the position estimates of a commercial handheld u-blox multi-constellation receiver (GPS+Galileo+GLONASS) were used, with the antenna and receiver placed inside the cockpit during the flight. The number of tracked satellites per constellation is shown in Figure 13. The effect of RFI can be clearly identified, starting shortly after 09:00 UTC and lasting until approximately 09:25 UTC. Initially, the number of satellites observed by the GPS receivers starts fluctuating between 8 and 6. After 09:15 UTC, the number of satellites begins to decrease more significantly to a minimum of just one satellite. For several minutes, the GPS receivers do not reliably track four or more satellites. The same applies for Galileo, for which the number of tracked satellites drops below four after 09:15 UTC, even reaching zero for a short period of time. In contrast, the number of tracked GLONASS satellites declined after 09:05 UTC from 7 to 5. This observation is consistent with a narrow-band jammer on the L1 center frequency having a weaker affect on GLONASS because of the use of slightly different frequencies in the frequency division multiple access scheme, compared with the code division multiple access scheme used in GPS and Galileo, as described by Osechas et al. (2022). As there were always sufficient satellites available for a position estimate, the GLONASS position is used as a reference. Figure 14 depicts the evolution of the distances between different position sources, with the distance between the positions determined by the GPS 1 receiver of the aircraft and the u-blox receiver shown in blue. The difference remains on the order of a few meters until approximately 09:04 UTC, when the NACP value dropped to zero. From that time point on, the distance increases over a period of approximately 25 min to a maximum of 2250 m. Shortly after 09:27 UTC, the distance suddenly drops again to just a few meters. This 23-min period corresponds to the time when the GPS receiver of the aircraft was in AIDED mode. This result is consistent with the distance between the GPS 1 and IRS 1 positions, shown in grey in Figure 14. During the first part of the flight, the distance varies between 1000 m and approximately 3200 m with a slow rate of change, consistent with the typical drift of an IRS. During the time when the aircraft was subject to RFI and the distance between the GPS 1 and u-blox positions increased, the distance between the position solutions of IRS 1 and GPS 1 remained relatively constant. After the aircraft reacquired its GPS position solution, at approximately 09:27 UTC, the difference jumped to a value of approximately 1500 m. The orange line in Figure 14 depicts the distance between the positions of the u-blox receiver and IRS 1. Until approximately 09:04 UTC, this distance primarily overlaps with the grey line (distance from GPS 1 to IRS 1), which is consistent with the small distance between the GPS 1 and u-blox positions. During the time period of RFI between 09:04 UTC and 09:27 UTC, the IRS–u-blox distance continues to evolve in a way that is consistent with both the behavior observed during the first part of the flight and the normal drift of an IRS. After reacquisition of the GPS 1 position, the grey and orange lines again show a high degree of overlap. It should be noted, however, that the flight data obtained did not indicate whether the GPS 1 position was aircraft-internally labeled as “Normally computed.”
Number of tracked satellites per constellation.
Distance between position estimates from different sensors The time when the GPS PRIMARY LOST message was displayed is indicated in red.
Figure 15 compares position solutions originating from the GPS 1 receiver, IRS 1, and the u-blox receiver. Furthermore, the positions at which the GPS PRIMARY LOST warning message was announced to the flight crew, as well as the position at which the position source reverted back to GPS as the primary mode, are indicated. Initially, the three position solutions show a distinct difference on the order of >1 km. For the distance between the GPS 1 and IRS 1 solutions, this difference remains consistent with the expectations regarding the drift of the inertial sensors. The difference between the GPS 1 and GNSS-based u-blox position solutions, however, is initially large until the aircraft is able to recover its GPS 1 position and again use GPS as the primary navigation source. For the remainder of the final approach, the GPS 1 and u-blox position solutions show good agreement.
Three different position sources during the final approach: GPS 1 (green), IRS 1 (red), u-blox (orange)
5 DISCUSSION
The analysis results showed that a significant number of flights in the European airspace were affected by RFI in the form of GPS/GNSS jamming. As the absolute number of affected aircraft alone depends on the size of the area under investigation and the amount of traffic flowing through, the numbers were normalized by the total number of aircraft. In this way, the share of flights affected by jamming could be identified. We discretized the AoIs into smaller hexagons to eliminate (to a large extent) the dependency on the chosen size of the AoI. Finally, a dependence on the observation period yields different shares of the total percentage of traffic affected if jamming occurs sporadically. The AoIs correspond to areas that were mentioned in EASA SIBs and where a sufficient number of aircraft encountered and reported navigation issues in order to raise concern. However, many small-scale interference events also take place on a regular basis, e.g., those caused by small jammers in cars. These events might have an affect on only a small number of aircraft and/or for only a short period of time. Whereas one aircraft meeting our criteria of being labeled as RFI-impacted may be the result of an equipment issue on board a single aircraft, more than one aircraft experiencing issues in the same area can be regarded as a strong indication for the presence of RFI and may be used to trigger further action for localizing and shutting down a potential source, if possible. An issue with detecting small-scale events based on ADS-B data may be the required effect on the navigation solution: if only some satellites are lost, the effect may be small enough that no drop in the ADS-B integrity parameters is triggered.
For a more detailed analysis of periods when jamming was observed, we discretized the observation period into 30-min intervals. Thereby, different days with high levels of RFI could be identified for the different regions. The periods with high RFI seemed to follow no particular pattern regarding the day of month, day of week, or time of day. The periods with high levels of RFI differed among all three AoIs, without any obvious similarities.
One general limitation of identifying regions with ongoing RFI based on ADS-B data is the dependence on air traffic operating in that region. If no air traffic is operating (e.g., over crisis areas), it is not possible to determine whether RFI is present in the area. Additionally, the method depends on a suitable network of ADS-B receivers to receive the signals transmitted by the aircraft. When crowd-sourced data from a terrestrial network of receivers are used, coverage may be limited, especially over water (e.g., the Black Sea) and in areas where it may not be possible to operate receivers and feed a network. Significant improvements to coverage can be achieved when monitoring ADS-B data from satellites. Space-based GNSS receivers can also be used to directly identify affected areas and even locate transmitters. Commercial services for both of those approaches are being operated, e.g., by Aerion and HawkEye 360, respectively. Finally, ADS-B data-based detection has the inherent problem that, if jammed, an aircraft will stop transmitting its position. Thus, extended periods of jamming bear the problem of receiving only the ADS-B status messages of the aircraft, without associated position information. As described by Osechas et al. (2022), it is also possible that GNSS receivers affected by RFI may not recover after the aircraft leaves the jamming area. In such cases, the aircraft could continue broadcasting low NACp values in areas that are not affected by jamming. Regarding differences between aircraft types, the shares of individual aircraft types of all flights roughly corresponded with the shares of individual aircraft types of all jammed flights. For some aircraft types, large differences between those shares were observed in one or two AoIs but never in all three AoIs. Hence, based on these results, no obvious relation between aircraft type and likelihood of being jammed could be determined. Due to different avionics and data sources used to feed the ADS-B, there may be differences in the reaction to RFI between different aircraft types as well as within aircraft of the same type. The analysis further showed that there were several flights that, according to flight plan data, were only equipped with GNSS but not with IRS or DME. These aircraft would likely require assistance from ATC. However, their number was rather small, indicating that the probability of an increased ATC workload in a given sector due to many aircraft requiring simultaneous assistance is rather small. Nevertheless, even a few simultaneously affected aircraft may lead to capacity issues in certain sectors and should be considered as a plausible scenario with appropriate procedures in place. To obtain a better understanding of the RFI effects on regular flight operations, as well as to link the observation of ADS-B data to actual warnings indicated to the flight crew, we compared observations from the flight deck, ADS-B data, and data from the operator’s FDM department. The first indication of RFI is announced to the crew as a loss of ADS-B reporting functionality, as previously presented by Fol & Felux (2022) based on the evaluation of pilot reports. The warning in the cockpit corresponds to the time when the NAC, NIC, and SIL parameters in the ADS-B messages drop to zero. Afterwards, the GPS position of the aircraft is aided by, and thus mainly based on, its IRS, showing a drift between the GPS position and the GNSS position of the handheld receiver in the cockpit that is consistent with the drift of an IRS. The GPS PRIMARY LOST message occurred several minutes after the aircraft apparently encountered RFI and is thus only a lagging indicator, consistent with the observations of Osechas et al. (2022). Upon leaving the area affected by RFI, the on-board GPS receivers recovered and resumed normal navigation. Additionally, the ADS-B transmissions resumed and transmitted position information and nominal NAC, NIC, and SIL values. For the time during which the aircraft was affected by RFI, the positions estimated by the flight control and guidance system remained close to the u-blox GNSS position, indicating that a combination of different alternative navigation sensors (most likely multiple DMEs in combination with the IRS) were used, providing a robust navigation solution. While not causing navigational issues, the encounter of RFI led to a slightly increased workload on the flight deck. The crew checked on the navigation system more regularly and had to handle the caution messages related to the ADS-B reporting and GNSS navigation faults. No impact on the conduction of the flight occurred, as the aircraft relied on conventional navigation systems whenever GPS was unavailable. For the period investigated during 2022, no widespread GNSS spoofing was ongoing. The detection metrics for labeling a flight as jammed specifically address the jamming case. If a flight is subject to spoofing, the ADS-B integrity parameters may likely not decrease in the same way as during jamming and thus require different modes of detection.
6 CONCLUSIONS AND OUTLOOK
This paper showed that GNSS jamming is an ongoing issue affecting a significant share of flights. While most aircraft have other alternatives for navigation (by either inertial sensors or conventional ground-based navigation aids), a GNSS outage generally produces alerts to the flight crews and increases their workload. Such conditions may not be a significant issue for en route flight; however, during times of high workloads, such as during arrivals and departures, and during more critical phases of flight, such alerts may distract the crew. From an operational perspective, the current impact of GNSS outages is minimal. This, however, may change with increasing reliance on PBN procedures foreseen in the road maps regarding modernizing the ATM system and making operations as efficient as possible. One key element in this navigation strategy is the use of GNSSs. To continue with the much-needed modernization of ATM to keep up with increasing traffic volumes, safe, secure, and reliable navigation is crucial. Thus, solutions to the GNSS jamming issue should be based on a variety of different measures. The recent observation of civil air traffic also being subject to spoofing further increases the need for measures to ensure continuous safe navigation. Suitable measures can include (i) increasing the robustness against RFI in on-board GNSS receivers, potentially by using advanced signal processing and antenna/receiver technology as well as multiple frequencies and constellations, (ii) increasing the integration of GNSS with other sensors and thus reducing the dependency on a particular navigation system, (iii) retaining a ground-based network of radio-navigation aids, such as DMEs and at least a minimum ILS network, with improved performance and the potential to support similar levels of accuracy and integrity as currently achieved by GNSS, and (iv) introducing measures to prevent harmful interference in the first place to the extent possible.
HOW TO CITE THIS ARTICLE
Felux, M., Fol, P., Figuet, B., Waltert, M., & Olive, X. (2024). Impacts of global navigation satellite system jamming on aviation. NAVIGATION, 71(3). https://doi.org/10.33012/navi.657
ACKNOWLEDGMENTS
Part of this work was funded by armasuisse W+T under contract number 047-29. Open access funding was provided by ZHAW Zurich University of Applied Sciences. The authors greatly appreciate the support!
Footnotes
↵1 There are exceptions in which DME/DME and inertial navigation are deemed sufficient for certain approved procedures.
↵2 The aircraft is equipped with three redundant IRSs; however, in the provided data set, only data from two systems were included.
↵3 RNAV5 implies that the total system error of the aircraft must remain below 5 nautical miles (NM) for 95% of the time.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Jump to section
Related Articles
Cited By...
- No citing articles found.