Impacts of ionospheric scintillations on GPS receivers intended for equatorial aviation applications



[1] This study examines the impacts of ionospheric scintillations on GPS receivers that are intended for equatorial or transequatorial aviation applications. We analyzed GPS data that were acquired at Ascension Island during the Air Force Research Laboratory (AFRL) campaign of the solar maximum year of 2002. Strong scintillations impacted the receiver-satellite geometry, leading to poor dilution of precisions and positioning accuracy. In addition, deep signal fades (>20 dB-Hz), leading to navigation outages were observed during most of the nights of the campaign. Under quiescent conditions, the C/No of satellites fluctuated slowly between 50 dB-Hz and 35 dB-Hz baselines for both L1 (1.5754 GHz) and L2 (1.2276 GHz) signals, depending on the satellite's elevation angle. The satellite's elevation angle and the effective scan velocity of the satellite's ionospheric penetration point (IPP) with respect to the magnetic field and plasma drift influenced the rate of fading of satellite signals.

1. Introduction

[2] The rapid expansion in global civil aviation is over-stretching the conventional sensor-based navigation infrastructure. Implementation of global navigation satellite system (GNSS) in the industry will improve performance, enhance safety, and as well save cost [Cabler and DeCleene, 2002]. GNSS is a generic name that encompasses the core constellations–GPS, GLONASS, upcoming Galileo, and augmentation systems–space based augmentation system (SBAS), ground based augmentation system (GBAS) and aircraft based augmentation system (ABAS). In this study, we have used GPS data as the representative of the response of GNSS to scintillation activity, because data from augmentation systems, such as SBAS were only integrated into the data collection system after the maximum of solar cycle 23.

[3] The major challenge that is delaying the implementation of GNSS for global aviation applications is ionospheric scintillation [Bandyopadhayay et al., 1997; Akala et al., 2011; Akala and Doherty, 2012]. Scintillation is the rapid fluctuations that are imposed on the amplitude and/or phase of radio signals that traverse irregularities in the Fregion of the ionosphere. It is a post-sunset event in the equatorial region, but can occur at any time of the day at the polar region. However, it is most predominant at the equatorial region, lesser at the high latitude and least at the midlatitude [Aarons, 1982; Basu et al., 1987, Basu and MacKenzie, 1988]. Scintillation is also known to be seasonal and solar activity dependent [DasGupta et al., 2004; Akala et al., 2011; Akala and Doherty, 2012], and it has correlation with geomagnetic activity [Ledvina et al., 2004; Basu et al., 2005, 2010]. Typically, scintillation is quantified by the S4 index (the standard deviation of the factor I/〈I〉 over a 60 s period, where I is the intensity of the received signal and 〈I〉 is its average value). It may cause amplitude fades in excess of 20dB-Hz on GNSS channels on active days [Aarons, 1982]. These fades could cause cycle slips, and stress the receiver to lose lock on the transmitted signals [Doherty et al., 2004], to be re-acquired at a later time [Carrano et al., 2005; Seo et al., 2009], leading to intermittent availability of service [Seo et al., 2011a, 2011b].

[4] The impacts of scintillation on the GPS tracking loop performance manifest by the fading being experienced by the signals of the scintillating satellites. Consequently, understanding the relationship between fading parameters and scintillation (scattering) environment in addition to the satellite's motion is very important. Rino and Owen [1980], with a further validation by Basu et al. [1987] showed the dependency of the decorrelation time on the Fresnel zone radius and the satellite's effective scan velocity under different scattering regimes.

[5] Dual frequency mechanism employs the combination of measurements at two frequencies (L1 and L2) to calculate and broadcast ionospheric, clock, and ephemeris corrections (and error bounds). Currently, dual frequency certified aviation GPS receivers onboard aircrafts are not yet available. However, SBAS reference station receivers operate on dual frequency mode. For now, SBAS avionics onboard aircrafts are L1-only receivers. For dual frequency mechanism, loss of lock on either L1 or L2 due to fading prevents these corrections from being calculated during the time period that lock is lost and even for a few seconds after lock is regained, thereby leading to navigation errors. Furthermore, GPS receiver loses lock on L2 signal more often than on L1 signal of the same satellite when traversing ionospheric irregularities [El-Arini et al., 2003, 2009; Strangeways, 2009]. Consequently, the present study concentrated on L1 signal data only.

[6] The central focus of the GPS modernization program is the addition of new navigation signals (L2C and L5) to the GPS constellation. The L2C signal will be dedicated for civilian applications, and it is expected to replace the current semi-codeless L2P(Y) signal. The combination of L1 and L2C signals will provide reliable ionospheric corrections that will improve navigation accuracy. The L5 on the other hand is exclusively reserved for aviation navigation services, and the frequency (L5) is designed with a protected spectrum, higher power, and greater bandwidth to support life-critical and high-performance applications [United Nations-Office for Outer Space Affairs (UNOOSA), 2010].

[7] For future aviation, GNSS will use dual frequency civilian codes (L1 and L5:1.17645 GHz frequencies) [Walter et al., 2008; Hegarty and Chatre, 2008]. The frequency diversity scheme at L1 and L5 frequencies is expected to mitigate scintillation impact on GPS aviation availability. For instance, if a receiver tracks both L1 and L5 signals from the same satellite, it can rely on one frequency to provide range measurements [Gherm et al., 2011; Seo et al., 2011a], even when it briefly loses the other frequency to deep fading. However, the overall scintillation robustness of this dual frequency mechanism will be ascertained during the upcoming solar maximum in 2012 and 2013 when combination of both L1 and L5 signals data, under severe scintillation conditions will be available to investigate the extent of their correlations.

[8] Furthermore, during the next solar maximum, interests will be geared toward testing of aviation GPS receivers for general tracking robustness under scintillations. Before ultimate confrontation with reality through field testing under severe scintillation conditions, GPS receivers are usually designed and subjected to bench tests via modeling and simulations to ascertain their capabilities [Hegarty et al., 2001; Conker et al., 2003; Humphreys et al., 2009, 2010a, 2010b], but these testing strategies can give misleading results if the scintillation time histories that are used for such studies are not realistic. For example, in field testing at Ascension Island during the solar maximum years of solar cycle 23, Bishop et al. [1998], Groves et al. [2000] and Ganguly et al. [2004] observed receiver performance degradations that were much worse than those anticipated by the simulations that were conducted prior to the campaign. This justifies the need for an archive of real equatorial scintillation data, on which scintillation time histories could be built for bench tests.

[9] The goal of this study is to analyze the responses of GPS satellite signals to equatorial scintillations for the betterment of modeling and simulations, and for the overall improvement of aviation GPS receiver architecture, especially those that are intended for equatorial and transequatorial applications. We analyzed the entire data for the period of the campaign, detailing the analysis of a worst night data (13 March 2002) of the campaign, and a comparative analysis of a less active night data (9 March 2002), hereafter referred to as ‘active night’ and ‘less active night’ respectively. The worst night was characterized by intense scintillation activities that caused the receiver to view less than four satellites at times. By implication, the night was also characterized with ionospheric irregularities. On the other hand, the less active night recorded minimal scintillation activities.

2. Observations

[10] The U. S. Air Force Research Laboratory (ARFL) conducted a 15-day (5–19 March 2002) campaign to monitor GPS scintillations at Ascension Island (7.96°S, 14.41°W, dip lat 16.0°S) during the solar maximum year of 2002. Several GPS receivers were used to acquire data during the campaign. However, the data acquired by the Ashtech Z-XII receiver at a sampling rate of 20Hz were used for the present study. In addition, using data from an earlier campaign (March 2000) at the same station,Groves et al. [2000]concluded that Ashtech Z-XII receiver is more robust in terms of amplitude and phase scintillations monitoring capabilities. Further, this receiver has the capability of reporting the health status of signals from all the satellites in its view. To maintain quality control as the receiver estimates the navigation solution, signals from satellites with corrupted pseudoranges due to scintillation impacts are often rejected by the receiver.Carrano et al. [2005] and Carrano and Groves [2010] had earlier analyzed the present data set to investigate signal availability under severe scintillations, and temporal decorelation of GPS satellite signals due to multiple scattering from ionospheric irregularities.

[11] Intense scintillation activity (S4∼1.0) was encountered on almost all the nights of the campaign and severe impacts on GPS tracking performance in terms of signal availability and positioning accuracy were consistently observed. It is important to mention that Ashtech Z-XII receiver is not an aviation certified receiver. In fact, the authors are unaware of scintillation tests performed with an aviation certified receiver before or during the maximum of solar cycle 23. However, based on its scintillation robustness, the data collected by the Ashtech Z-XII receiver is representative enough to help in defining preliminary investigations on the impacts of scintillations on aviation GPS receivers.

[12] During this campaign, a number of navigation outages were observed. A navigation outage occurs when a GPS receiver has fewer than four satellites in its view to estimate the navigation solution. Under such condition, absolute zero is recorded as the position solution by the receiver. From the receiver reported position dilution of precision (PDOP) and time dilution of precision (TDOP), we calculated the geometric dilution of precision (GDOP) by the conventional method: GDOP = [PDOP2 + TDOP2]1/2. A GDOP and PDOP that is higher than 6.0 will not guarantee accurate navigation solution [Kaplan, 1996; Misra and Enge, 2001; UNOOSA, 2010]. Almost all the nights of the campaign were characterized with strong scintillations, with up to five or more satellites signals scintillating at times. Generally, navigation outages were observed during all the nights of the campaign except five; 5th–6th and 9th–11th. The durations of the outages range from 1 to 50 s, although less than 10 s most of the time. An outage that lasted for 25 s was observed on the 7th, and two longer ones that lasted for 50 s each were observed on the 13th and 19th. The observed outages were generally localized between 2100 and 2300 UT of the nights. This observation is in agreement with the conclusion of Akala et al. [2011], which reported that intense scintillations, which impacts negatively on navigation performance are usually localized within the hours of 2000–2300 LT. For this reason, the data analyzed for each night of the campaign were those observed at 2000–2400 LT (UT = LT at Ascension Island).

[13] A GPS receiver maintains carrier lock when the fluctuations of its carrier-to-noise density ratio (C/No) are localized within the upper (50 dB-Hz) and lower (30 dB-Hz) threshold bounds, otherwise loss of tracking lock may be imminent (although this is receiver dependent). A deep fading in this paper is defined as a fade with a depth of 20 dB-Hz or more. The fade samples were determined by selecting the samples that fall below the specified C/No threshold values (30, 25, 20, 15, 10 dB-Hz).

[14] In a previous study by Seo et al. [2009]using 45 min data from an earlier campaign of March 2001 at Ascension Island, the statistical analysis of the signals from all the PRNs was mutually considered. Nevertheless, we believe that the signal from each satellite traverses ionospheric irregularities with characteristic path length, depending on the satellite's elevation, and as such, responds uniquely to ionospheric perturbations. Consequently, characterization of the extent of degradation of the signals that is caused by scintillation on satellite-by-satellite basis will provide useful information that could assist in improving GPS receivers tracking intelligence, especially under disturbed ionospheric conditions.

[15] On the active night, six outages were observed, with durations ranging from 1 to 50 s, while no outage was recorded on the less active night. The fading durations of the signals from scintillating satellites were determined and analyzed. In this paper, we defined fading duration as the time between when the C/No falls below 30 dB-Hz threshold level and when it recovers out of it. The choice of 30 dB-Hz threshold level in this work was due to the fact that most receivers are designed to maintain lock on signals at a lower threshold bound of 30–32 dB-Hz. The fading durations were determined by selecting the entering and exit points of each fade at a 30 dB-Hz threshold level. Thereafter, the time interval between the two points was estimated, but a fade that has not recovered above the threshold level, 10 s or more after its entering point has been identified is assumed to have been lost and discarded from the samples. Fading durations could as well be considered at other defined threshold values.

3. Results

[16] Ionospheric irregularities impact the C/No values of satellites, causing them to fluctuate rapidly. The extent of these modulations on a satellite signals depends largely on whether the signals from the satellites traverse a patch of ionospheric irregularities in the sky or not. In a situation where the satellite signals traverse a region of the sky where irregularities are not present, such signals will not suffer modulation even when the night is characterized with scintillation events. Figure 1a shows the C/No of a scintillating satellite (PRN 27) during a 30 min time window on the active night, while Figure 1b shows the corresponding S4 index and the satellite elevation angle trajectory (20–31°) during the period. Similarly, Figure 2ashows the C/No of a non-scintillating satellite (PRN 28) during the same period, whileFigure 2b shows the corresponding S4 index and the satellite elevation angle trajectory (35–37°) during the period.

Figure 1.

(a) Carrier-to-noise density ratio of a scintillating satellite during half an hour slide window of the active night (13 March 2002). (b) The corresponding S4 index and the satellite's elevation angle trajectory during the period.

Figure 2.

(a) Carrier-to-noise density ratio of a non-scintillating satellite during half an hour slide window of the active night (13 March 2002). (b) The corresponding S4 index and the satellite's elevation angle trajectory during the period.

[17] Figures 3a–3e show the number of satellites that maintained lock on L1 signals, the position dilution of precision (PDOP), the geometric dilution of precision (GDOP), the vertical position error, and the horizontal position error, respectively, on the less active night, while Figures 4a–4e show the same parameters on the active night. The vertical and horizontal position errors were evaluated from the receiver reported position data. Using standard technique by Leick [2004] and Hofmann-Wellenhof et al. [1997], the reported position data in world geodetic system (WGS-84) reference frame were converted to the Earth Centred Earth Fixed (ECEF) Cartesian coordinates. Thereafter, the median value of all the position values during the campaign was calculated. This median value was assumed to be the receiver true position in the absence of scintillations. Subsequently, we defined a topocentric coordinate (north, east, up) centered at the median value, and we rotated all the receiver position samples into the topocentric coordinate system. Since the origin of the topocentric system represents the receiver position in the absence of scintillation, the receiver reported position samples in this coordinate system represents the “errors” due to scintillation. The vertical error is the “up” component of each position sample, while the horizontal error is equal to the square root of the addition of the square of the “north” component and the square of the “east” component.

Figure 3.

(a) Number of satellites that maintained lock, (b) position dilution of precision (PDOP), (c) geometric dilution of precision (GDOP), (d) vertical position error, and (e) horizontal position error during the less active night.

Figure 4.

(a) Number of satellites that maintained lock, (b) position dilution of precision (PDOP), (c) geometric dilution of precision (GDOP), (d) vertical position error, and (e) horizontal position error during the active night.

[18] Figures 5a–5b show the C/No of the satellites in view of the receiver during the less active night and a worst data of the active night respectively, at a time window of 180 s. The choice of a 180 s time window was informed by the interest to observe the signal characteristics of each satellite during a short time period, especially near a navigation outage. Figure 6ashows the statistics of the observed fades at different C/No thresholds (30, 25, 20, 15, 10 dB-Hz) for each of the scintillating satellites during the active night.Figure 6b shows the percentage occurrence of fades at each C/No threshold for each satellite over the entire period of the campaign. The percentage of occurrence of fades below each C/No threshold level was determined based on the number of nights of the campaign, at which fades below the specific threshold was detected. Figure 7a shows the tracks of the IPPs over the previous hour period, while Figure 7b shows the velocities of the IPPs at 22:30 UT on the active night. The circles in both figures show the location of the ionospheric penetration points (IPPs) in the sky above Ascension Island during the period. The diagonal lines show the magnetic meridians. Figures 8a–8bshow the distributions of the fading durations for each satellite at a common C/No threshold of 30 dB-Hz, during the active night, and over the entire period of the campaign, respectively.

Figure 5.

Carrier-to-noise density ratio (C/No) for a 180 s time window (a) during the less active night, PRNs (elevation angle): 4 (55°), 7 (48°), 8 (71°), 24 (19°), 26 (12°), 27 (42°) and 28 (33°); (b) during the active night, PRNs (elevation angle): 4 (61°), 7 (45°), 8 (67°), 24 (24°), 26 (14°), 27 (31°) and 28 (34°).

Figure 6.

(a) Fades distributions below each C/No threshold for each satellite on the active night and (b) percentage occurrence of fades below each C/No threshold for each satellite, over the entire period of the campaign.

Figure 7.

(a) The tracks of the IPPs over the previous hour period. (b) The velocities of the IPPs at 22:30 UT on the active night. The circles represent the location of the ionospheric penetration points (IPPs) in the sky above Ascension Island during the period.

Figure 8.

Distributions of fading durations below 30 dB-Hz threshold for each satellite (a) during the active night and (b) over the entire period of the campaign.

4. Discussion

[19] During the active night, most of the satellites' signals experienced scintillation. Ionospheric irregularities modulated the C/No, leading to incursions in excess of 20 dB–Hz at times. The enhanced rapidity of the signals fluctuations could have led to navigation bit errors; cycle slipping, or loss of tracking. Invariably, even when lock is maintained, too rapid fluctuations on the C/No of satellites could impact negatively on the navigation solutions [Carrano et al., 2005]. Since the lost satellite signals cannot be used for position calculation, the geometry of the useable satellite constellation is degraded during periods of outages. Dilution of precision (DOP) quantifies the influence of receiver-satellite geometry on GPS positioning accuracy. The best DOP is obtained when the satellites in view of the receiver are evenly distributed in the sky. For instance, in a case of four satellites in view of the receiver, the best DOP will be obtained if three of the satellites are equilaterally distributed over the sky, and the fourth satellite is located directly overhead in the centroid of the equilateral triangle, collectively leading to tetrahedron geometry by their line of sights (LOS) to the receiver [Kaplan, 1996; Misra and Enge, 2001].

[20] Typically, HDOP values are between 1.0 and 2.0 [Kaplan, 1996]. VDOP values are larger than HDOP values, indicating that vertical position errors are larger than the horizontal position errors. We suffer this effect because all the satellites from which we obtain signals are above the receiver. The horizontal coordinates do not suffer a similar fate as we usually receive signals from all sides. On the less active night, the receiver showed good tracking capability, as it consistently maintained lock on six to eight satellites, and the GDOP and PDOP values were generally less than 2.2 with no single record of an outage (Figures 3a–3e). In contrast, on the active night, between 2200 and 2300 UT, the number of satellites that the receiver maintained lock on often reduced to four and in some cases less than four, leading to the observed six navigation outages (Figures 4a–4e).

[21] Ideally, the vertical alert limit/horizontal alert limit (VAL/HAL) must be greater than the vertical protection level/horizontal protection level (VPL/HPL) for approach with vertical guidance (APV) services to be available. By definition, VPL and HPL are confidence bounds for measurements in the position domain. APV is an approach operation that is suitable to provide lateral navigation with vertical guidance based on GPS measurements. The basic concept is to provide vertical guidance to aircrafts at a decision height (DH) above the conventional 200 ft noted for Cat I precision approaches [Van Dyke, 2001; Cabler and DeCleene, 2002]. There are two categories of APV service: APV-I and APV-II. According toInternational Civil Aviation Organization [2006], the VAL requirements are 50 m (164 ft) for APV I and 20 m (66 ft) for APV II, while the HAL requirement is 40 m (130 ft) for both APV I and APV II operations.

[22] Figures 3d–3e show the vertical position error (VPE) and horizontal position error (HPE) for the less active night. The VPE and HPE during the active night are characterized with spikes between 22.00 and 23.00 UT (Figures 4d–4e). The VPE and HPE vary between 0 and 30 m and 4–30 m respectively for the other periods.

[23] Conventionally, integrity approach was based upon a notion that the actual error distributions would be close to Gaussian, and that their convolutions would result to positioning errors that are also close to Gaussian, although, there might be deviations from the ideal behavior [Walter et al., 2010]. Integrity analyses are cumbersome and laborious. Walter et al. [2009, 2010] provided detailed explanations of integrity analyses, especially from the standpoint of WAAS. As of date, aviation integrity capabilities are designed for SBAS operations. For single frequency SBAS, Radio Technical Commission for Aeronautics (RTCA) [2006] defines

display math
display math

where Kv is the Gaussian tail (Kv ≡ 5.33) and σuis the cover-bound of true 1σ value of the vertical position error. KH is the Gaussian tail (KH ≡ 5.73) and σmajoris the cover-bound of true 1σ value of the horizontal position error.

[24] Wanner and Nelthropp [2006] and Walter et al. [2010] from three years data collected at twenty locations within the WAAS coverage, showed that VPE increases with VPL, and that the increments become more pronounced as the percentage of confidence bounds increases. Consequently, the reported spikes shown in VPE and HPE plots (Figures 4d–4e) during 22.00–23.00 UT of the active night might influence the VPL and HPL. Perhaps, compliance to APV requirements during this time window (22.00–23.00 UT) might become stringent. Further research efforts are required to validate the relationship between the VPE/HPE and VPL/HPL.

[25] In determining navigation solution, the number of satellites that are lost simultaneously to deep fading is very vital. A receiver with fast reacquisition capability will reduce the chance of simultaneous losses of satellites to deep fades, thereby providing better navigation performance [Seo et al., 2011b]. The U. S. Wide Area Augmentation System Minimum Operational Performance Standards (WAAS MOPS) requirement for the reacquisition time of aviation GPS receivers states that: “For satellite signal outages of 30 sec or less when the remaining satellites provide geometric dilution of precision of six or less, the equipment shall reacquire the satellite within 20 sec from the time the signal is reintroduced” [RTCA, 2006]. This implies that a satellite that is lost to deep signal fading could have its reintroduction into the position solution delayed for up to 20 s [Seo et al., 2011b].

4.1. Statistics of Signal Fading

[26] On the less active night, four of the satellites namely PRNs 7, 8, 24 and 28, at elevation angles (el. angles) of 48°, 71°, 19° and 33° respectively were observed to maintain perfect tracking, as their C/No fluctuated slowly within the threshold bounds in the order of increasing elevation angles, while the other three; PRNs 4, 26 and 27 at elevation angles of 55°, 12° and 42° respectively, experienced frequent fading (Figure 5a). However, since four of the satellites maintained tracking consistently at good geometry, navigation solution was continuously achieved. On the active night, signals from PRNs 8 and 28 experienced no scintillations as they were observed to fluctuate slowly along the 45 dB-Hz and 40 dB-Hz baselines respectively. The signals of other satellites that were in view of the receiver were significantly impacted. Overall, six navigation outages were observed during the active night. The first, which was the worst during the active night in terms of duration (50 s), occurred between 22.102 and 22.117 UT (Figure 5b). During this period, PRNs 8 (el. angle: 67°) and 28 (el. angle: 34°) maintained good tracking, while PRNs 7 (el. angle: 45°) and 27 (el. angle: 31°) were observed to fluctuate into and out of fades. The two satellites' signals fluctuated simultaneously into fade at about 22.117 UT. Prior to this time, PRN 7 signal was lost at two instances; 22.105 and 22.112 UT respectively. It is important to note that after a signal that is lost to fades recovers, it usually takes few seconds for the pseudo-range data from such satellite to be included in the navigation solution estimation. PRN 7 signal was lost at two instances, concurrently as PRN 27 signal had just recovered from fades, perhaps, still waiting for a clean bill of health for its pseudo-range data to be included in calculating the navigation solution. This could have impaired the number of satellites readily available to the receiver to calculate the navigation solution. PRN 4 (el. angle: 61°) showed up briefly at 22.116 UT only to re-appear at 22.12 UT and maintain lock to 22.15 UT. PRNs 24 (el. angle: 24°) and 26 (el. angle: 14°) only showed up briefly. The inability of the receiver to track up to four satellites with good geometry within this time period (22.102–22.117 UT) led to the observed 50 s navigation outage. After this outage, there was a continuity of service for about 14 min before a second outage which lasted for 5 s, followed by other four outages that lasted for 2, 3, 1 and 7 s respectively.

[27] Generally, over the entire campaign period, PRN 27 signals scintillated with characteristic prolonged fading duration (∼2.2 s at times). PRNs 4, 7, 13, 24 and 26 signals on the other hands were observed to scintillate with characteristic rapidity and brief fading durations (0.1–0.4 s) that resulted into intermittent loss of lock. Two satellites; PRNs 1 and 28 recorded the least percentage occurrences of fades at each C/No threshold, while PRNs 4, 7, 13, 24, 26 and 27 recorded higher percentage occurrences of fades (Figure 6b). Although, it is important to note that PRN 1 satellite appeared only at three nights of the campaign. At 15 dB-Hz threshold and above, PRNs 8 and 28 recorded percentage occurrences of fades within the range 33.3–53.3% and 20–33.3% respectively. PRN 1 satellite recorded percentage occurrences of fades generally lower than 15% at all threshold level. All other satellites, apart from PRNs 1, 8 and 28, recorded percentage occurrences of fades within the range 33.3–86.6% at 15 dB-Hz threshold and above. At 10 dB-Hz threshold, PRNs 1, 8 and 28 recorded 6.6% occurrences of fades each, PRNs 4 and 24 recorded 13.3% each, while PRNs 7, 11, and 13 recorded 20% each. Finally, PRNs 26 and 27 recorded 26.6% occurrences of fades each at 10 dB-Hz threshold.

[28] Humphreys et al. [2009] and Carrano and Groves [2010] stressed the importance of signal decorrelation time as a parameter that could quantify signal fading rates. The fading rate is related to the satellite's effective scan velocity with respect to the magnetic field and plasma drift as the satellite traverses ionospheric irregularities. The satellite's effective scan velocity is defined as the rate with which the radio LOS cuts across contours of equal correlation in the ionosphere [Fremouw, 1980]. It accounts for the effect that the anisotropy of irregularities has on the conversion of spatial structures to temporal fluctuations. The zonal drift was measured during the campaign by the spaced receiver technique at the UHF frequency (250 MHz). Carrano and Groves [2010]gave detailed explanation of the spaced receiver technique. The zonal drift velocity was 135 m/s at 22:00 UT during the active night. Furthermore, on the active night, the average effective velocities of the scintillating satellites during the observed outages are PRN (eff. vel.): 4 (146 m/s), 7 (88 m/s) and 27 (26 m/s). PRNs 24 and 26 showed brief or discontinuous appearances, while PRNs 8 and 28 were quiescent during the period; hence, there is no meaningful correlation between their effective velocities and the fading rates of their signals. The effective velocity was computed in the post-processing, using the locations and velocities of the GPS satellites–determined from GPS ephemeris and the technique described inRino [1979]. The average effective velocity was thereafter calculated by taking the average of the computed effective velocity data of each satellite during the active night.

[29] As shown in Figures 7a–7b, among all the satellites which were scintillating at the time (PRNs 4, 7, and 27), only PRN 27 had an IPP with a significant component in the direction of the zonal drift (i.e., the magnetic east direction). When the effective scan velocity is computed, the zonal drift velocity is subtracted from the IPP velocity. Hence, the effective scan velocity for PRN 27 becomes comparatively less than those of the others. Given the inverse dependency of signals decorrelation time on the effective satellite's velocity, and thus, the direct dependency of the fading rate on the effective satellite's velocity, it became apparent why PRN 4 showed characteristic rapid fading, followed by PRN 7, while PRN 27 showed relatively slow fading rate (large fading duration). Consequently, the chances of losing lock on PRN 4 by the receiver were comparatively higher than those of the other two scintillating satellites. Generally, the satellite elevation angle, as well as the scan velocity of the IPP of the satellite with respect to magnetic field and plasma drifts influences tracking sustainability at a given strength of ionospheric perturbations.

[30] During the active night, the statistics of the observed deep fades for different scintillating satellites shows that PRNs 8 and 28 did not experience fades, while PRNs 4 and 7 signals experienced frequent fades, with occasional loss of locks. PRNs 13, 24 and 26 signals generally faded below the lower bound of the tracking lock threshold (less than 30 dB-Hz baseline); although PRN 13 showed brief appearance and incursion into fades. PRN 27 signals maintained lock at some instances, but with repeated incursions into fades, thereby dwelling comparatively longer in fades before recovery, with occasional loss of lock. As shown inFigure 8a, the fading duration is less than 1 s most of the time for PRNs 4, 7, 24 and 26, with denser samples distribution within 0.1–0.4 s, during the active night. For PRN 27, a contrary distribution was observed. The observed fades were characterized with longer durations (∼2.2 s at times), with a distribution that peaks at 1.5 s.

[31] Figure 8b shows signals fading durations over the entire period of the campaign. The fading duration was binned into 0.2 s interval. Similarly, PRNs 1, 4, 7, 11, 13, 24, 26 and 28 showed denser samples distribution within 0.1–0.4 s, while PRN 27, showed a distribution that peaks at the 1.6 s bin. The distributions of PRNs 1 and 28 were generally low over the campaign period. Generally, rapid fades with shorter durations like those displayed by PRNs 4, 7, 11, 13, 24 and 26 could cause high phase dynamics which could further lead to cycle slips, loss of lock or complete outages [Humphreys et al., 2010a, 2010b]. Increase in the strength of the ionospheric turbulence could increase the rapidity of the fades of signals that traverse the ionosphere, resulting in abrupt phase transitions that could further lead to overbearing impacts on the receiver's phase lock loop (PLL) [Strangeways, 2009; Strangeways et al., 2011]. As reported by Humphreys et al. [2010a], prolonged fading like the one experienced by PRN 27, is usually accompanied by phase dynamics that are comparatively slow, thereby allowing broadband measurement noise to dominate.

[32] Adapting the signals to possible bandwidth enhancements may assist in improving GPS receivers' carrier loop tracking performance, although the major drawback of this option is that wider bandwidth tracking loops incur more noise. Therefore, it is important to carefully consider the trade-off between bandwidth enhancements and simultaneous increase in the receiver's noise level. As suggested byGanguly et al. [2004], an adaptive carrier tracking loop for GPS receivers may be reasonable. The loop could operate by using only the PLL portion with a tight bandwidth. When phase and amplitude scintillations are detected, loop bandwidth will be adaptively increased to accommodate new signal conditions. On the other hand, when a deep amplitude fade occurs, there is no signal to track, and as such, only the PLL will lose lock. At this point, the frequency lock loop (FLL) portion of the loop will be activated in order to facilitate the re-acquisition of the signal as it recovers from fade.

[33] Furthermore, international cooperation on multiconstellation operations will increase the number of satellites in the sky, and this will increase the chances of always tracking more than four satellites by the receiver, even during disturbed ionospheric conditions. Although, we also have a caveat on this option, as we have no control over which portion of the sky that would be covered by scintillation and to what extent. In other words, scintillation being a nature-made phenomenon may cover LOS of most of the satellites if their tracking locations are in the region of scintillation in the sky. Moreover, SBAS satellites are geostationary satellites. This implies that they have a permanent IPP, and if they are covered by scintillation patches, until the patches move away, they will continue to suffer fading because they (geostationary satellites) are not moving [Bandyopadhayay et al., 1997]. Applications of other augmentations such as GBAS and ABAS will be of significant advantage in enhancing aviation safety under these conditions.

5. Conclusion

[34] Equatorial scintillation impacts negatively on GPS navigation measurements by reducing the number of satellites that are readily available for the receiver to calculate the navigation solution. Ideally, a minimum of four satellites is required to calculate a valid navigation solution. We analyzed GPS data from Ascension Island during the AFRL campaign of the solar maximum year of 2002 (5–19 March 2002). Deep signal fades, which further led to navigation outages, were observed during most of the nights of the campaign. This is in agreement with the result of Seo et al. [2009, 2011a, 2011b]. At times, scintillation impacts cause signals from some satellites to plummet below the lower threshold bound.

[35] In addition to the satellite's elevation angle, the effective scan velocity of the IPP of satellite's signals with respect to magnetic fields and plasma drifts in the plane of the ionosphere influences the signals fading rates under a given strength of ionospheric turbulence [Rino and Owen, 1980; Fremouw, 1980; Rino, 1982; Carrano and Groves, 2010]. The present results are in agreement with the conclusion drawn by Carrano and Groves [2010]. They explained that the rates of fading and fading depth are important factors often responsible for loss of lock on signals by receivers. Additionally, geometrical factors are also important. Strong scintillations impact on the receiver-satellite LOS geometry, leading to poor dilution of precisions and positioning accuracy.

[36] During quiescent conditions, the C/No of satellites normally varied slowly along 45 dB-Hz and 40 dB-Hz baselines. The observed navigation outages during severe scintillations in the equatorial region are particularly worrisome. Adapting satellites' signals to possible bandwidth enhancements may improve the receiver's sensitivity to scintillations. Furthermore, international cooperation on multiconstellation operations, with augmentation systems will increase the number of satellites readily available in the sky for receivers' tracking. The present study may support modeling and simulations efforts for overall improvement of aviation receivers' architecture, especially those intended for equatorial applications.


[37] The first author thanks the U. S. Government for the Fulbright Scholarship grant, and the Institute for Scientific Research, Boston College for hosting him. The authors thank Christopher Hegarty of MITRE Corporation for his useful discussions on VPL/HPL for APV operations.