Evening scintillation is an aspect of space weather occurring primarily in the equatorial region due to scattering of satellite signals by ionospheric F-region irregularities. In order to quantify scintillation-related performance degradation in GPS (Global Positioning System), we operated single frequency (L1) Ionospheric Scintillation Monitors (ISM), sampling at 50 Hz, at 6 sites in the South East Asia/Oceania region during the most recent solar maximum period 1998–2002. Scintillation indices, horizontal position, receiver lock time and dilution of precision were recorded every minute. Data reveal that individual one-minute positional errors can increase from a few meters to 30 m during scintillation, whilst statistics at the hourly upper decile level reveal an underlying seasonal modulation from about 5 m to 15 m. Error maxima occur during the equinoxes and are more pronounced at the equatorial anomaly crests than at the magnetic equator, as expected. Such errors may concern users of systems that are dependent on GPS.
 The performance of GPS receivers in the equatorial region may be affected by amplitude and phase scintillations imposed on the satellite signals by ionospheric F-region irregularities [Kelley et al., 1996]. At equatorial latitudes, such irregularities reach a peak in occurrence after sunset and during periods of high solar cycle activity [Davies, 1990]. There is also a seasonal variation, with peaks in activity in the Australian/Asian longitude sector appearing around the equinoxes, in March and September [Fang and Liu, 1983; Cervera et al., 2001].
 The formation of equatorial ionization irregularities is attributed to the Rayleigh-Taylor Instability [Kelley, 1989] in which an ionization perturbation forming on the bottom-side of the ionospheric F2 layer is amplified to produce an upward moving depletion or bubble, subsequently developing a spectrum of smaller irregularities with characteristic scale sizes from meters to tens of kilometers that map down magnetic field lines to the north and south Appleton anomaly crests located about 12 to 15 degrees from the magnetic equator [Aarons, 1982]. The irregularities present a random diffraction screen to signals passing through [Yeh and Liu, 1982], causing scintillations in amplitude and phase to be imposed, and possibly causing receiver tracking loops to lose lock on one or more satellites. The consequent degradation in GDOP (geometric dilution of precision) can be one source of GPS positional error. In addition, errors in the GPS signal delay correction in single frequency receivers may be caused by the depletion of ionization within plasma bubbles and this will also contribute to positional error at times of scintillation.
2. Receiver Network
 The 6 stations of the network [Thomas et al., 2001] monitor an extended geographic area (see Table 1). Data gathering commenced at Parepare, Pontianak and Marak Parak in December 1997, at Darwin in June 1998, at Vanimo in August 1999 and at Fang in June 2000. From 12 May 2001, the receiver at Fang was relocated to Chiang Rai, 65 km to the east. In this paper, we compare data from the northern anomaly crest (Fang/Chiang Rai), the magnetic equator (Marak Parak) and the southern anomaly crest (Vanimo).
Table 1. GPS Station Locations (in Degrees)
IGRF Geomagnetic Latitude
1998 Geomagnetic Dip
 The six ISMs are based on a NovAtel 11-channel single frequency (L1) C/A (Coarse Acquisition) code receiver fitted with an oven-controlled crystal oscillator for low phase noise performance [Van Dierendonck et al., 1993, 1996]. The received signals are sampled at a rate of 50 Hz to provide processed data every minute, including scintillation parameters, position, receiver lock data and GDOP. Amplitude scintillation is monitored through the S4 index, which is the normalized standard deviation of de-trended signal power.
Cervera and Knight  modeled the carrier tracking response of a second order Costas loop to both amplitude and phase scintillation, finding that loops can experience an increased level of phase noise, cycle slips and loss of lock, due to deep amplitude fades and rapid signal phase fluctuations. A wide loop bandwidth (say, 20 Hz) results in increased noise, poorer signal-to noise and increased susceptibility to fading, whereas a narrow bandwidth (5 Hz) is susceptible to Doppler shifts related to phase scintillation. Our receiver, with a relatively wide phase-lock-loop noise bandwidth of 15 Hz, should be robust through strong phase scintillation, but it is inevitable that tracking loops will be stressed by severe equatorial conditions that apply during equinoxes at solar maximum.
3. Observed Effect of Scintillation
 Receivers can suffer degraded performance as tracking loops become stressed, cycle slips occur, lock is lost and GDOP increases. Performance can be directly monitored every minute through measured scintillation indices, the elapsed time since the last loss of lock event, GDOP and errors in the receiver position solution. We choose to base our statistical measure of positional accuracy on the R95 parameter, that is, the radius of the circle in the horizontal plane, centered on the receiver's true or reference position, that encloses 95% of all position measurements taken from within a defined time interval. We divide data into hourly intervals, each comprising 60 position estimates. Our reference position is the daily mean (up to 1440 data points).
 Before looking at long-term statistical variations in GPS performance, we first examine the effects of scintillation on a single day, 13 September 2000, at the Fang station. Figure 1 shows the variation with time of day at 60s intervals of the power scintillation index S4 from satellite PRN27, together with satellite elevation angle (0°–90°) over-plotted on S4, the time (s) since the previous loss of receiver lock, GDOP and hourly R95 (m). Strong scintillation occurred around 15UT (about 22LT) at a satellite elevation too high to be vulnerable to multipath propagation. Signal lock was lost on several occasions (as measured by the number of occurrences of lock time falling to near zero) and GDOP rose from about 2 to 7. The bottom panel of Figure 1 shows that R95 during this day increases to a maximum of about 15 m during the main scintillation event. We have noted also on other occasions that it is possible to observe increases in R95 without significant rises in GDOP, suggesting erroneous GPS delay corrections in the presence of equatorial bubbles.
 For the same day, Figure 2 compares the scatter in position measurements and the size of the R95 circle for time interval 15–16UT during the scintillation event (crosses) with another time (23–24UT) when scintillation was absent (filled circles). There is a dramatic contrast between the two times, the maximum error during the scintillation event reaching more than 30 m, and R95 reaching about 15 m compared with only about 2 m at the quieter time. The small offset of the quiet time data from the reference position is most likely due to an imperfect correction applied within the receiver for background ionospheric delay.
 We monitor the long-term variation of receiver performance in terms of two quantities, (a) the daily occurrence rate in loss of lock events, summed over all satellites at elevations above 40°, and (b) the daily maximum of R95. The relatively high elevation mask angle of 40° localizes data to within about 3° of the station's geographic position, assuming an ionospheric irregularity height of 300 km. Three stations have been selected, (a) Fang/Chiang Rai under the northern anomaly crest, (b) Marak Parak near the magnetic equator, and (c) Vanimo under the southern anomaly crest. We compare the long-term behavior of the 3 stations starting January 2000 and ending December 2002.
Figure 3 shows the long-term variation of the daily loss of lock count. A clear seasonal modulation, with equinoctial maxima, is evident at the two anomaly crest stations but absent from the magnetic equator station. This is consistent with the enhanced level of irregularity formation and scintillation at the anomaly crests compared with the equator [Aarons, 1982]. Figure 4 displays the daily maxima of R95, smoothed with a running 3-point median filter to better reveal the underlying seasonal trend of receiver accuracy. Data before 2 May 2000 are off-scale due to activation of Selective Availability (SA). After this date, the ensuing improvement in instrumental accuracy easily reveals a seasonal modulation in R95, maximizing around the equinoxes. A statistical examination of the times of day at which receivers suffer maximum R95 shows a strong occurrence peak in the evening between about 8pm and midnight local time, confirming that the effect is scintillation-related. The presence of a small positional accuracy modulation at the equator in the absence of any significant loss of lock (Figure 3) is further evidence for erroneous GPS delay corrections in the presence of equatorial bubbles.
 Whilst removal of SA has offered much improved accuracy to users of single frequency receivers, it has also made scintillation effects more evident. It is clear from these data taken at solar maximum using single frequency (L1) C/A code receivers that equatorial scintillation events can cause frequent loss of channel lock and can have a deleterious effect on GPS positioning. As a result of scintillation, we find that receiver loss of lock occurrences and subsequent positional errors show significant seasonal modulation with peaks occurring at the equinoxes. As expected from the known latitudinal structure of the equatorial anomaly, maximum effect is achieved under the anomaly crests (Fang/Chiang Rai and Vanimo) rather than at the magnetic equator (Marak Parak). The small seasonal modulation observed in positional accuracy for Marak Parak, without loss of lock, suggests that erroneous GPS delay correction, due to depletion of the ionization within equatorial bubbles, also contributes to degradation of positional accuracy. Scintillation-related degradation in positional accuracy at low latitudes occurs preferentially in the evening hours before midnight. If high reliance is being placed on GPS for system management (for example, aircraft control), the effects reported herein may need to be considered.