Global Positioning System and solar radio burst forensics



[1] On 6 December 2006, a solar radio burst associated with a class X6 solar flare demonstrated that GPS receiver operation is vulnerable to solar radio burst noise at 1.2 GHz and 1.6 GHz. Within 8 days, two more solar radio bursts confirmed the initial results. These solar radio bursts occurred at solar minimum when they were least expected. Given that measurements of solar radio bursts extend back to at least 1960, why did 40 years pass before anyone realized that solar radio bursts could be so intense or pose a potential threat to the continuous availability of GPS operations? An investigation has been conducted to see if archived solar radio burst data or GPS data could be used to detect intense solar radio bursts. With the exception of the intense solar radio bursts of December 2006, we find that when both GPS data and Radio Solar Telescope Network (RSTN) data are available, they agree within the limits presented by differing reception frequencies and unknown polarization. However, inconsistencies and lapses within the RSTN data set were also discovered, making it unlikely that we will ever know the true number of intense (>150,000 solar flux unit) solar radio bursts that may have occurred during the last 40 years.

1. Introduction

[2] On 6 December 2006, an X6-class flare erupted, producing a record-setting solar radio burst (SRB). This SRB was so intense that it disrupted the operation of many GPS receivers, particularly those providing precise positioning and using the L2 P(Y) signal to remove ionospheric errors [Cerruti et al., 2008]. This SRB exceeded 1000 kSFU at L1, roughly ten times more intense than any previously measured SRB and 100 times more intense than any previously measured SRB at solar minimum. Within 2 weeks, two more remarkable solar radio bursts occurred, less intense than the 6 December event, but with one still more intense than any previously recorded event and one nearly more intense.

[3] The historical record of continuous measurements of SRBs is quite brief, extending back to 1960, just four and one-half solar cycles. Hence when an unexpectedly intense SRB occurs, several questions immediately occur. Was the 6 December SRB a simple outlier on a well behaved statistical distribution such as, for example, a 100-year flood? Is the sun exhibiting a trend in which more intense solar radio bursts might be expected in the future? Or has the method of monitoring SRBs been inadequate so that the power of previous intense SRBs was underestimated? Conversely, now that solar radio bursts are a recognized threat, what is the probability that they will affect GPS or other GNSS receiver operations over the next solar cycle? None of these questions have clear answers. Here we examine archived GPS data and RSTN solar radio burst data to investigate previous intense solar radio bursts as a form of GPS solar radio burst forensics.

[4] GPS operates at two frequencies called L1 (1575.42 MHz) and L2 (1227.6 MHz) and is a right-hand circularly polarized (RHCP) signal. The nearest frequency employed by RSTN is 1415 MHz and polarization information is not archived. We have examined two forms of RSTN data archived at the NOAA National Geophysical Data Center (NGDC): fixed-frequency event listings and RSTN one-second (1-s) data. Fixed-frequency event data archive the start time, stop time, and peak power fluxes of solar radio bursts at a variety of fixed frequencies. Again, we use 1415 MHz data. These data may be obtained at the Web site

[5] We obtain the GPS carrier-to-noise ratio (C/N0) from the IGS (International Global Navigation Satellite System Service) and GeoNet networks to estimate the power of previous SRBs and then compare them with SRB power measured by the RSTN. The IGS and GeoNet networks have the advantage of easily accessible observation files archived at (and elsewhere) but the disadvantage of only extending back about 10 years for many sites. In addition, C/N0 is not a required observable by the IGS, so many, if not most, stations do not record C/N0. The GPS response to solar radio bursts yields a lower bound for the SRB power flux since GPS only measures the right-hand circular (RHC) power. Nonetheless, a lower bound is adequate to examine RSTN and its estimates of SRB power.

[6] To find examples of solar radio bursts, we first examined GPS data coincident with ten separate X-class flares. The dates inspected were 2 November 1992; 9 July 1996; 6 November 1997; 23 and 27 April; 6 May; 18 August; 28 November 1998; 22 November 2001; and 28 October 2003. For the two oldest dates, no GPS data could be found. For two other dates, the effect on GPS receivers was ambiguous or nonexistent. The remaining six dates have data showing the effect of SRBs on various GPS receivers. Of these six dates, the RSTN only had 1-s data available for the two most recent: 22 November 2001 and 28 October 2003.

[7] Next we searched the solar radio burst event data at NGDC at 1415 MHz for any additional intense events and found one additional event on 15 April 2001 for which there are GPS data but no RSTN 1-s data.

[8] This paper is organized by first reviewing the relationship between GPS signals and solar radio burst intensity. Then we reexamine the 6 December 2006 events to understand previous inconsistencies between the RSTN, the Owens Valley Solar Array, and GPS C/N0 data. We then show examples from 11 November 2001 and 28 October 2003. We will limit our presentation to L1 CA (coarse acquisition) C/N0 amplitudes only (1575.42 MHz), since they exhibit the most robust tracking in stressed environments.

2. Effect of Solar Radio Bursts on GPS C/N0

[9] To compare the reduction in GPS L1 signal amplitude with the observed SRB power at 1415 GHz, we use the expression derived by Cerruti et al. [2006]:

equation image

where Pn is the system noise power, Aeff is the effective area of the antenna, and Δ(C/N0) is the change in the carrier-to-noise ratio caused by the SRB. This expression was verified using L1 CA power estimates from multiple receivers and comparing them with results from the OVSA RHCP measurements of SRB power. Using the numbers appropriate for the IGS receivers found in the work of Cerruti et al. [2006], we compute that

equation image

where Δ(C/N0) is in dB-Hz. Only GPS signals from satellites with elevations above 30° are used for quantitative comparisons so that elevation-dependent antenna gain is not an important factor, although, in principle, such elevation effects can be taken into account [Carrano et al., 2009].

3. Reexamination of the 6 December 2006 Event

[10] To use archived RSTN solar radio burst data, we first need to know if the reports are accurate. Cerruti et al. [2008] suggested inconsistencies between the fixed-frequency event data at 1415 MHz and the fluxes implied by reduction of C/N0 in GPS receivers. To address this question, an independent measure of SRB power is required. The first step is to compare the 6 December 2006 RSTN data with the Owens Valley Solar Array (OVSA) measurements. To interpret the comparison, note that RSTN uses an 8 MHz bandwidth with an unknown response time whereas OVSA uses a 100 MHz bandwidth with a response time that is small compared to its 8.1-s cadence.

[11] This comparison has already been made by Cerruti [2008] and is summarized in Table 1. The first three rows show the reported time of peak flux from the fixed-frequency event listings, the value of the peak flux, and then the flux at that same time as observed by OVSA. The last row shows the peak flux time and value observed by OVSA. For example, OVSA observed a peak flux at 1933 UT of 1003 kSFU (kilo-Solar Flux Units) whereas the Sagamore Hill RSTN station reported a peak at 1917 UT of only 13 kSFU. At 1917 UT, OVSA observed 53 kSFU. The Palehua RSTN station also observed a peak flux at 1917 UT but of a higher value (150 kSFU) than the Sagamore Hill station and later reported an equally large flux of 150 kSFU at 1931 UT. At 1931 UT, OVSA reported a flux of 341 kSFU. The conclusion from Table 1 is that the Radio Solar Telescope Network (RSTN), at least for this one example, underestimates the SRB peak powers and is not consistent from station to station. The discrepancy for the Palehua report for 1931 UT is understandable, since the description of the RSTN performance states that power fluxes exceeding 100 kSFU are not accurately reported [Springer et al., 1980]. The reports from both RSTN sites at 1917 UT, however, remain puzzling.

Table 1. Measured Peak Power at Two RSTN Stations Compared to Peak Power Measured at OVSA on 6 December 2006
StationTime of Peak Power (UT)RSTN 1415 MHz (kSFU)OVSA 1400 MHz RHCP (kSFU)
  • a

    Not applicable.

Sagamore Hill, Massachusetts19171353
Palehua, Hawaii191715053
Palehua, Hawaii1931150341

[12] After our original comparison of RSTN data with the GPS and OVSA observations, the Sagamore Hill 1-s data became available. Figure 1 compares the Sagamore Hill RSTN signal power at 1415 MHz with the OVSA signal power at 1405 MHz over a roughly 1.5 h period encompassing the intense solar radio burst of 6 December 2006. The blue + symbols represent OVSA 1400 MHz data, and the red dots are the Sagamore Hill 1415 MHz data. The agreement is remarkably good for both receivers at SRB fluxes greater than 0.1 kSFU except for periods between 19.5 UT and 19.7 UT when the Sagamore Hill radio fluxes saturate at nearly 100 kSFU and report inaccurate values. The value of the fixed-frequency event peak flux for Sagamore Hill is shown as a green line. The fixed-frequency event data neither accurately report the peak flux nor the end time of the SRB. Hence the Sagamore Hill 1-s data clearly agree with the OVSA data when not in saturation and clearly disagree with the fixed-frequency event listing summarized in the first row of Table 1. NOAA responded to inquiries for clarification of this discrepancy. The discrepancy was acknowledged but it was not possible to determine the cause (E. H. Erwin, personal communication). In addition, confirmation was received that the Palehua 1-s data were not available for 6 December 2006 nor are they likely to become available, so inconsistency between the 1-s data and the fixed-frequency event data could not be investigated for rows 2 and 3 in Table 1. We conclude that the RSTN reports are not optimum for investigating previous extreme events that could affect GPS for three reasons: (1) the RSTN radiometers saturate at 100 kSFU, (2) the RSTN radiometers do not distinguish RHCP flux from LHCP flux, and (3) although the RSTN 1 s data appear to be accurate except where saturated, the reported RSTN peak fluxes may contain errors.

Figure 1.

Comparison of 1-s Sagamore Hill radio burst data with similar data from OVSA and the reported peak flux from fixed-frequency event data.

[13] We now proceed to investigate in more detail the two events of our sample that have comparable GPS and RSTN 1-s data. The first of the events also has available OVSA data for comparison.

4. The 22 November 2001 SRB Event

[14] In Figure 2 (top), the C/N0 amplitudes are expressed in dB-Hz at 1575.42 MHz for all of the satellites in view from GPS site KARA (Trimble 4000SSI GPS receiver, latitude 43.61°, longitude 169.78°). The station was observing eight satellites on L1. C/N0 was calculated by taking C/N0 = 27 + 20 * log (SNC) dB-Hz where SNC is the signal-to-noise count that Trimble receivers record in their RINEX files. The C/N0 amplitudes varied from about 33 dB-Hz to 56 dB-Hz. A significant reduction in C/N0 amplitude for all signals occurred between 22.0 and 22.2 UTC. Other smaller simultaneous amplitude decreases for all signals can be seen at several additional times, for example, at 22.8 UTC. In Figure 2 (bottom) is an estimate of SRB power from the Palehua, Hawaii RSTN station expressed in SFU (blue), as well as a SRB power estimate from OVSA (green) and the power estimate derived from the GPS fade depth (red). The RSTN power estimate is for the sum of both right-hand circular (RHC) and left-hand circular (LHC) power. The peak SRB power is between 22.0 and 22.2 UTC with an amplitude of 46.4 kSFU. The peak value measured at the OVSA was 49.1 kSFU, and the GPS measurements indicate a peak burst of 59.6 kSFU. The OVSA power shown is also the sum of RHC and LHC power, since RHCP measurements were not available on that day. However, Nobeyama Radio Polarimeter measurements starting after 2300 UT indicate that the burst is strongly RHCP between 1 and 2 GHz. Other smaller peaks in SRB power occurred at later times, some of which coincided with reductions in GPS signal C/N0 amplitudes.

Figure 2.

C/N0 GPS L1 CA signal amplitudes (dB-Hz) shown for (top) all satellites in view and (bottom) RSTN SRB at 1415 MHz measured at Palehua, Hawaii, OVSA SRB at 1600 MHz, and SRB predicted by GPS power fade.

5. The 28 October 2003 SRB Event

[15] In Figure 3 (top), the C/N0 amplitudes are expressed in dB-Hz at 1575.42 MHz for all of the satellites in view from GPS site SIMO (Rogue SNR-8000 GPS receiver, latitude 34.188°, longitude 18.4°). The station was observing eight satellites on L1. C/N0 was calculated by taking C/N0 = 20 * log (SNRv) − 3 dB-Hz where SNRv is the 1-s voltage SNR recorded by Rogue SNR-8000 receivers in their RINEX files. The C/N0 amplitudes varied from about 33 dB-Hz to 53 dB-Hz. The 28 October 2003 event, one of the famous “Halloween” solar events [Tsurutani et al., 2006], was somewhat more complex than the 22 November 2001 event. There are three periods of reduced C/N0 amplitudes at about 11.1, 11.8, and 11.9 UTC. In Figure 3 (bottom) is an estimate of solar radio burst power from the San Vito, Italy RSTN station expressed in SFU as well as the SRB power predicted by the GPS measurements. The three peaks in SRB power coincide with the reductions in the GPS L1 C/N0 signal amplitude. The largest SRB power was 20.4 kSFU at 11.8 UTC, whereas the GPS data indicated a burst of 34 kSFU. The OVSA was in darkness during this period and has no data for the event.

Figure 3.

C/N0 GPS L1 CA signal amplitudes (dB-Hz) shown for (top) all satellites in view and (bottom) RSTN SRB at 1415 MHz measured at San Vito, Italy, and SRB predicted by GPS power fade.

[16] The observation of SRB interference at GPS frequencies is confirmed by Chen et al. [2005], who measured the rate of loss of phase tracking at the L2 frequency from the IGS receiver NKLG, located at 2.1° latitude and 9.4° longitude near the subsolar point. The L2 frequency is 1227.6 MHz and L2 phase tracking is more susceptible to interference because of the required semicodeless tracking algorithms that raise the noise floor [Woo, 2000]. The NKLG signal showed peak values for loss of L2 phase tracking at 11:05 UT, 11:45 UT, and 11.58 UT, virtually coinciding with the RSTN SRB amplitude peaks and reductions in C/N0 GPS L1 signal amplitudes shown in Figure 2. Other IGS stations examined by Chen et al. [2005] showed similar behavior.

6. Conclusions

[17] GPS receivers operate with an RHCP signal architecture and use RHCP antennas. For a randomly polarized signal, GPS receivers respond to only half of the signal power. For SRB, which can be primarily RHCP, LHCP, or a mixture of both, GPS receivers will only respond to RHCP power, which has been confirmed by Cerruti et al. [2006]. Thus, the power estimated from a reduction in C/N0 signal amplitude caused by a SRB is a lower bound estimate of the true SRB power.

[18] Table 2 is a comparison of the RSTN fixed-frequency event peak fluxes with peak fluxes estimated GPS signals and the OVSA instrument. Beginning with the 22 November event, the time history in Figure 2 shows rough agreement with all three (RSTN, OVSA, GPS) estimates of solar radio burst power, with OVSA and GPS implying somewhat larger power than the RSTN. This is confirmed in the top row of Table 2 where the comparison is now made with RSTN fixed-frequency event data. The agreement for the 28 October event is not as clear. The time series data in Figure 3 show that the GPS data imply a larger SRB power except for impulsive events just after 1200 UT. The second row of Table 2 implies reasonable agreement between GPS-based estimates and RSTN. However, the RSTN fixed-frequency event peak flux occurred in the impulsive structures just after 1200 UT and did not coincide with the peak flux time implied by the GPS signals. Hence this apparent agreement in Table 2 is less convincing.

Table 2. A Comparison of SRB Power Estimated From GPS L1 CA Receivers and Measured by RSTN at 1415 MHz and OVSA at 1600 MHz
DateΔC/N0 (dB-Hz)SRB Power Calculated From GPS L1 Signal (kSFU)SRB Power Measured by OVSA (1600 MHz) (kSFU)RSTN Fixed Frequency Event Peak Flux (kSFU)
  • a

    Not available.

22 November 20018.359.649.146.4
28 October 20036.333.9N/Aa20.4

[19] There are reasons to suspect that significant spectral gradients can occur in SRB over the L band range of frequencies. Cerruti [2008] demonstrated in one event that the difference in solar radio burst power between the L1 and L2 frequencies can be 2 orders of magnitude, with L2 being smaller than L1 [see also Carrano et al., 2009]. The difference in frequency between L1 and 1415 GHz is about half of that between L1 and L2, so it is impossible to rule out a sharp spectral gradient in power as an explanation for the difference between the SRB power implied from GPS signals and RSTN observations. However, it cannot be ruled out that the scaling factor in equation (1) may require refinement. Additional joint C/N0–solar-radiometer events of sufficient accuracy are needed to investigate this possibility.

[20] The RSTN receiver description [Springer et al., 1980] states that the largest signal amplitude observable at 1415 GHz is 100 kSFU, which accounts for the differences between amplitudes measured by the OVSA and RSTN shown in Figure 1 during the 6 December 2006 event. However, the RSTN fixed-frequency event peak flux of 13 kSFU does not agree with the RSTN 1-s data from which it is derived, as shown in Figure 1. The peak flux is about 7 times smaller than that observed in the 1-s data.

[21] This brief study thus supports the conclusion that SRB power has been underestimated by the RSTN during the past decade. Furthermore, the large disagreement between the single-frequency event peak fluxes and 1-s data on 6 December 2006 implies that peak fluxes are unfortunately not reliable in a search for earlier intense SRB events. From the past 6 to 7 years for which reliable GPS C/N0 data and OVSA data (roughly 8 h/d, from 1600 to 2400 UT) are available, we can safely conclude that there were no SRB comparable to those during December 2006. Nobeyama Radiopolarimeter data extend the coverage from 0000 to 0800 UT and, indeed, two events exceeding 500 kSFU were reported in their flare list at 1 GHz in 1990 and 1993 (see For other times when only RSTN fixed-frequency event data are available, intense SRB may have occurred without being accurately reported.

[22] We conclude by noting that accurate solar radiopolarimeters that (1) are distributed worldwide, (2) measure RHCP, and (3) measure the continuous range of frequencies between at least 1–2 GHz, would be particularly valuable for verifying how often solar bursts occur that are capable of affecting GPS. In the absence of such a worldwide array of solar radiopolarimeter stations, GPS receivers themselves can be used to address this question, provided the carrier-to-noise ratio is recorded as part of the archived data.


[23] This research was supported by ONR grant N00014-04-1-0105 to Cornell University. OVSA is supported by NSF grants AST-0352915 and AST-0607544, and by NASA grant NNG06GJ40G to the New Jersey Institute of Technology.