Longitudinal correlation of equatorial ionospheric scintillation



[1] Studies at the Air Force Research Laboratory are investigating methods for the incorporation of expected data sets from the Communication/Navigation Outage Forecast System (C/NOFS) satellite into existing scintillation specification models for an improved knowledge of the regional scintillation environment. Results from this research will form the basis for ground-based empirical scintillation forecast algorithms. One such study involves a determination of the extent to which correlations can be made for scintillation over a range of longitudes providing valid upstream (westward) forecasts of scintillation activity based on near real-time observations. Using archived data collected on Scintillation Network Decision Aid (SCINDA) receivers, a comprehensive study was conducted to numerically evaluate the correlation of scintillation on link pairings separated by varying degrees of longitude representing up to 3 hours in local time.

1. Introduction

[2] With its low-inclination orbit (∼13°), the C/NOFS satellite will provide valuable information for use in existing equatorial scintillation specification models. Scintillation data recorded at multiple ground sites on VHF/UHF/L band signals from a coherent electromagnetic radio tomography tri-mode beacon onboard the spacecraft will be used in a new Space-Based Scintillation Network Decision Aid (SB-SCINDA) to complement the expanding SCINDA ground-site network. Additionally, SB-SCINDA will make use of in situ density profiles from C/NOFS in conjunction with a phase screen algorithm, resulting in a much improved regional scintillation specification.

[3] Observations of equatorial spread F (ESF) occurrence date to the 1930s when Booker and Wells reported on the scattering of radio waves by the F region of the ionosphere. In the 1960s, advanced techniques for observing irregularities were developed through the monitoring of signals from geostationary beacons [Basu, 1985]. Measurement of enhanced S4 levels (defined as the normalized standard deviation of the signal power incorporating both the intensity fades and the fade distributions) on satellite communication (SATCOM) links provides an indication of the presence of F region plumes or depleted regions in the ionosphere. These highly structured irregularities originate in the bottomside F region and have been observed on incoherent scatter radar to extend to altitudes greater than 1000 km above the magnetic equator, mapping along the magnetic field to ±15° in magnetic latitude [Groves et al., 1997]. Changes in the scale sizes of the irregularities lead to a frequency-dependent decay rate [Aarons, 1993] as these ESF plumes experience a zonal drift equal to that of the postsunset ionosphere, typically 150–200 m/s eastward, decreasing throughout the evening as inferred from ground-based observations.

[4] The driver behind SCINDA, a model which provides a situational awareness of the scintillation environment from ground-based observations, was developed at the Air Force Research Laboratory in the mid-1990s for use with fixed-link VHF data [Groves et al., 1997; McNeil et al., 1997] and later upgraded to incorporate L band scintillation observations from Global Positioning System (GPS) satellites [Caton et al., 2004; McNeil and Caton, 2002]. An inherent limitation to the current version of SCINDA is its lack of ability to offer any knowledge of the local scintillation environment to the west of (or upstream from) the monitored SATCOM links. F layer irregularities drift across the SATCOM links, where they are recognized by the model, are propagated eastward (or downstream) at the observed drift speed of the ionosphere, and are decayed with empirically based algorithms that have been extensively validated [Caton and McNeil, 2004a, 2004b]. At this point, however, no information is provided to users requiring specification of conditions to the west of these observations. Theoretical models have been proposed for understanding the longitudinal behavior of scintillation, most notably by Tsunoda [1985], and studies have been performed with in situ satellite data and ground-based observations to better understand the morphology of F layer irregularities as a function of longitude [Maruyama and Matuura, 1984; Aarons, 1993; Burke et al., 2004]. Results from such studies have been incorporated into the climatology available from the wideband ionospheric scintillation model (WBMOD) [Secan et al., 1995] and are used in the generation of SATCOM products. It has been a goal of the researchers working on SCINDA to provide a more accurate specification of the scintillation environment across as wide a region as possible through the use of real-time observations. The primary goal of the effort described here is to provide a method for the determination of expected scintillation intensities to the west (upstream) of observations made on SATCOM links on the basis of results from a statistical analysis of the longitudinal correlation of scintillation activity.

2. Methodology

[5] Validation studies of the SCINDA model [McNeil and Caton, 2000; Caton and McNeil, 2004a] have shown that predictions of expected scintillation levels 6° in longitude downstream (eastward) from the observation point can be made with greater than 75% accuracy. In an attempt to expand the capability of the current model, an investigation was carried out in which the ability to correlate scintillation observations between station pairs with varying longitudinal separations was examined statistically.

[6] Because of the nature of the ionosphere, one would suppose that if the conditions exist at a specific longitude to trigger the instabilities, leading to the formation of plumes and causing scintillation, the likelihood of observing scintillation a few degrees in longitude to the west of this location would be significantly high as well, particularly during the peak of the local scintillation season. Likewise, if observations indicate comparatively quiet conditions at a specific longitude, one might expect the likelihood of an inactive evening a few degrees to the west to be higher than in the previously mentioned case. Quantifying this longitudinal relationship would provide a significant enhancement to the SCINDA model's ability to offer situational awareness of scintillation activity by extending our region of coverage to the west (upstream) of the longitude of observation of ionospheric plumes.

[7] To accomplish this goal, a configuration-driven software tool was developed for the analysis of the correlation of scintillation activity on selected east-west link pairs. “ACTIVE” and “INACTIVE” designators were assigned on a nightly basis for each link on the basis of the number of hours of observed scintillation above a selected threshold level. Configurable parameters include (1) the local time range over which correlations are to be computed, (2) the number of hours of data available required for a “valid” evening, (3) a scintillation threshold in S4 (the normalized standard deviation of the signal power incorporating both the intensity fades and the fades distributions), and (4) the number of hours of scintillation above the threshold required for defining active conditions. For the results discussed in this paper, parameters 1 and 2 are fixed, requiring at least 3.5 hours of data to be available between 1.0 and 5.0 hours after local sunset for valid evenings. Parameters 3 and 4 will be varied.

[8] Stepping through the available data sets, nightly correlation scores were assigned with the following logic: simultaneous “ACTIVE” or “INACTIVE” nights on both east and west links return a correlation score of 1; when either link was “ACTIVE” while the other was “INACTIVE,” a correlation score of 0 was assigned. These nightly results were then binned by the activity level on the easternmost link (“ACTIVE” or “INACTIVE”) and averaged over running 31-night intervals. For this step, the beginning evening for each monthly interval was stepped by 15 each time, producing averaged results for nights 1–31, 15–46, 31–61, etc. Plots of the averaged correlation scores fluctuate noticeably throughout the year. Upon closer examination, it was noticed that these fluctuations could be associated with seasonal changes in scintillation activity.

[9] Stepping through the available data sets, nightly correlation scores were assigned with the following logic: simultaneous “ACTIVE” or “INACTIVE” nights on both east and west links return a correlation score of 1; when either link was “ACTIVE” while the other was “INACTIVE,” a correlation score of 0 was assigned. These nightly results were then binned by the activity level on the easternmost link (“ACTIVE” or “INACTIVE”) and were averaged over running 31-night intervals. For this step, the beginning evening for each monthly interval was stepped by 15, each time producing averaged results for nights 1–31, 15–46, 31–61, etc. Plots of the averaged correlation scores fluctuate noticeably throughout the year. Upon closer examination, it was noticed that these fluctuations could be associated with seasonal changes in scintillation activity.

[10] The method described by Carrano and Griffin [2004] offers the ability for determination of seasonal transition dates as a function of longitude. Using climatological scintillation output from WBMOD at the 85th percentile, this technique provides, as a function of longitude (with 5° resolution) and day of year, a quantity known as the “Total Mean Hourly S4” (TMHS4). The THMS4 index provides an indication of the intensity and duration of scintillation on a nightly basis and is defined as the sum of the average S4 observations throughout each night. The climatological S4 from WBMOD is typically saturated at 1.0; therefore a THMS4 index of 3.0 on a given night implies at least 3 hours of severe scintillation. Contour plots of global THMS4 values as a function of the day of year supply the information required for an algorithm providing a seasonal delimiter as a function of longitude and day of year. Seasonal transition dates were compiled for each VHF link designated for use in this study. Because the transition dates are dependent upon longitude, there will occasionally be overlaps when comparing data from any two links. For example, on a particular day of the year, a link at Guam may be in a “TRANS” season while the Manila longitude sector is in the “OFF” (or “ON”) season. In a case such as this when nightly correlations between these two links are desired, the “TRANS” season was given precedence over the “OFF” or “ON” season. With this information, the 31-night correlation scores could be further binned, providing a seasonally dependent estimate of the correlation of scintillation activity over a known longitudinal distance. This process was repeated with station pairs of differing longitudinal separations and again for varied definitions of active conditions, parameters 3 and 4, leading to the development of a set of seasonally dependent curves detailing the spatial correlation of scintillation at various levels of activity.

3. Longitudinal Scintillation Correlation Study With GPS Data

[11] SCINDA network scintillation monitors have been operational since the mid-1990s. Beginning with the installation of VHF receivers at Ancón, Peru; Antofagasta, Chile; and on Ascension Island, the present configuration consists of 12 real-time data sites with both GPS and VHF receivers monitoring and reporting scintillation measurements in near real time. A handful of additional sites, supported by the Australian Defence, Science and Technology Organisation, are recording GPS scintillation data. Although these sites have yet to become part of our real-time operational network, the archived data sets provided from these locations have proven valuable for model validation and enhancement. From each site, the intensity scintillation parameter, S4, is archived as a function of time valid at the frequency for which it was observed. Archived data sets from L band GPS receivers on the SCINDA network contain the scintillation level recorded on each satellite visible as a function of time along with the satellite azimuth and elevation as measured from the station.

[12] A problem with the use of GPS scintillation data in this study is the ability to precisely determine the longitudinal separation of the east-west data links because of the fact that the observation of scintillation data at a ground site is actually made over a range of longitudes. Assuming a cutoff at 10° elevation, the longitude of the ionospheric penetration point at an altitude of 300 km of a GPS satellite can vary more than 15° from the longitude of the station. For the purposes of this study it was desirable to localize the scintillation observations as much as possible without dismissing too much of the useful data. To accomplish this, a minimum elevation of 35° was applied to the GPS data sets. In addition to localizing the observations (to a range of less than ±4.0° in longitude), this step also served to minimize the likelihood of multipath interference leading to false scintillation observations, a phenomenon often seen at low elevation angles. An additional constraint on this study is the variability of scintillation activity with magnetic latitude, particularly noticeable at L band frequencies. While the largest scintillation effects at L band are observed at stations near the equatorial crest (e.g., Antofagasta, Chile, ∼15° magnetic latitude (MLAT)) stations nearer the magnetic equator (e.g., Ancon, Peru, ∼0° MLAT) often report significantly reduced levels. For this reason, direct correlations of nightly scintillation activity between Antofagasta and Ancon, the stations with the most complete data sets available, provided no useful information. We therefore limited results shown here to three stations separated by approximately 45° in longitude and located between 5.2° and 11.5° of the magnetic equator.

[13] The longitudinal scintillation correlation tool was exercised with data collected at SCINDA stations located in Fang (Thailand), Manila (Philippines), and Guam providing three link pairs ranging from 21.8° to 45.5° separation in longitude (Table 1). Because the background noise level on the scintillation measurements varied from site to site, each station was assigned an individual scintillation threshold level for distinguishing “ACTIVE” and “INACTIVE” evenings, parameter 3 described in section 2.

Table 1. GPS Longitude Correlation Study
East LinkWest LinkLongitude Separation, degData Set

[14] A sample of the 31-night output detailing the night-to-night comparisons of data on the selected links is shown in Figure 1. Here, results are displayed for the Manila-Fang sensor pairing from nights 1–31 in 2002. The histograms show the number of hours of scintillation above the selected threshold observed at each station between 1.0 and 5.0 hours past sunset. In each case, the number of hours was determined by counting the number of 1-min data records in which the link to the GPS satellite with the “hottest” scintillation measurement (or highest S4) was above the station-independent scintillation threshold. Nightly histograms above the dashed line represent “ACTIVE” evenings and are indicated with diagonal lines while those below are labeled as “INACTIVE” and shaded gray. The negative value shown for evening 23 and represented in black is an indication of a case for which not enough hours of data had been collected for a valid comparison between the links. For this 31-night period, there were 30 evenings of data with a total “correlation score” of 20, meaning that on 20 nights, both sites were “ACTIVE” or “INACTIVE.” Conversely, during 10 of these evenings, one link was “ACTIVE” while the other was “INACTIVE.” The average correlation score for this period was 66.7%.

Figure 1.

Sample 31-night correlation results for 2002, Manila-Fang GPS.

[15] Compiling these statistics over an entire year leads to results as shown in Figures 2, 3, and 4 where yearly results from 2002 for the Manila-Fang station, Guam-Manila, and Guam-Fang pairs are shown. In each case, the running 31-night correlation averages are shown in the left plot with connected dots. The histograms in the right plot illustrate the number of simultaneous evenings of data between the two stations used for computing the correlation scores over each 31-night period. To weed out potentially invalid comparisons, a minimum of five evenings of data over each 31-night period were required according to the definition of parameter 2 in section 2 (shown as a dashed line). Negative values in this plot represent nights for which no correlation scores were computed because of a lack of data.

Figure 2.

GPS scintillation correlation results, Manila and Fang.

Figure 3.

GPS scintillation correlation results, Guam and Manila.

Figure 4.

GPS scintillation correlation results, Guam and Fang.

[16] The longitudinal separation of the station pair shown in Figure 2, Manila and Fang, is approximately 21.8°, while Guam is separated from Manila by nearly 23.7° (Figure 3). The distance between Guam and Fang (Figure 4) is approximately 45°. It is apparent from these results that the correlation of the nightly scintillation over these longitudes is dependent upon the time of year. While not shown here, when comparing these results to a plot showing the 5-min averaged scintillation levels over the entire year on a nightly basis, it becomes evident that the highest correlations fall within the distinguishable “ON” or “OFF” seasons while the lesser correlations are observed during the “TRANS” seasons.

[17] It was at this point that our focus was changed to VHF scintillation data recorded on SATCOM links to geosynchronous satellites. Because of the limited data set and our inability to precisely determine the longitudinal separation of the links when using ground-based GPS observations, we did not apply these L band results to our final results shown below; however, they were used as a check on the VHF-based correlations. While the correlations from the Guam-Manila and Manila-Fang pairings fit quite nicely with results from the VHF station pairings, the Guam-Fang correlations made over 45° of longitude did not.

4. Longitudinal Scintillation Correlation Study With VHF Data

[18] Exercising the tools developed for the examination into the correlation of scintillation activity over various longitudinal separations, our attention was turned to sets of archived VHF S4 data (∼250 MHz) collected on the SCINDA network. The configurable parameters 1 and 2, described in section 2, were again left at 1.0 to 5.0 hours past sunset, requiring a minimum of 3.5 hours of data for valid evenings. The parameters defining “ACTIVE” evenings of scintillation, (3 and 4), were varied as will be discussed here.

[19] The stations providing data for this study are located in Ancón (Peru), Antofagasta (Chile), Guam, and Manila (Philippines). With two VHF sensors at Ancón and Antofagasta, one looking to the east and the other monitoring scintillation levels on a SATCOM link to the west, this set of four stations provided a total of five link pairings ranging in longitudinal separation from approximately 5.97° to 29.94°. A list detailing these pairs and the longitudinal separations of the 300 km ionospheric penetration points is given in Table 2. ANCW, ANCE, ANTW, and ANTE represent the east and west links at Ancón and Antofagasta, respectively. Likewise, GUM and MNL are used to represent VHF links at Guam and Manila.

Table 2. UHF Longitude Correlation Study
East LinkWest LinkLongitude Separation, degData Set

[20] As was done with the GPS study, night-to-night correlations of scintillation activity over known longitudinal distances were scored using the algorithm described in section 2. This time, however, the nightly correlations were binned into separate “ACTIVE” and “INACTIVE” categories on the basis of the level recorded on the eastern (or downstream) link prior to the computation of the monthly averages. For example, the results shown in Figure 5 with 31-night correlations of scintillation activity from 2001 with ANTE/ANTW for the evenings of 241–271 are similar in format to those shown in Figure 1, with the exception that they are now binned by “ACTIVE and “INACTIVE” activity levels on ANTE. In this case, the configurable parameters 3 and 4 were set to require 1 hour of S4 above the level of 0.6 in order for an evening to be labeled as “ACTIVE.” Only those evenings in which less than 1 hour of scintillation above this threshold (“INACTIVE,” gray histogram) are included in the results shown on the left-hand side of Figure 5 while active nights (“ACTIVE,” diagonal-hatched histogram) at ANTE are compiled into the results displayed on the right. Here, you can see the reasons for sorting the results in this way. These results could be interpreted to say that for this 31-night period, when one ANTE was “INACTIVE,” on the basis of our selected definition, there was an 81.2% chance of an “INACTIVE” night at ANTW, nearly 6° in longitude upstream. However, on evenings when a significant amount of activity was detected at ANTE, the measured correlation in the observed scintillation levels was significantly lower at only 55.6%.

Figure 5.

VHF 31-night correlations for ANTE-ANTW binned by activity observed on ANTE. “INACTIVE” is on the left and “ACTIVE” is on the right.

[21] After separating the VHF correlation statistics into seasonal (ON/OFF/TRANS) results, these were further binned by longitudinal separation. Seasonal correlations from the three separate link pairs separated by nearly 6° in longitude were averaged together and were included with results from the ∼12° and ∼30° pairings in a plot of the correlation percentage versus longitudinal separation. An exponential curve fit provides coefficients representing the statistical percentage of correlation as a function of longitude, season, and activity level.

[22] As described earlier, the configuration-driven code developed for this study required the selection of parameters defining “ACTIVE” and “INACTIVE” conditions, parameters 3 and 4. Selections were made independently for each link. If one were to require, say, 1 hour of S4 above 0.6 on the east link to define an “ACTIVE” evening, correlations with scintillation observed on the upstream (western) link could be determined at an equally severe level of 0.6 or, if desired, at a less severe level of an hour above 0.3. Fully exploiting such a tool for every possible level of activity on each link would, in return, provide a climatologically valid algorithm driven by real-time data observations providing the probability of scintillation activity at any specified level over a range of longitude.

[23] Results from two specific cases are provided in Figures 6 and 7. In each case, samples of the seasonally dependent longitudinal correlations are broken into “ACTIVE” (solid circles, solid lines) and “INACTIVE” (open circles, dashed lines) evenings. In Figure 6, we exercised the method described here for rather moderate levels of activity on both links, requiring only 30 min of scintillation above a threshold S4 level of 0.3. On the left are exponential-fit curves to the average correlations returned during the “ON” season. The “TRANS” season results are presented in the middle plot, while those from the “OFF” season are displayed in the right-hand plot. The set of curves shown in Figure 7 are presented in the same format as described for Figure 6 and were made with from results with runs where “ACTIVE” evenings were more strictly defined, this time requiring a full hour of S4 observations on the eastern link above 0.6 and on the western link above 0.3.

Figure 6.

Seasonal correlation curves: half-hour S4 > 0.3 on east, >0.3 on west. Solid lines and filled circles represent “ACTIVE” evenings. Dashed lines and open circles represent “INACTIVE” evenings.

Figure 7.

Seasonal correlation curves: 1 hour S4 > 0.6 on east, >0.3 on west. Solid lines and filled circles represent “ACTIVE” evenings. Dashed lines and open circles represent “INACTIVE” evenings.

[24] A noticeable feature highlighted by these results is the comparatively poor correlation of “INACTIVE” evenings during the “ON” season and likewise for “ACTIVE” nights during the “OFF” season. The instabilities triggering the rare “ACTIVE” nights during the “OFF” season appear to be rather localized in longitude, even over a distance of only 6° or less than 700 km. You will notice in Figure 7 that there is no 30° separation point displayed for “ACTIVE” evenings during the “OFF” season. This was due to the fact that there were no nights observed on the VHF receiver at Guam that met the criteria described for “ACTIVE” evenings for this run (60 min of S4 observations greater than 0.6).

[25] Similar sets of curves could be generated ad infinitum for a wide range of scintillation activity levels. Taken together, they could be used to provide probabilities of various scintillation intensities as a function of longitude from a fixed observation point. Such a tool would be particularly beneficial in offering tailored specifications of the scintillation environment to the west (upstream) of real-time data links, a region in which no current data-driven information is provided. A future study on the temporal dependence of these correlations should serve to further expand the effectiveness of this algorithm as would the inclusion of a dependency on solar and geomagnetic conditions.

5. Conclusions

[26] An analysis was performed on the correlation of scintillation activity over variable longitudes in the equatorial region for the purpose of providing a reliable predictor of scintillation upstream (westward) from observations. Nightly correlations between selected SATCOM link pairs were averaged over monthly periods, then binned by activity level and season, resulting in sets of curves offering expected correlations of various scintillation intensities as a function of longitude. Higher correlations were observed for “ACTIVE” nights during the “ON” season and “INACTIVE” nights during the “OFF.” The relatively low correlation noted for “ACTIVE” nights during the “OFF” season, particularly when using a comparatively strict definition for active conditions, is an indicator that the instabilities leading to F region irregularities are localized in nature. With the inherent limitations on GPS data in this statistical study, only fixed-link VHF measurements were used in the final analysis; however, the longitudinal correlations from GPS sensors with longitude separations of ∼21° and 23° in longitude fit rather well the VHF results. Not surprisingly, the correlations seem to break down over 45° of longitude as the differences in seasonal onset over that distance become apparent.

[27] When performing a detailed comparison of scintillation levels on various SATCOM links to geosynchronous satellites, it is important to take into consideration possible differences in the observed levels due to geometry effects. An observer detecting scintillation on a link directly overhead might expect to record lower S4 levels than a second observer communicating through the same structure but at a lower elevation angle. Rino [1979] defines the parameter CsL as the height-integrated strength in the irregularity spectrum at a scale size of 6.28 m (2π m). This parameter was rescaled to 1km and was renamed CkL for use in WBMOD [Robins et al., 1986; J. A. Secan, private communication, 2006]. CkL is a frequency-independent indicator of the electromagnetic scintillation irregularity strength. By converting measured S4 levels to an equivalent CkL using the geometry of the observing link, then converting back to an S4 at a fixed look angle, we are able to remove the effects of geometry when making comparisons of activity levels between two SATCOM links. While the results shown in this paper were achieved by direct comparisons of S4 values, this study was repeated with S4 values first being converted to CkL then back to S4 for SATCOM links with identical geometry. Because of our use of thresholding in defining “ACTIVE” and “INACTIVE” conditions, the resulting correlations were nearly identical to those presented here with the exception of the Guam-Manila link pairing for which not enough data were available for a valid statistical comparison during certain seasonal periods. With this in mind, it would be worthwhile repeating this study in the future with data from the expanded SCINDA data network.


[28] The work at AER/Radex, Inc., was supported by AFRL contract F19628-00-0089.