In order to provide a more reliable situational awareness of scintillation impacts on users of space-based communication/navigation systems, L-band scintillation measurements from GPS satellites are ingested, in real time, to the Scintillation Network Decision Aid (SCINDA) model. SCINDA is a real-time, data-driven communication outage forecast and alert system developed by the Air Force Research Laboratory at Hanscom Air Force Base and made operational through a joint effort between the U.S. Air Force and the U.S. Navy. UHF and L-band scintillation parameters are measured, modeled, and propagated in time to provide a regional specification of the scintillation environment in an effort to mitigate the impacts on the satellite communications (SATCOM) community. In an effort to provide UHF SATCOM users with a more consistent estimation of scintillation impacts on their systems, GPS sensors, measuring S4 at L-band frequency, are now used as a supplement to the stationary UHF links dramatically increasing the regional coverage at each individual station, particularly those located outside the anomaly crest. Three-dimensional representations of GPS observed plumes are mapped to the ground from the location of a selected satellite, producing a “scintillation specification map detailing “outage” regions on the globe. In this paper we present results from a validation study of the GPS proxy model and demonstrate its usefulness as an adjunct to the SCINDA model in producing the most accurate and reliable nowcast and forecast scintillation specification products available to UHF SATCOM users.
 The Scintillation Network Decision Aid (SCINDA) was developed as a proof-of-concept tool to provide UHF satellite communications (SATCOM) users a situational awareness of the near-space environment factors that may inhibit their ability for clear communication. In 2000, a SCINDA-based product was made operational as part of the Operational Space Environment Network Display (OpSEND) [Bishop et al., 2002], currently running at the Air Force Weather Agency at Offutt Air Force Base, Nebraska. UHF scintillation specification maps are produced in real time and made available to SATCOM users via a secure Web interface. The products have proven to be a valuable resource for communicators in the detection or, as we call it, “nowcasting” of equatorial transionospheric scintillation.
 The system works by listening continuously to geostationary beacon satellites at UHF frequencies (∼250 MHz) from ground stations positioned at various strategic points around the globe [McNeil et al., 1997; Groves et al., 1997]. When scintillation is detected along a particular link, a model of the scintillation structure giving rise to the scintillation, often referred to as a “bubble” or “plume,” is invoked. This model projects the scintillation up the magnetic field line to the mirror point, the point on the other side of the geomagnetic equator at the same magnetic latitude, providing a latitudinal dimension to the structure. The longitudinal extent is determined through the incorporation of zonal ionospheric drift velocity measurements from a spaced-receiver configuration along with the time between detection and cessation of scintillation activity on the link. The structures themselves move eastward with the zonal drift and decay in intensity at rates derived from detailed studies of the historical data. These model structures are then projected to the ground for a specific satellite, resulting in what we call a scintillation specification map which describes the situation from any latitude and longitude in terms of “yellow” or “red” conditions, for moderate and severe communication disruption, respectively. A detailed description of the model follows in section 2.
 In spite of the success of SCINDA, there are limitations to its capabilities. Recording scintillation on a signal from a geosynchronous satellite at UHF frequency limits the monitoring to the line of sight between the sensor and the stationary satellite. Since the ionospheric drift is west to east during the period when equatorial scintillation is observed, this limitation prevents the detection of structures that develop to the east of the ionospheric penetration point (IPP) to the geosynchronous satellite. The IPP is the intersection of the station to satellite line of sight with the altitude where the scintillation disturbance is most intense. Additionally, structures that may exist to the west of this point cannot be included in a SCINDA product until they have been observed to drift across the link between the antenna and the satellite. Furthermore, the location of a particular SCINDA station relative to the geomagnetic equator can prevent the detection of all but fully developed, severe scintillation structures. Stations positioned at, say, 18° off the geomagnetic equator may miss newly developing plumes that extend to perhaps 10° from the geomagnetic equator. These limitations, taken together, tend to underestimate the potential for communication problems across a particular theater.
 It has been a long-time goal of this effort to minimize these limitations and extend the SCINDA communications outage predictions to longitudes farther from the actual station location [Caton et al., 1999]. This was accomplished in the latest upgrade of the model delivered as part of the OpSEND suite of products by incorporating scintillation data from GPS sensors measuring S4 at L-band frequency as a supplement the stationary UHF receiver links. The greatly enhanced coverage area allows for the detection of plumes that may be missed by UHF links, particularly plumes developing to the east of the IPP to the UHF geosynchronous satellite and equatorward in latitude of stations located a significant distance off the geomagnetic equator. Because the GPS L-band data are used to provide UHF SATCOM scintillation maps, this upgrade is referred to as a GPS “proxy” model for UHF scintillation.
 The existing SCINDA network consists of 18 equatorial data collection sites. Of these, real-time data feeds are currently received from 12 locations for use in the generation of products. Each of the real-time sites consists of both UHF and GPS sensors, while GPS-only measurements are made at the remaining sites. Since the GPS proxy model can supplement and even replace an actual UHF sensor, these GPS-only sites can be incorporated into the scintillation specification map product when they come online.
 The benefits of the GPS proxy model can be quantified geographically by looking at the expanded coverage area from a particular location. Assuming that peak scintillation occurs at an altitude of 350 km, and assuming an elevation cutoff of 10° for use with GPS data to eliminate multipath errors (discussed later), the GPS sensors can “see” about 9° in longitude to the east and west of the station. More important, for a station located farther than 10°–12° off the magnetic equator, such as Ascension Island, a GPS receiver can detect scintillation from 6° to 9° in latitude north and south of the station. These numbers depend in large part on the geographic latitude of the station. For example, at the SCINDA data collection site in Pontianak, Indonesia (located near geographic equator), the use of GPS adds approximately 9° of coverage in all directions while only 6° of coverage is added to the south at Ascension Island. This extra 6° is sufficient and critical, though, in detecting plumes that do not rise as high as Ascension's 18° magnetic latitude.
 This is a considerable addition to the existing SCINDA stationary links, even those within the anomaly crest region, since, as previously mentioned, plumes moving in from the west can be detected earlier. At a typical early evening zonal drift velocity of 125 m/s, a plume on the move could be detected approximately 1 hour before it reached a stationary overhead link to a station. Another fact to consider is that scintillation structures tend to form beginning at a fixed local time, about 2 hours past local sunset. Since GPS sensors can see up to 9° to the east of a station, they sample later local times and therefore can detect the formation of plumes to the east of the station at an earlier point in time. This 9° in longitude adds more than 30 min to the lead time for plume detection, assuming all plumes begin to form at a fixed time after local sunset. The extension of the detection area both in latitude and longitude (time), then, is the primary motivation in introducing the GPS proxy model into the UHF version of SCINDA.
2. How the GPS Proxy Model Works
 The first step in generating scintillation structures from GPS data is the editing out of spurious S4 values caused by multipath interference at low elevation angles. This is accomplished by the application of a station-specific “mask” to eliminate data according to elevation and azimuth angle. In spite of careful placement of the GPS antenna, at some stations the mask must be set as high as 20° at all azimuth angles and certain azimuth ranges may require masking up to as high as 40° due to nearby obstructions such as large antennae or buildings. At other stations, a simple 10° mask suffices. Once the data have been masked, it is then sorted by time and by the identifying pseudorandom noise (PRN) number of each satellite. The azimuth and elevation recorded by the sensor is converted to a geophysical latitude and longitude of the IPP at 350 km altitude. Our studies have indicated that the most intense scintillation occurs in the altitude range 300–350 km. Of course, the scintillation plumes themselves, which are pinned to magnetic field lines, extend as high as perhaps 1500 km. However, it would seem that the 300–350 km range is what is important as far as the scintillation intensity is concerned [McNeil and Caton, 2000]. This number was arrived at based on results from the aforementioned SCINDA validation effort conducted in 1999 where a 350 km topside provided the most accurate cessation times on the eastern link of Antofagasta based on propagation of data collected on the western link. In support of this, theoretical models of scintillation view the interfering medium as a two-dimensional “phase screen” [Rino, 1979] so that the idea of a thin model around 300 km is in keeping with past formalism. We have chosen the upper limit of this range in order to give the GPS “bubbles” as large a physical dimension in latitude and longitude as is reasonable, in keeping with the objective of erring on the high side as far as scintillation is concerned.
 A three-point smoothing is carried out on the data from each PRN on the recorded 1-min S4 levels. A plume recognition algorithm then examines the data where a threshold S4 of 0.2 is used to signal the existence of a scintillation bubble. It is possible to use a threshold this low because the noise level of the GPS units is actually quite low, excepting multipath interference, which is effectively removed by masking. A plume is flagged when two successive S4 values are detected above the threshold. Likewise, a plume is terminated when two successive points are found below the threshold. Since our GPS receivers routinely lose lock above a scintillation index S4 of 0.6, the code allows for a bubble to persist for up to 1 hour through a data gap, provided that there are at least two points on either side of the gap with an S4 index above the threshold. On the other hand, to reduce spurious bubbles, we require that all bubbles be at least 15 min long in GPS detection time.
 After the completion of the plume recognition algorithm, the longitudes of the IPPs at the “edges” of the plumes are propagated according to a zonal drift velocity derived from the Richmond (National Center for Atmospheric Research) drift velocity model [Richmond et al., 1980]. The “edges” are defined by the IPPs of the first and last contact of the GPS link with scintillation. These are drifted eastward, according to the model from the time of detection to the “valid time,” which may be up to 3 hours ahead of the current map production time in the case of a forecast map. At this point, the north/south extent of the bubble is set in geomagnetic coordinates based on whichever one of the edges extends farther from the geomagnetic equator. This means that GPS-detected plumes essentially stop where they cease to be detected since no information exists as to whether they extend farther from the equator than the detection point. As opposed to the plumes based on the UHF scintillation data, GPS-detected plumes, as modeled in SCINDA, are not allowed to grow in north/south extent. Each plume is traced from this point, along the magnetic field line, to the mirror point on the opposite side of the geomagnetic equator, giving the plume its latitudinal extent. The bottomside of the bubble is assigned an altitude of 300 km while the topside is assigned a maximum value of 600 km if the actual height of the bubble at any point is greater than this value as it is traced along the magnetic field line. As mentioned, although we believe that most of the scintillation intensity is confined to the 300–350 km range, we have chosen to make the topsides of the GPS bubbles somewhat higher so that they project to more expansive structures on the scintillation specification maps. This again is in keeping with our desire to overestimate the scintillation, if we err at all.
 With the limits of each plume defined, an intensity value is assigned based on an approximation of the effects on UHF communications. It is well known that scintillation at GPS frequencies is considerably less in terms of the scintillation index S4 for the same structure than it is for UHF scintillation. To assign a UHF scintillation level, we examine the average value of the GPS S4 measurements during the detection period. If this is greater than 0.5, an initial UHF scintillation level of 0.9 is assigned to the structure. For values above the lower threshold but less than 0.5, an equivalent UHF magnitude of 0.8 is used. These values correspond to the scintillation level at a magnetic latitude of 15°, near the anomaly crest. The latitudinal variation is adapted from the crest model in the wideband model (WBMOD) [Fremouw and Secan, 1984; Secan et al., 1995]. The particular version of the crest implemented in the most recent version of the code is essentially flat from the equator down to 15° magnetic. The intensity then falls off relatively sharply, becoming negligible by 20°. This same crest model is used in the SCINDA model for plumes detected by the stationary UHF links.
 For GPS-observed plumes detected prior to local midnight, the scintillation level is maintained at peak intensity for 1 hour before the level is reduced exponentially with a time constant of 3 hours. For plumes detected after local midnight, decay begins immediately, with a time constant of 1 hour. These values were also adopted from the SCINDA model for the stationary UHF links and were determined from analysis of 7 years of UHF scintillation data.
 To produce the resulting operational product, both GPS and UHF detected plumes are projected to the ground for a specific satellite with accommodation for the variations of S4 with geometry and angle to the magnetic field. These are referred to as “shadows”. The shadowed areas provide a user with a scintillation specification map for a particular geographic theater, satellite, and time. Because the products are produced to supply UHF scintillation specifications to the SATCOM community, shadows resulting from scintillation observed on UHF links are displayed overtop the GPS-based shadows in deference to the fact that UHF measurements are more reliable in terms of scintillation intensity than are the GPS measurements.
3. Inherent Limitations in the GPS Proxy Model
 There are certain limitations involved in using L-band GPS scintillation measurements to generate structures intended to represent UHF outages. These limitations can be understood by considering the nature of equatorial scintillation at these two frequencies. While it is generally true that the presence of structures giving rise to scintillation at GPS frequencies implies scintillation at UHF frequencies, the converse is not always true. For example, scintillation at GPS frequencies follows UHF scintillation more often than not during solar maximum periods. However, at solar minimum, UHF scintillation persists, albeit with somewhat lower probability, while GPS scintillation is virtually absent [Basu et al., 1988]. We will also show that the seasonal behavior is strikingly different at UHF and GPS. This is to say that strong UHF scintillation may be obtained along a particular link, while a transmission along that same link at GPS frequencies may be affected little or not at all. The use of GPS data in mapping UHF structures therefore can generate only a subset of all possible structures. This is truly a subset of the UHF structures present, since UHF scintillation is present whenever GPS scintillation is present. More simply, it is possible to tell when a UHF link should be scintillating through use of GPS measurements; however, it is not necessarily possible to tell when a UHF link is not scintillating judging only by the absence of scintillation along a GPS link. This limitation should be understood by all users, particularly in circumstances where UHF measurements are not available. On the other hand, for reasons mentioned in the previous sections, the use of GPS is certainly better than providing no information at all.
 The seasonal and diurnal behaviors of the UHF and GPS scintillation are known to differ. This can be seen in Figure 1, where we show the fraction of time that the S4 parameter exceeds selected levels for both UHF and GPS as functions of both day of year and time, measured from local sunset. These data are from Antofagasta, Chile, in 1999. We have chosen a scintillation index S4 level of 0.6 for the UHF and 0.3 for the GPS in order to bring out the behavior, but this choice is not critical for defining the season or diurnal behavior. The choice of 0.6 as a threshold for the UHF scintillation does not affect either the diurnal or seasonal behavior to a large extent because the distribution of UHF S4 is bimodal; that is, the value is either near zero or saturated near unity. The choice of a higher threshold for the GPS S4 would only serve to amplify the differences in season and local time, since both are shorter for GPS than for UHF. We see that there are two important differences between the behavior of scintillation at UHF and GPS frequencies, at least at this station. Although both show peak scintillations near equinox with nearly nothing in the summer months, the scintillation at UHF continues and even appears to intensify through the winter solstice. The GPS scintillation, on the other hand, shows the peaks at the equinoxes with very little activity at winter solstice. We can therefore say that the GPS model would add little information to a UHF scintillation specification map during winter solstice period, at least at Antofagasta. This would certainly not mean that UHF scintillation was not taking place, only that the GPS links were not picking up this scintillation. This interesting comparison suggests some difference in the scintillation structures and/or the ionosphere itself between equinoxes and winter solstice that we believe is not completely understood.
 Also noticeable in Figure 1 is the fact that UHF scintillation above the selected thresholds tends to last longer, perhaps by an hour or two, than does scintillation at GPS frequencies. This is likely due to the difference in decay rates of structures giving rise to UHF and GPS scintillation. From this information, it should be obvious that the GPS-based model cannot predict UHF scintillation as late into the evening as well as a set of corresponding set of UHF links. Since such links are not available, the extra information provided by GPS is of considerable importance, though, obviously.
 An additional concern in the application of the GPS proxy model is the dependence on the solar cycle in the severity of scintillation at L-band frequencies. Although UHF scintillation generally occurs with regularity at solar minimum as well as at solar maximum, our own data, which now span one half a solar cycle, as well as previous studies [Basu et al., 1988] suggest that scintillation at GPS frequencies is nearly absent at solar minimum. Because of this, the GPS portion of the SCINDA model is not expected to be valuable during solar minimum conditions. Even still, the use of the GPS proxy model at solar minimum in no way detracts from the resulting scintillation specification map, since the bubbles from the fixed UHF links will still be displayed.
4. Operational Impact
 How valuable is the GPS proxy model in providing a real-time scintillation specification to UHF SATCOM users? The example displayed in Figure 2 was made with data collected at the SCINDA stations located in Bahrain and Diego Garcia (shown as black diamonds) on an evening in November 2000, a typical night during the scintillation season. The magnetic equator is shown as the heavier dashed line in the center of the images while the lighter dashed lines represent the anomaly crests, where scintillation is expected to peak, at ±15° magnetic latitude. Scintillation plumes detected by the in-theater UHF sensors are shown in the frame on the left while structures detected on the GPS sensors are shown on the right. The darker (red) features indicate “severe” scintillation areas where the modeled S4 values are greater than 0.6 while the lighter (yellow) areas indicate “moderate” disturbances with S4 values between 0.3 and 0.6. Although these thresholds are to some degree arbitrary, they are derived from actual experience in the field where scintillation measurements have been made concurrently with attempts at satellite communication (K. M. Groves, personal communication, 2000).
 Because the Bahrain and Diego Garcia sensors are located at 19° and 15°, respectively, off of the magnetic equator, UHF links to the geosynchronous satellite used for scintillation measurements often do not “see” plumes that have not yet grown or never will grow to reach these latitudes. This situation is especially severe for Bahrain, where UHF scintillation on the fixed link is a relatively rare occurrence. Additionally, with an approximately 22° longitude separation between the stations, on an evening when simultaneous UHF scintillation is observed at both locations, one might expect plumes to extend throughout the entire longitudinal region. In the case shown here, the gap in the observed UHF structures, at the same longitude as Diego Garcia, is partially filled in by GPS-observed plumes. This plume has apparently not extended far enough south to be detected at Diego Garcia. This is also the case at Bahrain, where we see a narrow plume in the GPS model directly south of the station that does not show up in the UHF measurements. The redundancy in the GPS-observed plumes in longitudes where the UHF links also detected scintillation is a quite comforting validation of the usefulness of the proxy model in situations where UHF data are not available. With the increased coverage from the GPS constellation, the chances for observing these structures, if they exist, are greatly enhanced. Although the GPS proxy model was originally developed for stations at high magnetic latitudes, the model has been incorporated for use at all SCINDA stations whenever GPS data are available.
5. Validation of the GPS Proxy Model
 Because a comprehensive discussion of the validation effort carried out on the GPS proxy model for use in the Scintillation Network Decision Aid is not possible in the space available here, the focus of this section is primarily on the phase of the study involving comparisons of the GPS proxy model with images from an all-sky imager at Ascension Island during a campaign in late March to early April of 2000 observing airglow at 6300 Å, an example of which is shown on the left of Figure 3. Depletions are indicated by dark patches in the emission lines for recombination of O2+, which further indicates an absence of plasma density in these regions. Scintillation plumes are thus detected by the depletion of plasma density within them. Positions of the depletions in the all-sky images indicate the relative azimuth and elevation with respect to the station. In making the comparisons, results from the GPS proxy model were evaluated and then converted to azimuth/elevation coordinates using a standard height of 350 km. Again, this altitude is where we expect the scintillation to originate for the most part.
 Comparisons from the evening of 27 March 2000 are shown in Figure 3. In each image the center represents the area directly overhead of the station. Observed plumes will drift eastward or right to left across the picture. Images from the all-sky camera are on the left-hand side, while the results from the GPS proxy model are shown in the same azimuth/elevation coordinates on the right. In the top panel, produced at 2130 UT, a depletion is seen almost directly over the station and perhaps slightly to the east (left-hand side). The GPS proxy model has produced two bubbles very near the azimuth and elevation of this structure. The model result of “two” plumes is actually just one plume being detected by two different satellites. This, i.e., the simultaneous detection of plumes of the same dimension by two satellites, is validation in and of itself. Depending on the track of the particular satellite, the longitudinal extent of the detected plume may vary. In this example, both satellites have detected the east (left) edge of the plume, but the satellite intersecting the plume at lower latitudes produces a plume that does not extend as far to the west as the plume with the south edge at a higher latitude. This redundancy in the model is somewhat comforting, since the most intense plumes are generally detected by more than one satellite.
 Half an hour later, at 2200 UT, a depletion is seen developing slightly south and west (right) of the station in the all-sky image. Two different GPS satellites also detect this plume. By 2230 UT (not shown) the plume to the far west in the all-sky image has combined with the plume nearly overhead, producing an extremely wide structure. Again, the GPS model results track this feature well in showing a considerably widened plume. Similar results were obtained from data recorded by the all-sky images on additional evenings.
 The bulk of the validation, which is given in detail by McNeil and Caton , was done through extensive case studies. The model is suited to case studies rather than statistical analysis because, as we have explained earlier, there is not a one-to-one mapping of the UHF and GPS scintillation in every case. High magnetic latitude stations with a fixed UHF link may miss the UHF scintillation present, but GPS may pick it up. On the other hand, GPS scintillation may simply be absent although scintillation in the UHF is observed. There is one case where we can calculate a statistical parameter, and this is for our station at Antofagasta, Chile. Because this station is near the anomaly crest at 11.1° magnetic, its fixed UHF link generally observes scintillation whenever it is present. We can ask, then, In how many cases where we observe GPS scintillation do we also observe UHF? For the solar maximum year of 1999 (with sunspot number 90) there were 85 nights in which an hour or more of scintillation was observed with the GPS sensor. Of these nights, 84 also showed scintillation in the fixed UHF link. This gives us a very high correlation of 99%.
 The effectiveness of the SCINDA model, which provides a situational awareness of equatorial scintillation to users in real time, has been considerably enhanced by the implementation of a GPS proxy model for UHF scintillation. Although recorded at L-band frequencies, GPS scintillation measurements have proven extremely useful in providing additional coverage for the production of UHF scintillation specification maps. With an understanding of the limitations involved in using data at GPS frequencies to represent conditions at UHF frequencies, the recent upgrade to the SCINDA model is considered to be a significant advancement over previous versions.
 Work by Radex, Inc., on this project was funded through contract F19628-98-C-0054 with the Air Force Research Laboratory's Space Vehicles Division. The work at AFRL was partially supported by AFOSR grant 2311AS.