The first observations of large-amplitude GPS L1 amplitude scintillations in the midlatitude ionosphere were made on 25–26 September 2001. The scintillations, which were intense at times (≥20dB, S4 ≈ 0.8), occurred during the main phase of a moderate geomagnetic storm (Dst = −100 nT) when a storm-enhanced density (SED) event coupled with a broad and structured ionospheric trough occurred over the eastern United States. The preliminary results, using a modified single-frequency GPS receiver and a dual-frequency (L1/L2) GPS receiver and measuring total electron content at Cornell University (53.2°N magnetic latitude), have been published previously. In this effort we examine the scintillation autocorrelation functions using data from the fast sampling (50 Hz) single-frequency GPS receiver. The intention is to derive a measure of the scintillation pattern velocity which can be estimated from the Fresnel radius and the width of the autocorrelation function. We show that a minimum in the autocorrelation function width, corresponding to a maximum in the velocity, is located near the equatorward boundary of a broad trough. We also demonstrate that estimates of the scintillation pattern velocity are consistent with other measures of ionospheric drift in midlatitude storm time convection. In at least two of the GPS signals the inferred velocity has narrow peaks of several hundred meters per second, characteristic of subauroral ionospheric drifts (SAID) and perhaps driven by plasma pressure gradients in the inner magnetosphere. This presentation, along with the paper by Ledvina et al. , strongly demonstrate that the SED and associated SAID can cause debilitating scintillation levels for GPS receivers.
 Radio wave scintillation investigations of the ionosphere are one method for sensing ionospheric irregularities primarily in the E and F regions [Yeh and Liu, 1982]. Aarons et al.  investigated equatorial ionospheric scintillations at VHF to determine the presence of ionospheric irregularities. In the auroral regions, scintillation studies provided further evidence of the ability to detect irregularities in the ionosphere [Basu et al., 1983; Coker et al., 1995; Aarons and Lin, 1999]. A comprehensive look at the fundamentals of ionospheric scintillations can be found in the work of Yeh and Liu .
 Radio wave scintillation in the ionosphere can be viewed as wave propagation in a random media. In general, the ionosphere can be thought of as a continuous region of relatively smooth plasma density occupying roughly 70–1200 km above Earth's surface. However, under certain circumstances, most frequently in equatorial and in auroral regions, localized structuring of electron density can occur. This electron density structuring is commonly denoted as irregularities, patches, or bubbles. The presence of these structured irregularities can affect radio waves that propagate through them both diffractively and refractively [Yeh and Liu, 1982].
 The midlatitude ionospheric region is generally considered to lack the mechanisms required to produce scintillations, unlike the scintillation-encouraging equatorial and auroral regions. Recent works have indicated that the this general claim of an inactive midlatitude ionosphere may be a poor assumption [e.g., Fukao et al., 1991; Kelley et al., 2000; Makela et al., 2000; Swartz et al., 2000]. Furthermore, from Foster [1995, and references therein], it is known that the midlatitude ionosphere, at latitudes corresponding to the northeastern United States, is subject to F region electron density structuring owing to space weather effects of penetrating magnetospheric disturbance electric fields. More recently, Foster et al.  have demonstrated the existence of large-scale ionospheric intrusions over the United States during local evenings that begin with storm-enhanced densities at the southern border of GPS chain observations and move northward, perhaps into the polar cap. These recent works suggest that the notion of an inactive midlatitude ionosphere should be re-examined.
 It is well known that the nighttime F region trough moves equatorward during times of increased geomagnetic disturbances producing scintillations on both radio stars and satellite signals at HF and VHF frequencies [see, e.g., Kersley et al., 1972; Nichol, 1973; Kersley et al., 1975; Schunk et al., 1976]. Further studies have shown that the edge of the ionospheric trough, where density gradients are typically the largest, causes VHF and UHF scintillations [Bowman, 1991; Bishop et al., 1993]. Within the trough the ionosphere may exhibit very fast flows, called subauroral ion drifts (SAID) by some authors and subauroral polarization streams (SAPS) by others, that reach speeds in excess of 3 km/s [Reddy and Mayr, 1998]. These are among the fastest flows observed anywhere in the ionosphere, far exceeding those observed at equatorial latitudes, and scintillations observed in this environment should have different temporal properties than those observed at equatorial latitudes.
 We use the term “scintillation” to refer to the diffractive effects of irregularity structures in the ionosphere. In order for scintillations to occur, a wave must traverse through structures on the order of the Fresnel radius, , where λ is the signal wavelength and z is the mean distance from the structures to the observer. The observable effect is a time-varying diffraction pattern in amplitude (or intensity) and phase projected on the ground, correlated both meridionally and zonally. This diffraction pattern traverses the ground with a velocity that is related to both the satellite velocity and the velocity of the irregularities [Kintner et al., 2001]. In terms of the Global Positioning System the signal sources are the GPS satellites and the observation equipment is a stationary GPS receiver on the ground. The scintillations are fluctuations in amplitude and phase which are measured and recorded by the GPS receiver.
 The same processes that cause scintillations can also cause disruptions in communication and global navigation systems that rely upon satellite communications. For instance, loss of lock of a GPS satellite can occur under strong scintillations [Kintner et al., 2001] or satellite phone usage may become unavailable [de Paula et al., 1999]. Disruptions of these types, especially in regions of large population density, can have adverse effects on the perceived reliability and dependability of such systems.
 Previously, Ledvina et al.  presented the first evidence of intense GPS L1 scintillations at midlatitudes. This observation was a fortuitous “accident” in that the instrumentation used for the observation in Ithaca was being tested prior to being deployed to Haleakala, Hawaii. Consequently, measurements of scintillation amplitude and total electron content (TEC) are only from single receivers. Drifts between receivers were not measured because there was only one fast sampling scintillation receiver operating. Instead, to estimate ionospheric drifts, we use a different approach by examining the width of the scintillation amplitude autocorrelation function. This paper is organized in the following manner. First, the techniques and equipment used to measure L band scintillations and the path integrated electron density are described. Then the observations of amplitude scintillations, total electron density, and signal power autocorrelation function width are presented. The ionospheric velocity is estimated using the autocorrelation function width and is compared with recent previous works. Finally, the last section consists of concluding commentary.
 Using a single-frequency (L1) GPS receiver modified at Cornell University [Beach, 1998; Beach and Kintner, 2001] and further modified to run in Linux [Ledvina et al., 2000], we recorded the signal power (at 50 Hz) of the GPS satellite signals. The signal power is the sum of the squares of the in-phase and quadrature components of the GPS signal. To avoid confusion, we will use the term “amplitude scintillations” to refer to the fluctuations in the measured signal power. As an additional measurement tool, a NovAtel dual-frequency (L1/L2) GPS receiver was used to determine the path-integrated electron density or TEC. TEC is determined by computing the relative code and phase delays on the L1 and L2 signals [e.g., Lanyi and Roth, 1988]. Typically, 6–8 GPS satellites are in view at any given time. All measurements were recorded at Cornell University, located at 42.4°N latitude and 283.5°E longitude (53.2°N magnetic latitude and 359.5°E magnetic longitude) during the evening of 25–26 September 2001.
 Numerous approaches are available for quantifying the level of scintillation intensity; the most popular and accepted being the S4 index. The S4 index is defined as the unity-normalized standard deviation of the received signal power. Physically, it gives a measure of the standard deviation of the time rate change of electron density fluctuations as seen by an observer. During the evening of 25–26 September the S4 index, which is satellite-dependent, reached a peak value of approximately 0.8. Figure 1 shows the S4 index for four different satellites over a period of several hours on 25–26 September, with magnetic Kp = 6, and 19–20 September, with the Kp = 2, a relatively quiet evening.
 Each panel corresponds to a single satellite and contains two traces: The solid trace for the night of 25–26 September and a dashed trace corresponding to a quiet night on 19–20 September 2001. Since GPS satellites have an orbital period of one-half sidereal day, the daily contribution to S4 from contaminating multipath repeats itself daily with a slow, roughly 4 min, drift from solar day to solar day. Hence the dashed trace represents a noise floor level above which the S4 index indicates the presence of irregularities and a disturbed ionosphere. Clearly, a disturbed ionosphere existed during the evening of 25–26 September 2001.
 TEC measurements provide path integrated electron density which, viewed over a period of time, can be an indicator of the presence of irregularity structuring associated with scintillations [Beach, 1998]. Bhattacharyya et al.  have shown that a time-varying TEC profile is indicative of the structures that cause L band scintillations in equatorial regions. Figure 2 shows TEC measured from Ithaca for four different satellites over several hours for the evening of interest and a previous quiet evening. In this figure, both examples of structuring and sharp density gradients are apparent. The specific details are difficult to understand without considering the spacecraft puncture points and velocities, which is done in the work of Ledvina et al. . For our discussion, it is sufficient to note the satellite psudo-random numbers (PRNs) 8 and 27 were moving from north to south with PRN 8 leading PRN 26 by about one hour while PRNs 3 and 31 were moving from south to north with PRN 3 leading PRN 31 by about one hour. The equatorward edge of a trough in the storm-enhanced density (SED) then appears at about 2700 UTC for PRN 8, at about 2600 UTC for PRN 27, and about 2400 UTC for PRN 31. For PRN 3, the trough boundary is apparently encountered over an extended region from 2330 UTC to 2500 UTC. We are using the word “trough” in a generic sense here. Usually the expression “ionospheric trough” refers to a density depletion formed from enhanced F region recombination rates in narrow regions of rapid (1–4 km/s) plasma flow [Schunk et al., 1976]. However, in our case, the situation is not so clear. Overall, the values of TEC were elevated over the entire night with depletions embedded in the elevated TEC that were still larger than the nominal TEC values. This may have been produced by the convection of lower density regions or by the remnants of previous rapid ionospheric flows and enhanced recombination.
 Next, we examine the width of normalized autocorrelation functions at a value of one half (3 dB point). Similar studies have been performed at the equatorial anomaly [Kintner et al., 2001] to demonstrate the relationship between the autocorrelation width and the drift velocity measured between multiple spaced receivers. These results demonstrate typical widths of about 2 s to more than 10 s when resonance occurred between the GPS signal ionospheric puncture point and the ionospheric motion. Since the nighttime equatorial ionosphere moves at roughly 100 m/s while convective drifts driven by the inner magnetosphere are several 100 m/s or more, we expect the data presented here to exhibit shorter timescales. Figure 3 shows the normalized autocorrelation widths, called τ3dB and measured at half height, where the autocorrelation function is determined from 40 s blocks of 0.02 s samples and where the S4 index exceeded 0.15. The width of the normalized autocorrelation function is determined at an amplitude of 0.5. All four GPS satellites are shown.
Figure 3 demonstrates several important features. First, the autocorrelation widths are mostly less than 2 s as opposed the equatorial case, where 2 s is a typical value. The minimum autocorrelation widths are about 0.2 s, which is smaller than the equatorial case. The simultaneous minima values of PRNs 3, 27, and 31 at about 2430 UTC is happenstance. Each of these spacecraft were located in a region of depleted SED or trough with low TEC values which are expected to accompany the fastest flows. Finally, there is large variation in the autocorrelation width from 0.2 s to at least 2 s. This too is expected in the storm time convective electric field where large-velocity shears have been previously measured [Reddy and Mayr, 1998].
 Next, we estimate the ionospheric velocity from one GPS signal autocorrelation width. This is a somewhat speculative process since there are several unknowns. The formal process we use for the estimate is to calculate the Fresnel radius (), where z is the distance to the scattering source, to multiply this by , corresponding to the length between diffractive peaks or nulls, and to divide by 3τ3dB. This yields the equation
where λ is the L1 wavelength. The Fresnel radius contains the unknown of z which we assume to be 350 km modified by the elevation angle, although in the subauroral trough irregularities may exist at higher altitudes where flow-enhanced recombination is less effective. Hence the calculated Fresnel radius may be an underestimate. Next, the number 3 is used to scale the width (τ3dB) of the autocorrelation function to the distance between peaks in the spatial scintillation pattern and is calculated using a sinusoidal amplitude scintillation pattern. Direct estimates of this scaling factor have been made at the equatorial anomalies using spaced receivers as a reference, and values from 3.2 to 4.6 were determined [Kintner et al., 2001]. The two underestimates, one in the numerator and one in the denominator, tend to cancel each other, so we believe that the values are reasonably accurate. Finally, one should note that we have not removed the effect of the moving GPS signal ionospheric puncture point because it is a vector correction and we can only calculate a scalar speed. Formally, the estimated speeds correspond to the scintillation pattern ground speed which should be corrected for signal puncture point speeds to obtain ionospheric drift speeds. Typical puncture point speeds for high-elevation GPS satellites are 30–60 m/s.
Figure 4 shows the velocity calculated using the formal process described above for just two GPS signals, PRN 27 and PRN 31. The velocity estimate vary from roughly 50 m/s to nearly 700 m/s and are highly variable. Most values are between 100 and 400 m/s, with both signals exhibiting a narrow maximum between 400 and 700 m/s. These estimated velocities are larger than those found at equatorial latitudes (100–150 m/s) and are typical of midlatitude convection during geomagnetic storms. The two peaks in speed at 2430 UTC for PRN 27 and PRN 3 are suggestive of SAID, SAPS, or simply the ionospheric trough because they have a narrow latitudinal extent, <0.5°, although the magnitude of the speed is somewhat smaller than typical cited values of 1–4 km/s [e.g., Bourdillon et al., 1982; Providakes et al., 1989]. The PRN 27 and 31 ionospheric puncture points were separated by 200–250 km roughly east-west at 2430 UTC.
 The production of L band scintillations requires the simultaneous occurrence of large ionospheric densities and irregularities at the Fresnel length, roughly 350–400 m. This is one reason for GPS scintillations being common at the equatorial anomalies where large plasma densities are driven by dirunal tides and irregularities can be produced by spread F drifts. These features do not occur commonly at midlatitudes. The event described herein occurred in the aftermath of a ionospheric storm similar to the one described in the work of Foster et al.  and by A. J. Coster (personal communication, 2002). The main plume, or SED, had formed over the eastern seaboard of the United States and moved westward when the scintillations and ionospheric gradients were observed. This scenario was further supported by TEC measurements. Figure 2 provides evidence of a trough-like structure aligned longitudinally within a region of S4 enhanced electron density, particularly at the equatorward edge, which is consistent with satellite observations of the ionospheric trough at midlatitudes [Evans et al., 1983]. Recently, Vo and Foster , using radar data, provided similar results for the northeastern United States that show the ionospheric trough positioned over the region of interest with adjacent SED at the equatorward edge.
 From the first publication of this data [Ledvina et al., 2002] it is obvious that the strongest scintillations coincide with steep TEC gradients and that the strongest gradients occurred during the main phase of moderate geomagnetic storm when Dst reached a peak negative value of approximately −100 nT. Basu et al.  recently published data taken from Hanscom Air Force Base in the northeastern United States showing VHF and UHF scintillations occurring during two different geomagnetic storms. Basu et al.  show that in each case the onset of the scintillations coincides with the main phase of the storm, suggesting a possible relationship between the two. The relation ship may be the creation of SED accompanied by an ionospheric trough during geomagnetic storms.
 During March 2001, J. J. Sojka (personal communication, 2002) demonstrated an occurrence of modest (S4 ≤ 0.2) GPS L1 scintillations correlated with the development of a magnetic storm occurring in the western United States during a geomagnetic storm. In this example the S4 index responds in parallel with the local horizontal magnetic field, which in turn is presumably produced by the development of a ring current. This suggests that penetrating electric fields, associated with the ring current development, are also associated with the development of irregularities.
 The contribution of this paper is to show that L band scintillations are created in regions of large ionospheric velocities during the main phase of a geomagnetic storm. These large velocities are suggestive of SAID and the subauroral ion trough. The large velocities may be related to the formation of irregularities at the Fresnel radius that are responsible for L band scintillations. Turbulent flows in the presence of large-velocity shears is a well-known phenomena in virtually all fluids including MHD fluids. Investigating the relation between velocity shear, density gradients, and irregularity formation in the subauroral ionospheric trough should be a feature of further research in this area.
4. Concluding Remarks
 We have presented further evidence associated with observations of GPS L1 scintillations at midlatitudes. The L band scintillations occurs during the negative phase of a geomagnetic storm, while the ring current is growing. Measurements of velocity, computed from the 3 dB width of the signal power autocorrelation function and the Fresnel radius, compare favorably with previous studies of subauroral velocities in the midlatitude ionosphere during geomagnetic storms. Velocities of several hundred meters per second were typical and were observed on all four GPS signals investigated.
 The formation of irregularities producing the L band scintillations may be related to the large velocities and accompanying velocity shear. However, scintillations also depend on absolute ionospheric density, and a process that both increases ionospheric density, compared to typical values, and produces irregularities in the density is required.
 Scintillations of this magnitude and the associated electron density fluctuations can cause considerable problems for commercial GPS receivers. The scintillations can cause the signal power to dip below a detectable level, and the phase fluctuations caused by the time-varying TEC can easily extend outside the bandwidth of tracking loops.
 This research was supported by ONR grant N00014-92-J-1822.