The spread F Experiment ( SpreadFEx ) : Program overview and first results

1NorthWest Research Associates, Colorado Research Associates Division, Boulder, CO 2Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos, Brazil 3Universidade Federal de Campina Grande, Campina Grande, Paraı́ba, Brazil 4Purdue University, West Lafayette, IN 5Utah State University, Logan, UT 6University of Illinois, Urbana, IL 7Universidade Estadual da Paraı́ba, Campina Grande-PB, Brazil 8National Center for Atmospheric Research, Boulder, CO 9IP&D—Universidade do Vale do Paraı́ba—UNIVAP, São José dos Campos—SP, Brazil


Introduction
Considerable progress has been made in recent years in understanding the occurrence, morphology, and variability of equatorial spread F (ESF) and plasma bubbles penetrating to higher altitudes. Despite this, an understanding of the conditions seeding ESF and plasma bubbles has remained elusive, with neither observations nor theory sufficiently comprehensive or persuasive to be definitive. Computed Rayleigh-Taylor Instability (RTI) growth rates based on known flow structures are typically too small, and it is unclear whether seeding can be entirely a plasma process or whether neutral perturbations are necessary to enhance plasma instability growth rates. Gravity waves (GWs) have been suggested as a trigger in many studies, either directly via density and/or velocity perturbations at the bottomside F layer or indirectly via mapping of electric field or other field-line-integrated perturbations generated by GWs interacting with sporadic E (E s ) layers or other fields at lower altitudes off the dip (magnetic) equator. However, definitive proof of GW seeding, the "smoking gun", observational or theoretical, has yet to be found. These issues motivated a combined experimental, modeling, and theoretical program initiated with two field campaigns performed in Brazil during "moon down" conditions from September to November 2005. Our measurements employed a suite of airglow cameras, VHF and meteor radars, digisondes, and GPS receivers at a number of fixed and temporary sites and correlative measurements with the Jicamarca Radio Observatory in Peru and via satellite (GUVI aboard TIMED and CHAMP) to characterize the neutral atmosphere and ionosphere as fully as possible during these periods. The primary goal of the spread F Experiment (SpreadFEx) was to test the theory that GWs play an essential role in the seeding of ESF and plasma bubbles extending to much higher altitudes. The purpose of this paper is to summarize our experimental design and describe our initial analysis results.
More detailed analyses of SpreadFEx data, supporting theoretical studies, and assessments of specific potential GW influences at bottomside F layer altitudes will appear in a forthcoming SpreadFEx special issue of Annales Geophysicae.
We begin in Section 2 by reviewing what is known of ESF and plasma bubble dynamics based on ∼30 years of observations, modeling, and theory, and what we understand of GW excitation and propagation into the thermosphere/ionosphere (TI). Our SpreadFEx campaign objectives and measurement strategy employed for our two field campaigns and an overview of our initial analyses are described in Sections 3 and 4. Section 5 provides a discussion of potential direct GW contributions to ESF and plasma bubble seeding at the bottomside F layer, as these are the only contributions we can assess within the scope of this overview. A summary of our initial results is presented in Section 6. The majority of the data discussed here are from 24 to 27 October 2005 during our second SpreadFEx campaign interval. However, some of these data (particularly airglow and some GPS measurements) were collected under clear skies during our first campaign interval.

ESF and plasma bubbles
ESF on the bottomside of the F layer yields ionosonde signatures virtually every night of its occurrence season. Strong ESF and plasma bubbles typically arise during the pre-reversal enhancement (PRE) of the zonal electric field when upward E × B plasma drifts elevate the F layer sufficiently (Heelis et al., 1974;Fejer et al., 1999). RTI is believed to be responsible for these plasma instabilities, and nonlinear RTI growth causes flux tube-aligned plasma de-pletions to rise to the topside ionosphere, where they may attain equatorial apex heights of ∼1000-1500 km in well developed cases. Plasma bubbles that penetrate to high altitudes occur on only ∼30 to 60% of nights, are apparently uncorrelated at closely spaced longitudes, but may share seeding conditions, and exhibit seasonal, solar-cycle, geographic, and geomagnetic dependencies (McClure et al., 1977;Sobral et al., 1980aSobral et al., , b, 2002Mendillo and Tyler, 1983;Abdu et al., 1992;Aggson et al., 1992;McClure et al., 1998;Fejer et al., 1999;Sobral et al., 2001;Hysell and Burcham, 2002;Rodrigues et al., 2004;Stolle et al., 2006;Su et al., 2008). In particular, increasing solar flux correlates with greater PREs of plasma drift, earlier ESF seeding and irregularity appearance, and higher initial altitudes (Hysell and Burcham, 2002).
GW perturbations have been suggested to lead to local current and electric field fluctuations having the wavelength of the GW (Klostermeyer, 1978;Kelley, 1989). More recently, GW interactions with sporadic E (E s ) layers were suggested by Prakash (1999) to yield electric field perturbations that map efficiently to the bottomside F layer; a related mechanism involves the mapping of large-scale polarization electric fields arising from E s layer instability to these same altitudes (Tsunoda, 2005(Tsunoda, , 2006(Tsunoda, , 2007. These fluctuations are suggested to seed RTI extending to greater altitudes. Major challenges to these proposed mechanisms, however, are apparent random associations of E s layers and ESF . Observed plasma bubble scales and modeled initial conditions vary considerably, but are typically in the range of 10's to 100's of km in the plane normal to the magnetic field (Ossakow, 1981;Tsunoda, 1981;Haerendel et al., 1992;Sultan, 1996), seeding altitudes are ∼200 to 300 km at the dip equator, but may be 100 km or more lower away from the dip equator, and perturbations must apparently be sufficiently aligned along magnetic field lines so as to yield a field-line-integrated perturbation of sufficient magnitude (Huang and Kelley, 1996c). Indeed, GW density perturbations of a few % or vertical velocity perturbations of a few m s −1 are critical seed elements in models that have sought to describe RTI and plasma bubble growth and morphology to date (Huang and Kelley, 1996a, b). We also note that the anticipated arrival time at the bottomside of the F layer of large-amplitude GWs arising from tropical convection is ∼1 to 2 hours after maximum convective activity (Vadas and Fritts, 2004) and coincides closely with early evening times of strong ESF and bubble onset (Swartz and Woodman, 1998).
Because of the field-aligned nature of plasma fluctuations, much of the modeling of ESF and plasma bubbles has been two-dimensional (2D). These efforts captured some of the gross features of the morphology and examined sensitivity to seeding conditions and scales (Scannapieco and Ossakow, 1976;Keskinen et al., 1980;Zalesak and Ossakow, 1980;Hysell et al., 1990;Huang et al., 1993;Sekar et al., 1995). Other efforts accounted approximately for variations of key parameters along the magnetic field (and with altitude) and for key observed ESF and bubble features via field-line-integrated methods (Zalesak et al., 1982;Keskinen et al., 1998). Fully three-dimensional (3D) linear and nonlinear studies have been performed more recently that have delineated the impacts on growth rates and structure and yielded reasonable agreement with observations (Basu, 2002;Keskinen et al., 2003). The study by Keskinen et al. (2003) also reproduced bubble structures with sharp density gradients extending up to the equatorial anomaly.
Studies specific to GW seeding of strong ESF and plasma bubbles have examined the necessary conditions for such seeding and the structure of resulting plasma bubbles, including dependence on GW amplitudes and zonal wavelengths, the implications of initial amplitudes for RTI enhancements, the need (or lack of) for spatial resonance, the influences of zonal shear on plasma bubble scales and depths, and the impact of Pedersen conductivity on GWinduced perturbations (Hysell et al., 1990;Huang et al., 1993;Sekar et al., 1995;Huang and Kelley, 1996a, b, c;Sekar and Kelley, 1998). These studies have addressed the physical mechanisms enabling GW seeding of ESF and plasma bubbles. Another statistical study of potential GW seeding by McClure et al. (1998) demonstrated potentially significant correlations between regions of enhanced (depleted) F region irregularity occurrence and regions of anticipated strong (weak) tropical convection. They even noted that an anomalous enhancement accompanied enhanced convection and sea surface temperatures during a major El Nino. Thus, statistical correlations suggest the potential importance of the mechanisms modeled by various groups and point to deep tropical convection as a potentially important source of such GW perturbations.

ESF and RTI growth rates
Numerous theoretical efforts have addressed the neutral and plasma conditions contributing to ESF and RTI growth rates. While various theories have provided insights into the likely environments favoring ESF, RTI, and plasma bubble formation, they have also frequently failed to yield sufficiently large growth rates for plausible large-scale neutral and plasma flows. This has been a large part of the motivation for invoking GW perturbations, as one component of the geophysical noise spectrum, in the seeding of RTI and plasma bubbles. Most previous efforts have addressed RTI growth rates, which depend largely on the plasma environment, but with potential GW influences exerted either directly at the bottomside F layer or at lower altitudes that then map to the bottomside F layer (Ossakow, 1981;Zalesak et al., 1982;Sultan, 1996;Prakash, 1999;Abdu, 2001;Tsunoda, 2005Tsunoda, , 2006Tsunoda, , 2007. However, a recent series of studies has considered an alternative instability mechanism depending directly on the neutral zonal wind that appears to achieve sufficiently large growth rates that explicit GW seeding of instabilities was suggested not to be required (Kudeki and Bhattacharyya, 1999;Hysell and Kudeki, 2004;Hysell et al., , 2005Kudeki et al., 2007).
A flux tube-integrated version of the generalized RTI linear growth rate based on Sultan (1996) can be written as where E is the zonal electric field (which arises from the evening PRE in the zonal electric field due to F layer dynamo), g is the acceleration due to gravity, E,F P are the field line integrated conductivities for the E-and F-region segments of a field line, U P FT is the conductivity-weighted, flux tube-integrated vertical wind (normal to B), L FT is the length scale of the density gradient, β is the recombination loss rate, ν eff is the effective ion-neutral collision frequency, and subscripts FT stand for flux tube-integrated quantities. In Eq. (1), GW influences can only be expressed through their impacts on the various plasma quantities affecting the RTI growth rate.
An alternative theory for preconditioning the bottomside F layer for ESF initiation and RTI claimed by the authors to not require GW seeding (Kudeki et al., 2007) yields a growth rate where U n and U p are the zonal (eastward) neutral and plasma velocities and L is the length scale of the density gradient. These conditions differ in important respects from those considered previously to be key to RTI growth itself.
The dominant requirements appear to be 1) a significant eastward neutral wind at the bottomside F layer, 2) a westward plasma drift at these altitudes, such that the eastward wind in the plasma coordinate is U ∼ 200 m s −1 , and 3) a vertical plasma gradient having an effective scale length of L ∼ 20 km or less. Kudeki et al. (2007) have argued that such conditions favor an instability having a westward phase tilt with altitude of 45 • induced via vertical Pederson currents. The obvious question, to be addressed further in Section 5, is whether GWs can contribute meaningfully to, or modulate, this latter mechanism, either through initial conditions or altered growth rates or both. We focus on the latter mechanism here because we can easily estimate direct GW influences at the bottomside F layer, but are unable to easily assess the potentially competitive mechanisms involving the creation of electric fields at E s layers at lower altitudes and their mapping to the bottomside F layer in this initial SpreadFEx overview. More detailed assessments of GW influences on plasma dynamics and instabilities will be provided in our SpreadFEx special issue (see, in particular, Fritts et al., 2008a, b;Abdu et al., 2008;Kherani et al., 2008).

GWs in the thermosphere/ionosphere
GWs are now understood to play a number of key roles in neutral atmosphere dynamics extending from the Earth's surface into the lower thermosphere (LT). GW effects in the mesosphere and lower thermosphere (MLT) are well documented and arise due to the attainment of large amplitudes, transports of energy and momentum from sources at lower altitudes, and turbulence, mixing, and flux divergences accompanying instability processes. The roles of GWs in ionospheric and plasma processes, in contrast, are less well known but also potentially important. As in the neutral atmosphere, GW importance in the ionosphere hinges on their penetration to high altitudes and their attainment of large amplitudes. GWs have been suggested to play a role in the seeding of RTI, strong ESF, and plasma bubbles penetrating to high altitudes, though their roles continue to be debated and definitive observational evidence has yet to emerge.
Of the various sources of GWs in the MLT, deep tropical convection is arguably one of the most significant, correlating with enhanced variances in the stratosphere (Tsuda et al., 2000) and with GWs having large amplitudes and penetrating to and well above the mesopause (Taylor and Hapgood, 1988;Piani et al., 2000;Lane et al., 2001;Sentman et al., 2003;Vadas and Fritts, 2004, 2006Vadas, 2007). Models of GW structure at OH airglow altitudes (∼87 km) that arise from convection are able to describe the qualitative character and some of the more quantitative aspects of the observed GW field quite well. Importantly, tropical sources apparently remain competitive with the auroral zones even at very high altitudes (Hocke and Schlegel, 1996). Convection does not obviously explain all of the GWs apparently propagating to high altitudes in the TI, however. Evidence of such motions in ISR data from Arecibo Observatory and the MU radar suggest penetration to much higher altitudes than implied for the spatial scales typically arising from convection (Oliver et al., 1997;Djuth et al., 1997Djuth et al., , 2004. This, and the inability to excite and propagate GWs having very large spatial scales from the lower atmosphere, suggest an additional source in the MLT, likely secondary GW generation due to GW instability and mean-flow interactions (Vadas andFritts, 2001, 2002;Vadas et al., 2003;Vadas, 2007;Fritts and Vadas, 2008).
GWs arising from individual convective plumes having frequencies of ω ∼ N /3, where N is the buoyancy frequency, can penetrate to lower thermospheric altitudes of ∼150 to 200 km altitudes under solar minimum conditions (Vadas and Fritts, 2004, 2006. Under solar maximum conditions, during which thermospheric densities increase and kinematic viscosities decrease dramatically, the same sources yields GWs that can penetrate to substantially higher altitudes (Vadas and Fritts, 2006;Vadas, 2007). Larger-scale GWs, likely not arising from individual convective plumes but rather as secondary radiation at higher altitudes, can penetrate to substantially higher altitudes under all solar conditions (Vadas, 2007). Together, these GW sources yield perturbations that span the spatial scales observed in plasma bubbles, horizontal scales of ∼40 to 400 km, and may reach substantial amplitudes at higher altitudes . Despite their small amplitudes at lower altitudes, the ∼5 to 15 scale heights between GW source regions and the bottomside F layer, and an expected suppression of GW instability processes that constrain amplitudes at lower altitudes (by the exponentially increasing kinematic viscosity), appear to allow GW perturbations to be large at the bottomside F layer. Specifically, horizontal and vertical winds, vertical displacements, and perturbation electron densities and electric fields due to GWs will likely be substantial at these altitudes (see below). While our SpreadFEx results to date are preliminary, it appears difficult at this stage to argue that GW perturbations cannot make plausible contributions to ESF and plasma bubble seeding conditions. Indeed, we believe that our initial data analyses, summarized below, suggest a significant role for neutral atmospheric waves in general and GWs in particular. These results will be elaborated further by Abdu et al. (2008), Fritts et al. (2008b), and  in the SpreadFEx special issue of Annales Geophysicae.

Overview of the Spread F Experiment (Spread-FEx)
The primary goal of SpreadFEx was to perform observational and modeling studies that would quantify the potential roles of GWs in seeding ESF and RTI in the bottomside F layer leading to plasma bubbles penetrating to higher altitudes. Because of the statistical links of plasma bubbles to tropical convection, we designed an experiment that would provide sensitivity to both neutral atmosphere responses to deep convection and plasma instabilities and structures at higher altitudes over Brazil. The specific link envisioned was via GW coupling from deep convection, with both GW perturbations and mean responses contributing to potential seeding of ESF, RTI, and plasma bubbles at higher altitudes, as depicted in Fig. 1. Specific questions motivating these efforts include 1) Do GWs play a significant role in the generation of ESF and plasma bubbles at greater altitudes? 2) How do GW perturbations to the bottomside F layer alter the seeding conditions conducive to RTI and ESF generation?

3) If GWs are an important component of ESF and bubble
generation, what are the geophysical parameters controlling their influences? 4) Are the GW roles in ESF and bubble generation sufficiently correlated with other measured geophysical parameters to allow a parameterization of these roles?
To address these issues, we designed a research program comprising three components. The first component was experimental, and included two measurement campaigns during "moon down" conditions from 25 September to 10 October and from 23 October to 8 November 2005. These measurement campaigns employed a suite of airglow cameras, VHF and meteor radars, digisondes, and GPS receivers at a number of fixed and temporary sites, as well as correlative measurements via satellite (GOES 12 and GUVI aboard TIMED) to characterize the neutral atmosphere and ionosphere as fully as possible during these periods. The second component of SpreadFEx was a series of analysis efforts. A third component included modeling of GW excitation and propagation in response to deep convection and plasma simulations of GW seeding to aid in the interpretation of our field data.
Our field measurements were performed in central and eastern Brazil and employed ground-based optical and radar instrumentation in collaboration with colleagues at Instituto Nacional de Pesquisas Espaciais (INPE) in Brazil, Utah State University (USU), and Purdue University (PU). INPE optical and radar instrumentation was located at several fixed sites, including São Luis, Fortaleza, Cariri, and Cachoeira Paulista extending from ∼2 to 23 • S geographic latitude and from ∼0 to 17 • S magnetic latitude, with GPS instrumentation more widely distributed. Additional tempo-   (de Paula and Hysell, 2004), meteor radars defined MLT winds at Cariri and Cachoeira Paulista Buriti et al., 2007), digisondes defined electron densities at several dip latitudes (Batista and Abdu, 2004), airglow cameras defined both GW structures at MLT altitudes and plasma bubble structures at higher altitudes in the thermosphere at several locations south of the magnetic equator in order to assess the spatial and temporal variability of these processes (Medeiros et al., 2004), and GPS sensors were employed to attempt to define the spatial and temporal variations in electron densities (Lanyi and Roth, 1988). The instrument locations and their relation to the magnetic equator are shown in Fig. 2 and listed in Table 1.

Initial Results from SpreadFEx
4.1 Convection, GW sources and propagation 4.1.1 GOES 12 indications of deep convection GOES 12 visible, IR, and water vapor measurements over Brazil (at 1-, 4-, and 8-km resolutions, respectively) provided the best way to quantify the locations, scales, and intensities of deep convection leading to GW generation and propagation to higher altitudes during SpreadFEx. Cold cloud-top temperatures indicate deep convection, and temporal variations between successive images provide an indication of updraft intensities and durations. These data are being employed to estimate the spatial and temporal scales of convective plumes that are assumed to launch GWs propagating into the MLT and to higher altitudes. An example of the GOES 12 IR data showing deep convection, color-coded to highlight the coldest convective plumes, is shown for reference in Fig. 3. This image shows a number of regions of active, deep convection. Five exhibit extensive cold temperatures, assumed to be the largest and most mature convection, and the major sources of GWs at this time. These occur near (2 • S, 68.5 • W), (2.5 • S, 67 • W), (6 • S, 62 • W), (8.5 • S, 62 • W), and (10 • S, 50.5 • W). Brasilia is at (15 • S, 47 • W), our temporary optical site was ∼100 km north, and the dip equator is shown for reference. There are also a series of smaller cells having cold cloud tops, but smaller spatial scales, a number of which would also be actively exciting additional GWs. The largest and deepest convection is the most efficient source of GWs that penetrate to the highest altitudes, but the horizontal extent of the cloud top IR signatures often overestimates convective plume widths due to their generation of cirrus cloud shields as they mature. Indeed, recent high-resolution numerical studies by Lane et al. (2003) and Lane and Sharman (2006) show that such large-scale convection is often modulated by smallerscale ∼5-km plumes.
4.1.2 Ray tracing of GWs into the TI Based on initial assessments of plume scales in the GOES 12 imagery, the five dominant convective plumes seen in Fig. 3 were assumed to have full-width, half maximum (FWHM) spatial scales of (10, 10, 5) km for the 1st, 2nd, and 4th plumes,   Fig. 3. The profiles were defined by balloon soundings below 30 km, meteor radar observations between 80 and 100 km, and a TIME GCM simulation initialized with the NCEP reanalysis data for this same period from 35 to 70 km and from 110 to 350 km, with linear interpolation elsewhere. and (5, 5, 5) km for the 3rd and 5th plumes listed above. Assumed updraft velocities for the five were 30, 30, 35, 30, and 15 m s −1 , respectively. GWs assumed to arise from these plumes were ray traced from an assumed source altitude of 14 km upward through the zonal and meridional wind fields and a temperature field defined in the following manner. Balloon data were used to specify these fields from the surface to 30 km, meteor radar data from Cariri were used to define the low-frequency wind field between 80 and 110 km, and data from a thermosphere-ionospheremesosphere-electrodynamics (TIME) general circulation model (GCM) simulation, initialized with National Center for Environmental Prediction (NCEP) reanalysis data for this interval, were used to define the large-scale wind and temperature fields between 35 and 80 km, and again between 110 and 350 km. The resulting wind and temperature profiles used for this purpose are shown at (45 • W, 10 • S) in Fig. 4. There is, however, considerable variability in the wind field, and the GWs that survive to the highest altitudes are thus very sensitive to their source scales and locations. The resulting GWs were ray traced using the methodology of Vadas and Fritts (2004) and incorporated viscous dissipation at higher altitudes to assess maximum penetration altitudes (Vadas and Fritts, 2005).
Of the spectrum of GWs arising from each of the dominant convective sources, only a small fraction having suitable propagation directions and relatively large horizontal scales succeeded in surviving to 200 km altitude. The momentum flux distributions accompanying the surviving GWs for all five sources are contoured in Fig. 3 at 21:55 and 22:15 UT, the first time approximately an hour after GW generation. Note that the earlier responses (blue contours) are due to GWs having smaller horizontal wavelengths (∼100±50 km) and higher frequencies, hence higher vertical group velocities, with the response at later times (red contours) having slightly larger horizontal wave- October 2005. Note that movement of structures in both emissions is generally eastward, but is faster and at ∼100 to 500-km zonal wavelengths in the TI and slower and at ∼20 to 400-km zonal wavelengths in the MLT lengths (∼115±70 km) and occurring at larger distances from the convective sources. The earlier response is composed largely of westward-propagating GWs having negative zonal momentum fluxes (dashed contours), while the later response has preferentially southeastward propagation and positive zonal momentum fluxes (solid contours). In all cases, the dominant responses at the higher altitudes are largely dictated by turning levels due to strong winds opposite to GW propagation in the LT. At all times, however, GWs having high phase speeds are found to propagate in a range of directions, depending on the local winds. Many more GWs succeeded in penetrating from convective sources into the MLT, and many of these were easily seen in the various airglow emissions measured at our three SpreadFEx optical measurement sites. These sites were typically east of the major convection, so the predominant eastward propagation appears to be consistent with primarily convective sources. Examples of GW propagation in the east-west plane are shown with Keograms of OH airglow brightness from Brasilia and Cariri in the lower panels of Fig. 5. Horizontal wavelengths were typically in the range ∼20 to 400 km, with the largest brightness variations occurring at intermediate scales of ∼50 to 200 km. While not displayed here, ray tracing of GWs having these horizontal scales, but lower frequencies than those penetrating into the TI, suggest largely eastward propagation and typical propagation times of several hours from their convective sources to the MLT. These results are generally consistent with the scales and largely eastward propagation seen in Fig. 5. Based on these observations and the accompanying ray tracing, we must be careful in attempting to link GWs in the MLT with plasma instabilities and bubbles at higher altitudes, as what dominates the airglow brightness modulations may not fully characterize the GWs that survive to 200 km and above. More quantitative and complete assessments of GW sources, generation, and spatial scales attaining MLT and higher altitudes are provided by Taylor et al. (2008), Takahashi et al. (2008), , , Fritts and Vadas (2008), and Fritts et al. (2008b) in our SpreadFEx special issue.

Bores in the MLT
While not a central focus of the SpreadFEx measurement campaigns, our airglow, radar, and related satellite measurements also provided sensitivity to mesospheric bores occurring at MLT altitudes and to the wind and temperature fields in which they occurred. As bores are believed to be excited by GW instabilities, dissipation, and/or mean-flow interactions, their study represents an interesting extension of our campaign objectives and specific GW studies, and it benefits significantly from our comprehensive SpreadFEx data set. One example of a bore seen at Cariri on 1 October is shown in Fig. 6 at two times separated by 20 min. This bore structure was most visible in the OI 5577Å emission, propagated towards the NE, and had an initial wavelength that appeared to shorten and a number of crests that appeared to increase over the 20-min interval displayed. Analysis of this event suggests a complex ducting structure  and helped motivate our initial modeling studies of bore excitation and evolution in thermal and Doppler ducts.
Bore structures, evolutions, and excitation have been studied previously by Picard (1998, 2001), Seyler (2005), Fechine et al. (2005), and Medeiros et al. (2005), largely in highly idealized flows. Observed bores, however, occur in more complex environments in which both temperature and wind fields play a role in defining the ducting environment. In general, the separate contributions of temperature and winds (yielding Doppler ducts) do not occur at the same altitudes and are expected to impact bore (and linear ducted responses) differently, depending on propagation direction. To illustrate some of the variability   For the broad thermal duct (middle panels of Fig. 7), there are two distinct ducted features, one occurring somewhat below the thermal duct propagating to the left and a second somewhat above the thermal duct propagating to the right. Each feature exhibits a largely anti-symmetric response in the temperature (and vertical velocity) field because of the symmetric horizontal forcing and normal dispersion, with the larger wavelengths having larger phase and group velocities than the smaller wavelengths. The larger and smaller wavelengths within each response are also centered at somewhat different altitudes. This is because the mean velocity varies across the thermal duct, and smaller-scale motions having smaller phase speeds are more influenced by mean winds and tend to occur nearer the max-imum velocity in their direction of propagation. In fact, the larger wavelength ducts are defined more by the thermal structure, and the smaller wavelength ducts are defined more by the wind structure.
The lower panels in Fig. 7 shows the response when the thermal duct is both narrower (3 rather than 8 km) and stronger (a peak N 2 ∼ 40 times, rather than ∼28 times, the background). Both responses in this case are much more confined about the thermal duct, and there is now very different dispersion between the left-and rightpropagating modes. The left-propagating packet closely resembles that in the upper panels, except that it is shallower. The right-propagating response, however, is both shallower and exhibits apparently different dispersion for the larger and smaller wavelengths. The larger wavelengths propagate near the thermal duct as before, but the smaller wavelengths now appear to comprise a distinct wave packet having a much smaller group velocity and occurring near the peak of the Doppler duct rather than the thermal duct. Indeed, it appears possible that the two parts of this response are exhibiting different responses due to anomalous dispersion expected in a dual-ducting environment (Fritts and Yuan, 1989). These features will be explored in greater detail elsewhere in this, and the SpreadFEx, special issues Laughman et al., 2008).

Electron densities, TEC, and GWs in the F layer
We now turn to an overview of the observational evidence of possible neutral and plasma coupling at the bottomside F layer obtained during the SpreadFEx measurement campaigns. Digisonde measurements at São Luis for 24 to 27 October 2005 are shown for reference in the upper panel of Fig. 8. Expanded versions of these data (18 to 03 UT), and the corresponding data from Fortaleza, are shown for the first two of these nights (24/25 and 25/26 October) in the middle and lower panels, respectively. Electron density fluctuations computed from the digisonde data from São Luis for all three nights from 1800 to 1200 UT are shown in Fig. 9. These perturbations appear larger than often observed elsewhere, but are indicative of the potential for large plasma density perturbations in the F layer, though the specific roles of these perturbations in generating ESF and plasma bubbles are unclear at this time.
Each night exhibits the expected rise in the F layer accompanying the PRE and a series of oscillations thereafter having large amplitudes and apparent vertical displacements. Observed periods range from ∼20 min to 2 hr, with the largest electron density fluctuations (∼40%) occurring at the longer periods. Also seen are apparent downward phase progressions, especially at altitudes of ∼280 km and below. These are most apparent in the electron density fluctuations at the longer periods from ∼00 to 04 UT for each day shown in Fig. 9. There are also several regions in each panel where opposite phases of the perturbation overlap each other. This provides some indication of apparent GW vertical wavelengths. More quantitative estimates of GW amplitudes, scales, and likely propagation conditions inferred from these data are discussed in Section 5 below.
Prompt penetration electric fields have been suggested to lead to strong uplift of the F layer (Abdu et al., 2003;Huang et al., 2004) and are expected to occur on a global scale (Kelley, 1989). Substorms may likewise contribute significant short-term perturbations at equatorial latitudes. Such effects are expected to be more pronounced under enhanced high latitude convection conditions, so we display the K p index during September and October 2005 in Fig. 10. These data reveal significant K p (∼4 from 00 to 12 UT) the evening of 24/25 October, but with substantially smaller values of ∼2 the next two nights. The D st index, which is available at significantly higher temporal resolu- tion, was likewise relatively quiescent, with mean values for 24 to 27 October of +1, −20, −21, and −15, with peak negative values of −39, −33, and −23 the last three days. ACE/MAG magnetic fields during these days were similarly weak to moderate. Thus, apart from the evening of 24/25 October, with a mean K p ∼ 4 and a mean D st of −20 nT, there is little evidence that there were significant geomagnetic influences. And even on this night, D st values were substantially smaller than −100 nT, and are thus relatively geo-magnetically quiet. Evidence of GWs at even higher altitudes is obtained from GPS estimates of the temporal derivative of integrated or total electron content (IEC or TEC), assumed to be representative of changes at the altitudes of maximum electron density, or sub-ionospheric point (SIP), ∼300-400 km, by the use of various GPS receivers in Brazil. Examples of such information as derived from data obtained with the GPS receiver at Fortaleza on 24 October are shown in Fig. 11 which suggest typical GW periods at these altitudes of ∼15 to 60 min. The various satellites viewed by the receiver during these times indicate similar temporal variations at the same times, suggesting spatial coherence among the various "piercing points" of the ionosphere. For reference, the vertical dashed red line in Fig. 11 indicates the anticipated start time for plasma bubble initiation (Woodman and LaHoz, 1976), suggesting that the d(TEC)/dt fluctuations seen at earlier times must be due to GW influences in the ionosphere rather than plasma bubbles themselves. These data are challenging to interpret unambiguously, however, because it is difficult to separate temporal variations that may be due to GW propagation from possi- ble spatial variations accompanying slow satellite motion. Nevertheless, they demonstrate fairly persuasively that GW perturbations in electron density also extend to higher altitudes than implied unambiguously by the digisonde electron densities.
To increase the potential for the detection of GWs and small-scale plasma structures in the ionosphere during SpreadFEx, five temporary GPS stations with an approximate spacing of 50 km between sites were also installed near the imager site in central Brazil at Fazenda Isabel, São João de Aliança (FAZ1), Parque Nacional, Alto Paraiso (ALPA), Ibama, Alvorado do Norte (ALVO), Flores de Goias (FLOR), and Teresina de Goias (TERE). The locations of each of these systems, along with permanent sites at Brasilia (BRAZ), Montes Ciaros (MCLA), and Uberlandia (UBER) from the Rede Brasileira de Monitoramento Contínuo (RBMC) GPS network surrounding the temporary GPS stations, are listed in Table 2. For these systems, the integrated electron content (IEC) was calculated from dual-frequency carrier phase observations (Lanyi and Roth, 1988;Hofmann-Wellenhof, 1994;Calais and Minster, 1998), with a 4th-order polynomial removed from each IEC time series to remove the effects of the diurnal variation in electron content. Array processing techniques developed to detect propagating disturbances in dual-frequency GPS time series (Calais et al., 2003) are currently being used to  Fig. 12(a), (b), and (c). Depletions in the IEC are clearly seen as troughs in the time series from all satellites that had tracks to the west of the array (PRN28, PRN4, and PRN8). Figure 12 Assuming a disturbance propagating perpendicular to a line through ALPA and FAZ1 that reached the SIP from TERE at a later time, a propagation speed of ∼80 m s −1 in a direction roughly ENE was inferred. Similarly, for PRN8 the propagation direction was approximately eastward at ∼101 m s −1 , while traces from PRN4 showed eastward propagation at ∼119 m s −1 . Indeed, the inferred propagation directions and speeds, as well as their variations with longitude and decreases with time, agree with similar estimates using airglow imagers (see Fig. 5). Future analysis will attempt to estimate vertical motions by combining SpreadFEx GPS and airglow data.

Spread F and plasma bubble occurrence, structure, and correlations
The evolution of the bottomside F layer into plasma bubbles extending to higher altitudes is illustrated in Fig. 13 with VHF radar RTI plots at São Luis for the nights of 24/25 and 25/26 October 2005. These data correspond to the first two nights of digisonde data and electron density fluctuations shown in Figs. 8 and 9. The RTI plots indicate the presence of ESF at the bottomside F layer beginning as low as ∼230 to 250 km in most cases, with apparent backscatter also occurring on 24/25 October as low as ∼200 km beginning just before 00 UT. Plasma bubbles appear not to be initiated until the bottomside F layer is elevated to altitudes of ∼250 to 300 km. We must be careful in our interpretation of these data, however, as plumes that are first observed at higher altitudes were necessarily seeded at earlier times to the west of São Luis.
Closer examination of Fig. 13 (top panel) suggests that ESF and RTI lead to bubble generation accompanying the strong upward motions of the bottomside F layer at ∼22 UT and just before 00 UT on 24/25 October that are also seen in the digisonde data of electron density fluctuations in Figs. 8 and 9. The periodicity of the plasma bubble generation appears consistent with the lower-frequency modulations (∼2-hr periods) in these other data sets that were suggested to indicate possible GW influences below ∼280 km. It is more difficult to estimate the time of initiation of the second plasma bubble, however, because the bubble had already penetrated to high altitudes at the time of its detection over São Luis. It is also not possible to assess the spatial scales or temporal evolutions of these bubbles from the RTI plots alone. Indeed, these data highlight  a challenge in interpreting our ground-based plasma bubble measurements, as the RTI plots display a mixture of spatial structures and temporal variability. Shorter period fluctuations (∼30 min) also seen in the electron density data did not obviously make distinct contributions to plasma bubble seeding or spatial structure. To the extent that the ∼2-hr oscillations at lower altitudes are indicative of GW perturbations, these data suggest that neutral atmosphere GW motions may have played an important role in the seeding process. The same periodicities were also present at the bottomside F layer on 25/26 October (Figs. 8,9,and 13,bottom,center,and bottom panels), and the longer periods (∼2 hr) again appear to correlate well with the vertical plasma motion at these altitudes. In this case, shorter periods were more prevalent in the digisonde and electron density data and may have played a larger role in seeding the apparently more complex bubble structures at higher altitudes. Again, however, it appears difficult to establish a clear correlation between fluctuations at the bottomside F layer and the plasma bubbles observed at higher altitudes after the initial bubble occurrence because the later bubbles were already fully developed when observed over São Luis. The potential for GWs present at the bottomside F layer to contribute to plasma instability seeding and growth rates is explored in greater detail below. To resolve ambiguities in the interpretation of RTI plots from the VHF radar, additional data sets or analysis methods are required. Fortunately, the SpreadFEx campaign benefits from several. Airglow measurements at 6300Å such as displayed in Keograms in the upper panels of Fig. 5, provide information on spatial scales and drift speeds. These suggest typical plume separations of ∼100 to 500 km, at least at the altitudes that contribute most to plasma depletions, will play an important role in our more quantitative analyses to follow. Additional information on plasma bubble geometries and evolutions can be obtained by interferometric analysis of the VHF radar data (Hysell, 1996), but these results will likewise await our Spread-FEx special issue. Finally, data from the GUVI instrument aboard the TIMED satellite will be seen below to yield a dramatic enhancement of our ability to quantify plasma bubble structures via tomographic inversions of 1356Å emissions at the times of these overpasses.

Plasma bubble structures inferred from space
The GUVI instrument aboard the TIMED satellite provides another valuable data set for plasma bubble studies, and there were several overpasses that were coincident in space and time with some of our SpreadFEx measurements. The primary measure of bubble structures by GUVI is via the 1356Å emissions, which show significant line-of-sight depletions when viewing bubbles near nadir. An example of the bubbles seen by GUVI over S. America during our SpreadFEx measurements is shown in Fig. 14. These data afford another opportunity to correlate bubble structures with 6300Å airglow images and VHF radar RTI plots in defining bubble scales and morphology.
Even more exciting is the ability to perform tomographic inversions with the GUVI data because the satellite motion and GUVI viewing have components normal to the magnetic field over Brazil. This allows extraction of electron density measurements in a plane normal to the field lines having horizontal and vertical resolution of ∼40 and 20 km, respectively. This allows for unprecedented resolution of the spatial scales of individual bubbles and much more quantitative comparisons with our SpreadFEx correlative data than would otherwise be possible. Two examples of these data from 1 and 2 October and the corresponding 6300Å airglow images from Cariri viewing the same bubble structures are shown in Fig. 15. While we did not perform an extensive analysis of these correlative data prior to writing this paper, this is the focus of the effort by Kamalabadi et al. (2008) in the SpeadFEx special issue.

Potential Neutral Atmosphere Contributions to ESF, RTI, and Plasma Bubble Seeding
The RTI and ESF preconditioning growth rates expressed through Eqs. (1) and (2) in Section 2.2 above indicate the expected sensitivities of these plasma instabilities to neutral motions able to impact these quantities. Here we examine the potential direct contributions of such neutral atmo- sphere motions to plasma instabilities accompanying neutral wave perturbations occurring at F layer altitudes. An assessment of possible GW influences at lower altitudes that map electric field or other perturbations to higher altitudes is more complex and beyond the scope of this summary of first SpreadFEx results. For more complete discussions of related SpreadFEx analyses, we refer the interested reader to Fritts et al. (2008a, b), Abdu et al. (2008), Kherani et al. (2008), and other papers in the SpreadFEx special issue.

Tidal influences at the bottomside F layer
A major contributor to TI structure that appears to have obvious relevance to the preconditioning arguments of Kudeki et al. (2007) is the neutral atmosphere tidal structure that penetrates to high altitudes, and is generated in situ, at equatorial latitudes. The migrating tides are expected to comprise the dominant motion field at F layer altitudes, and their temporal behavior described by the TIME GCM appears to make a clear contribution to the potential plasma instability processes envisioned by these authors (we note, however, that the TIME GCM tidal structures at F layer altitudes have yet to be validated, so their detailed amplitude and phase structures should be considered to be only suggestive at this time). Zonal tidal winds on 25/26 October 2005 over the dip equator in Brazil at 21, 00, and 03 UT for a model run spanning the SpreadFEx campaign period are shown in Fig. 16. These profiles suggest that zonal winds are small or negative (westward) at ∼21 UT and swing sharply positive (eastward) thereafter at F layer altitudes, which appears to correlate well with onset times for plasma bubbles noted by previous authors. If the dependence of ESF initial instability is indeed a strong function of eastward winds, as argued by Kudeki et al. (2007), then it seems likely that tidal winds will play a very large role in determining when strong instability is possible, or at the least, in modulating instability growth rates. Because tidal structures exhibit significant amplitude variability at lower altitudes, due both to variable forcing in the lower atmosphere and modulation of processes coupling to higher altitudes, they are also expected to impose this variability on ESF, RTI, and plasma bubble occurrence statistics at F layer and higher altitudes. A similar suggestion was voiced by Meriwether et al. (2008) based on long-term Fabry Perot interferometer measurements in Peru. Indeed, solar-cycle variations in thermospheric temperatures will also modulate tidal amplitudes and phases and could contribute, together with the modulation of GW penetration altitudes in the TI, to the solar-cycle occurrence statistics of plasma bubbles.

GW perturbations at the bottomside F layer
GW motions were suggested above to contribute to bottomside F layer electron density fluctuations and to corresponding neutral atmosphere wind and temperature perturbations and displacements. Here we attempt to estimate the magnitudes of these fluctuations in order to assess their potential to contribute to the plasma instability processes discussed above. To assign the most conducive GW geometry, we consider the idealized plasma instability environment envisioned by  shown in Fig. 17. This schematic, and the arguments by Kudeki et al. (2007), suggest that instability will be favored for perturbations that enhance both electron density gradients and neutralminus-plasma zonal motions (U n − U p ) in regions where the plasma has been elevated. This is the lower left quadrant of the clockwise circular plasma motion depicted in Fig. 17, where the plasma motion opposes the neutral motion and induces rising motion westward and above, which increases the perturbation displacements. As we will see below, GWs can enhance eastward motions accompanying either upward or downward motions, but enhanced neutral and plasma density gradients always accompany downward neutral motions.
Considering first the neutral GW perturbations implied by a given fractional electron density gradient, we assume mean neutral and electron density profiles of the form and ρ e (z) = ρ e e z/H e , where H and H e are the respective (positive) neutral and electron density scale heights. We also assume GW velocity, pressure, potential temperature, and neutral density perturbations of the form (u , v , w , p / p, θ /θ, ρ Fig. 17. Schematic of plasma instability growth geometry and its dependence on neutral wind and the plasma gradient (after . where primes denote perturbations and k = (k, l, m) is the GW wavenumber vector. Assuming that the plasma moves with the GW perturbation wind field and that there are no chemical or electrodynamic processes that also impact plasma densities, the electron (or ion) continuity equation, dρ e /dt = 0, may be written as ρ e /ρ e = ku /ω + lv /ω + mw /ω − iw /ωH e . (6) Employing the energy conservation and continuity equations  iω θ /θ = N 2 /g w and Assuming that motions are approximately hydrostatic such that θ /θ ≈ −ρ /ρ and employing the relation Finally, the GW polarization relations  yield relations for vertical and horizontal perturbation velocities in terms of fractional densities or in a form that is more commonly used In the above, u h and w are the horizontal and vertical GW perturbation velocities, ω = k h (c − U n ) is the GW intrinsic frequency, k h = (k 2 + l 2 ) 1/2 = 2π/λ h and m = 2π/λ z are the GW horizontal and vertical wavenumbers, c and U n are the GW horizontal phase speed and the neutral mean wind in the direction of propagation (assumed zonal in the discussion of plasma instabilities above), λ h and λ z are the GW horizontal and vertical wavelengths, and β ∼ (1 − ω 2 /N 2 ) 1/2 ∼ 1 for hydrostatic and small-scale GWs. With our degree of approximation, Eq. (10) is accurate for all GWs in the TI, whereas β ∼ 1 in Eq. (11) only when λ 2 h 16π 2 H 2 and ω 2 N 2 . These are reasonable assumptions at lower altitudes, but they are less accurate where GW vertical wavelengths exceed ∼100 km and intrinsic frequencies exceed ω ∼ N /3, as we expect to occur in response to increasing kinematic viscosity and thermal diffusivity in the TI (Vadas andFritts, 2005, 2006;Vadas, 2007;Fritts and Vadas, 2008). When ω > N /1.4 and GW scales increase, β < 1/2 and approximations appropriate for the lower atmosphere lead to overestimates of u h based on density measurements. We again note that we have considered only advective effects in the above discussion, and have specifically neglected chemical, electrodynamic, and magnetic influences on plasma densities hat may be competitive. Thus, plasma perturbations will almost certainly be smaller than suggested by the above equations.
Under solar minimum conditions, thermospheric temperatures above ∼200 km are ∼600 K and gradients are small, such that N ∼ 10 −2 s −1 and the buoyancy period is T b ∼ 10 min. Thus we should expect typical GW intrinsic periods of T i = 2π/ω ∼ 30 min (Vadas, 2007;Fritts and Vadas, 2008), and the GWs suggested by the electron density and TEC fluctuations discussed above are in the expected range. We also expect neutral and electron density scale heights of H ∼ 15 to 20 km and H e ∼ 20 to 50 km, such that the ratio of fractional electron to neutral density fluctuations is likely something less than (ρ e /ρ e0 )/(ρ /ρ 0 ) ∼ 5 or smaller. The 10 to 40% electron density fluctuations inferred from the digisonde data then imply horizontal and vertical GW perturbation velocities of u ∼ 10 to 40 m s −1 and w ∼ 3 to 10 m s −1 (or larger, if electron density fluctuations are under-estimated) if motions occurred at smaller scales and nominal intrinsic frequencies of ω ∼ N /3, with estimates for u ∼ 2 times smaller for higher frequencies and larger scales. Thus, the velocity perturbations of GWs believed to have been present in the bottomside F layer during plasma bubble seeding during the SpreadFEx measurements appear to have been sufficiently large to impact plasma instability growth rates if they existed in the relevant regions and induced the appropriate perturbations to the differential plasma and neutral zonal velocities and/or the length scale (H e ) of the electron density gradient. As suggested above, GWs also contribute to perturbations of the mean electron density gradient and may thus contribute to modulation of plasma instability growth rates through both differential neutral and plasma motions and enhanced electron density gradients.

GW impacts on plasma growth rates
We now consider the GW orientations and structures implied by our SpreadFEx measurements and their possible contributions to plasma instabilities and growth rates, with a focus on direct influences at the bottomside F layer. Our SpreadFEx airglow observations and our limited ray tracing of GWs from convective plumes to date imply a preference for approximately zonal GW propagation in the MLT, though this may simply be a consequence of our SpreadFEx observations occurring generally to the east of the most intense convection. The TIME GCM simulation results employed here imply strong zonal and meridional tidal winds (and associated GW filtering) exhibiting significant spatial (in altitude and latitude) and temporal variability at the bottomside F layer. However, the tendency for eastward tidal winds to arise in the early evening hours appears to be consistent with the needed orientation of neutral atmosphere perturbations potentially contributing to RTI seeding and plasma bubble generation. This also implies preferential westward GW propagation at these altitudes, though weaker winds at lower altitudes would allow more isotropic GW propagation.
Assuming first that ρ e rather than ρ e /ρ e0 is constant with altitude, because of the apparent maxima of the latter between ∼250 and 300 km and the apparent decrease above, we can express the local electron density gradient as (again assuming only advective effects) H e + ∂ ∂z ρ e exp(imz) Fig. 18. Profiles of zonal neutral wind (U n , solid) and plasma motion (U pl , dashed) in the bottomside F layer (left) expected under ESF and RTI seeding conditions. At right are shown surfaces of constant neutral and electron densities (green, assuming neutral advection) due to an upwardand westward-propagating GW with maximum upward and downward displacements (blue and red dashed lines) regions of minimum and maximum neutral and electron densities, respectively. Neutral zonal and vertical winds are shown with blue arrows, the black circle shows the expected plasma motion , and the pink region shows where neutral GW motions oppose horizontal and vertical plasma motions (contributing to polarization electric fields) and enhance the vertical electron density gradient, all of which enhance the expected plasma instability growth rates anticipated by Kudeki et al. (2007).
The ratio of perturbation to mean contributions is then ∼(ρ e /ρ e0 )m H e within the phase of the GW having the largest positive gradient. For a GW with a vertical wavelength of λ z ∼ 100 km, which is consistent with apparent vertical structure in the electron density fluctuations seen in Fig. 9 and with expectations of assessments of GW wavelength variations due to increasing kinematic viscosity in the TI (Vadas and Fritts, 2006;Vadas, 2007;Fritts and Vadas, 2008), this ratio maximum is ρ e ρ e0 m H e ∼ 0.5 to 1 for the ranges of reasonable GW amplitudes, wavelengths, and plasma scale heights H e anticipated at the bottomside F layer.
We are now in a position to assess how the various GW perturbations of horizontal and vertical winds and plasma density gradients may impact the plasma growth rates discussed above (with the above caveats about the neglect of chemical and electrodynamics effects). We begin by summarizing the mean structure of the neutral thermosphere and ionosphere anticipated to yield significant plasma growth rates by Kudeki et al. (2007). These conditions are shown in the left panel of Fig. 18. The important features appear to be 1) an eastward neutral wind at higher altitudes (due largely to tidal motions that only become significantly eastward at the time of observed plasma bubble formation), 2) a westward plasma motion partially decoupled from the neutral motions at the bottomside F layer, and 3) a strong vertical plasma (and electron density) gradient at the same altitudes. Considering GW influences on both neutral winds and electron density and plasma gradients, there are two possible GW orientations that imply different contributions to key terms in the instability growth rates. A GW having westward propagation and westward tilts of phase surfaces with altitude (as depicted in the right panel of Fig. 18) will provide the optimal enhancements of the important terms in the plasma growth rate suggested by Kudeki et al. (2007) in a common volume (or more likely in multiple similar volumes), specifically plasma density gradient and (U n − U p ) enhancements (see Eq. (2)). Westward-propagating GWs also have downward motions where eastward motions and electron density gradients are enhanced. But this is also considered to be beneficial for plasma instabilities, especially the RTI (Fejer, 1996).
Thus we assume an upward-and westward-propagating GW will contribute greater potential enhancements to the growth rate in Eq. (2) for the sake of this discussion. The right side of Fig. 18 shows surfaces of constant electron density (per unit mass, in green) for this GW orientation, with maximum upward and downward displacements (blue and red dashed lines) corresponding to the regions of minimum and maximum electron densities, respectively. Neutral GW wind perturbations (with zonal and vertical components anti-correlated) are shown with blue arrows, and the black circle shows the expected plasma motion leading to instability growth in the westward and upward phase of plasma motion . The pink region shows where neutral GW motions oppose horizontal plasma motions (increasing the neutral-plasma velocity difference by up to ∼20%, thus amplifying polarization electric fields) and enhance the vertical electron density gradient, all of which enhance the expected plasma instability growth rates anticipated by Kudeki et al. (2007) in this region of the flow. Thus, we conclude that GW perturbations may have substantial influences on plasma instability growth in two ways: 1) they may significantly enhance plasma instability growth rates through perturbations of the neutral and plasma quantities upon which the growth rates depend and 2) they may induce "seed" perturbations at suitable spatial scales, allowing more rapid attainment of finite-amplitude perturbations extending to higher altitudes. Finally, the westward intrinsic phase speed of the hypothesized GW of (c − U n ) ∼ N λ z /2π ∼ 160 m s −1 implies westward propagation, a phase motion close to that of the westward plasma drift, and hence a much longer time for coherent perturbations to act on the underlying plasma than suggested by the much higher GW intrinsic frequency. Similar arguments can be made about the potential effects of GWs propagating eastward or superpositions of GWs having various propagation directions (see Fritts et al., 2008b). These influences will vary with the relative contributions of the GW to horizontal and vertical velocity and plasma density gradient fluctuations. In either case, it appears to us that it is likely that large-amplitude GWs at the bottomside F layer contribute to growth of the plasma instabilities and plasma bubbles extending to much higher altitudes.

Summary of Initial SpreadFEx Results
While our summary of SpreadFEx analyses presented here is not complete, there are a number of preliminary statements that we can make about the neutral and plasma dynamics observed during the campaigns based on these results. We list here those results that appear to be defensible at this stage, and refer readers to our forthcoming Spread-FEx special issue of Annales Geophysicae for more comprehensive analyses. Current conclusions include: a) deep convection appears to be a significant and viable source of GWs penetrating into the MLT and to the bottomside F layer, depending on GW scales, frequencies, and propagation directions; b) GWs spanning a wide range of scales and frequencies can reach the MLT, but only GWs having large scales and high frequencies can reach higher altitudes; c) GWs arising from convection are strongly influenced by lower atmospheric and MLT winds that dictate which GW frequencies, scales, and propagation directions allow penetration to the bottomside F layer and above; d) digisonde electron densities suggest GWs having downward phase motions (upward propagation), periods of ∼20 min to 2 hr, and significant electron density perturbations at the bottomside F layer, with corresponding implied neutral density and wind perturbations; e) GPS data suggest GWs of similar periods extending to even higher altitudes before and after plasma bubble occurrence; f) smaller-scale GWs and wave-mean flow interactions appear to excite bores in the MLT that exhibit a variety of dispersive and nonlinear responses; g) VHF radar and digisonde data suggest a connection between rising plasma motions, decreasing electron densities, increasing plasma gradients, and plasma bubble initiation; h) VHF radar, digisonde, GPS, airglow, and GUVI data suggest an ability to quantify plasma perturbations in altitude, longitude, and time; however, it remains unclear how these perturbations are related to plasma bubble generation; i) neutral zonal tidal winds appear to play an important role in enabling spread F seeding, given the transition from westward to eastward neutral motions at ∼250 km and above in the early evening hours; j) the multiple correlative data sets collected during the SpreadFEx campaigns are among the most comprehensive available, and will likely enable more quantitative conclusions upon completion of our various analyses; k) our initial results lead us to conclude that GWs may have sufficient perturbation amplitudes and impacts on instability growth rate at the bottomside F layer that their potential role in ESF and plasma bubble seeding should not be discounted; and l) it remains to be quantified fully how the GW neutral and electron density perturbations identified in our SpreadFEx experiment affect the generation of ESF and plasma bubbles extending to higher altitudes.