Previous incoherent radar studies at Arecibo Observatory, Puerto Rico have demonstrated that ∼1–3% electron density “imprints” of internal gravity waves are routinely present in the Arecibo thermosphere (∼118–500 km). A special radar technique involving photoelectron-enhanced plasma waves (PEPWs) was used for these observations. Recently, it was discovered that the trails of the gravity waves can be detected in standard incoherent scatter power profiles when properly filtered. This result was validated using simultaneous PEPW observations. This new development opens up the possibility of monitoring thermospheric gravity waves day and night. Preliminary studies indicate that gravity waves are continually propagating through the Arecibo thermosphere, and that “sets” of waves separated by approximately 20–60 min are typically present. With the aid of additional radar tests, it may be possible to unlock Arecibo power profiles recorded over the past 30 years for gravity wave studies. The precise origin of the waves is currently unknown.
 In the past, very accurate measurements of electron density were made at Arecibo Observatory, Puerto Rico, by applying the coded long-pulse (CLP) radar technique [Sulzer, 1986] to plasma line echoes enhanced by daytime photoelectrons (PEPL). In the lower thermosphere above Arecibo, background neutral waves couple to the ionospheric plasma, routinely yielding ∼1–3% electron density “imprints” of the waves [Djuth et al., 1997]. These imprints are present in all observations made to date; they are decisively detected at 30–60 standard deviations (σ) above the “noise level” imposed by the measurement technique. Complementary analysis and modeling efforts provide strong evidence that these fluctuations are caused by internal gravity waves. Properties of the neutral waves such as their period and vertical wavelength are closely mirrored by the electron density fluctuations. Frequency spectra exhibit a high-frequency cutoff at the Brunt-Väisälä frequency, and vertical half wavelengths are typically in the range 2–25 km between 115- and 160-km altitude and 50–150 km above ∼170 km altitude.
 In general, the observed electron density imprints vary smoothly with altitude and their vertical wavenumber (km−1) spectrum is characteristically narrow-banded. The bandwidth ranges from 0.01 km−1 to 0.003 km−1 below 130-km altitude and is less than 0.003 km−1 at higher altitudes. Djuth et al.  estimate that perturbations in the horizontal neutral wind field as small as 2–4 m/s can give rise to the observed electron density fluctuations. The basic mechanism entails redistribution of ionospheric plasma in the vertical direction under the action of a mostly-horizontal oscillating neutral wind field. The propagation/oscillation of this acoustic-like wave is made anisotropic by gravity. Above ∼130 km, neutral motion parallel to the geomagnetic field moves plasma up and down field lines; at lower altitudes plasma motion is primarily determined by the wind-induced Lorentz force that acts on the ions. The required wind speed for imprints can be significantly greater than the above minimum values depending on the orientation of the neutral wave's horizontal wave vector relative to the geomagnetic field. Horizontal wind vectors are not known for any existing CLP PEPL observations. Limited observations with extended altitude coverage indicate that wave imprints can be detected at thermospheric heights as high as 500 km.
 Overall, the application of the CLP technique to PEPLs yields important information about neutral wave dynamics in the upper atmosphere, but it does have its limitations. This technique is effective only during the daytime when large fluxes of photoelectrons are present in the Arecibo ionosphere. On average, good observing conditions exist for ∼7 hours a day. The limited diurnal coverage is arguably the greatest shortcoming of the technique.
2. Discovery of Gravity Wave Trails in Incoherent Scatter Power Profiles
 In the past, incoherent scatter radar measurements of electron density, electron/ion temperature, and ion velocity have been used to detect gravity waves in the thermosphere [e.g., Hocke and Schlegel, 1996; Kudeki et al., 1999]. The Arecibo observing program that led to the detection of gravity wave trails in power profiles entails the cyclic transmission of three different types of radar pulses: Barker-coded pulses for power profiles (BPP), a pseudo-random phase-coded pulse for CLP PEPL observations, and a multi-frequency radar pulse used to determine plasma temperatures from the ion-line autocorrelation function. The repeat time of this pulse cycle is typically 30–40 ms.
 The BPP data are digitally processed to isolate time varying, medium scale, electron density perturbations. Raw BPP data are averaged in 3000-pulse intervals (120 s), and then the corresponding system noise is subtracted from the profile. Thereafter, the profiles are range-squared-corrected and normalized by the average system noise for the entire run. Subsequently, a digitally filtered background profile is derived for each processed raw profile. A simple two-way averaging filter having a length (in range) of ∼30 km is used for this purpose. The filtered profile is then subtracted from the processed raw profile to obtain the final residual profile expressed in units of signal-to-noise ratio, S′/N. Each 120-s residual profile represents one column of pixels in a Range-Time-Intensity (RTI) display. Because the BPP sensitivity to electron density perturbations is much less than that of the CLP PEPL, well-defined gravity wave imprints [Djuth et al., 1997] are not detectable with the BPP. Wave “trails” corresponding to CLP PEPL imprints are observed instead. By definition, trails are sloping contours of enhanced radar backscatter that correspond to the regions of enhanced CLP PEPL electron density. Because of statistical limitations, trails do not resolve the detailed spatio-temporal structures evident in the CLP PEPL results. The trails have a statistical significance of only about 2–3 σ compared to 30–60 σ for the CLP PEPL electron density variations. Nevertheless, the BPP trails provide an effective means for detecting gravity wave imprints and pave the way for more comprehensive Arecibo studies involving nighttime observations.
Figure 1 illustrates the typical signature of the gravity wave imprints as observed with the CLP PEPL technique. All PEPLs are corrected for electron temperature, Te. Figure 1 is similar to others presented by Djuth et al. . Comparisons between simultaneous CLP PEPL results and filtered BPP observations are provided in Figures 2 and 3. These data were acquired in 1992 and 1998, and comparable results were obtained in 1991 and 1993. The plot on the far left in Figure 2 contains residual CLP PEPL electron density data expressed as a percentage of the mean profile. Data are displayed in a color contour representation instead of the waterfall plots shown in Figure 1. This was done to facilitate comparisons with the BPP measurements. The center panel contains digitally filtered BPPs in RTI format. “Trails” evident in the BPP results match one-to-one with “sets” of downcoming waves detected with the CLP PEPL. A “set” is a collection of profiles charting the movement of electron density enhancements from high altitude (∼180–200 km) to low altitude (∼120 km). The time period between sets is measured using consecutive electron density maxima (in units of positive percentage) in the altitude interval ∼180–200 km; only high-altitude maxima having underlying electron density enhancements near 120 km altitude are employed. The times between gravity wave sets range from ∼20–60 min in the observations made to date. The downward tilt in the trails and CLP PEPL electron density enhancements approximates the vertical phase velocity of the gravity wave. At lower altitudes, the waves curve toward increasing time because the wavelength (and therefore the phase velocity) decreases as the wave moves downward. This is a product of gravity wave propagation in a dispersive medium. The tilt in the trail arising from the downward gravity wave phase velocity together with the trail curvature at low altitudes are unique signatures of gravity wave sets/trails. The panel on the far right of Figure 2 is the range-squared corrected power profile from which the center panel was derived. Careful inspection of the full power profile reveals subtle variations caused by the gravity wave trails. However, it is clear that in the absence of digital processing of the profile, trails would never be seen with clarity. The percentages of wave-induced perturbations in the incoherent scatter power profiles roughly match those of the CLP PEPL technique. They are not identical because the power profiles are not corrected for Te/Ti, where Ti is ion temperature.
 The data contained in the left and center panels of Figure 3 show features similar to those of Figure 2, but these observations were made five and a half years later. The panel on the far left provides an expanded view of the CLP PEPL results, and the center panel contains the filtered BPP data. The panel on the far right shows BPP results obtained on the day following the example provided in Figure 2. As is typical, trails in the lower thermosphere are evident and some of the trails extend to higher heights near 300 km and beyond. The ability of the filtered BPP subtraction technique to detect trails in the upper thermosphere is dependent in part on the bottomside gradient and the scale length of the F region perturbations. The filter does not have a particularly favorable response to a steep bottomside gradient, and this yields a region of diminished sensitivity in the filtered profiles (purple region between 200 and 250 km in Figures 2 and 3). Nevertheless, we retained the filter for the current study because it is very simple and beyond reproach. In addition, it is clear from Figure 13 of Djuth et al.  that while a filter length of 30 km may exhibit good performance at altitudes below ∼170 km, at higher altitudes filter lengths greater than 60 km are more appropriate. In principle, we should increase the filter length with altitude, but we elected not to do this to minimize the potential for introducing artificial features into the filtered results.
 All existing daytime data sets containing simultaneous high-temporal resolution CLP PEPL and BPP data have been processed and the results are similar to those of Figures 1–3. The observations consist of one data set acquired in May 1991, three obtained in July 1992, one taken in July 1993, and one secured in February 1998. Each data set contains ∼6–7 hours of contiguous observations with gravity waves present throughout. In addition, one CLP PEPL data set from 1993 was processed without BPP data, and we recently discovered and processed two evening/nighttime power profile data sets from September 1994 and one from February 1998. All current data samples exhibit gravity wave sets/trails, and it is possible that most, if not all, Arecibo observations will yield similar results.
3. Nighttime Power Profile Observations
 Nighttime power profile data with full altitude coverage are currently lacking, and therefore the pervasiveness of gravity waves in the upper thermosphere cannot be established at this time. However, wave activity in the lower thermosphere can be investigated. Residual power obtained from one of three nighttime data sets with limited altitude coverage is presented in Figure 4. In this case, we employed a filter length of 37.5 km instead of the 30 km value used for the daytime observations. During the periods 1800–2200 AST (local time) and 0330–0600 AST, there is little background ionization in the lower thermosphere from ∼120 to 220 km altitude, and during the period 2300–0330 AST no significant background ionization is present between ∼130 km and 260 km altitude. Tilted structures resembling the daytime gravity wave trails are evident during the time intervals 1730–2200 AST and 0600–0730 AST at altitudes between 125 km and 260 km. The tilted structures observed from ∼1730 AST to 1830 AST in the altitude interval 220–260 km and from 0630 to 0730 AST in the height range 195–260 km are imposed on a background plasma of significant density (ne = 2 − 4 × 104 cm−3, where ne is electron density). These structures are most likely caused by the redistribution of the background plasma by gravity waves in a manner similar to that observed during the daytime.
 The tilted structures in the altitude interval 125–200 km between 1730 and 2200 AST and between 0630 and 0730 AST represent relatively weak signals. A similar very weak, tilted structure is also observed near 140 km at ∼0530 AST. The presence of such weak, tilted backscatter in the lower thermosphere at night was first noted by Mathews et al. , who termed it “plasma rain.” This terminology will be adopted here. In Figure 4, most of the backscatter from plasma rain in the lower thermosphere is so weak that small biases created by the action of the averaging filter on the bottomside F region, the decaying E region and sporadic E yield measurable S′/N offsets from 0. Thus some of the structures shown as a purple color (−0.2 S′/N) are actually slightly positive perturbations. In Figure 4, the plasma rain phenomena occurs both in the evening and early morning when the bottomside F region is sufficiently low in altitude (∼200–250 km altitude). After sunset in the lower thermosphere, the detectable radar backscatter from background O2+/NO+ ionization quickly becomes diminishingly small. Thus, there is no observable background plasma for gravity waves to redistribute. Nevertheless, the continuity between the gravity wave trails at higher altitudes and the plasma rain at lower heights (e.g., in the time interval 1730–1900 AST) indicates that the two phenomena are related. Most likely, the downward phase progression of gravity waves and the associated electrodynamic response of the ionosphere are transporting a small amount of plasma from the bottomside F region into the lower thermosphere. This transport also takes place during the daytime, but the plasma redistribution effect tends to overwhelm the downward transport process. By comparing the plasma rain S′/N values of Figure 4 with the daytime S′/N values of Figures 2 and 3, one finds that plasma transport on average contributes less than 10% to trail amplitude. When the bottomside of the F region moves above the 260-km data cutoff between 2200 and 0400 AST, no plasma rain is evident. However, the downward descent of the ionization near 130 km may be gravity wave related.
4. The Arecibo Mystery
 Although the nature of the observed CLP PEPL gravity waves is clear, the reason why the waves should be coherent (i.e., narrow-banded in vertical wavenumber) and ubiquitous is a mystery. Recent observations of coherent waves in the Arecibo D region between ∼65 and 85 km by Zhou  add to the mystery. An intriguing aspect of the CLP PEPL/BPP data described above is that the gravity waves tend to be grouped into sets separated by 20–60 min.
 Our initial Arecibo results indicate that the waves are not episodic, but are probably always present above Arecibo. Long wavelength ocean waves are a potential source of the observed gravity waves, but this has yet to be established. If this were true, it might explain why seed irregularities needed to excite the generalized Rayleigh-Taylor instability are periodically present in the equatorial ionosphere. The connection between gravity wave filtering and wave breaking/saturation in the Arecibo mesosphere and the observations in the thermosphere is not known, but the resulting local forcing of the mean flow may produce secondary gravity waves that propagate into the thermosphere [e.g., Fritts et al., 2002]. Moreover, the coherence of the gravity waves seen in the daytime thermosphere with the CLP PEPL technique may require an atmospheric tuning process in the lower thermosphere [Walterscheid et al., 1999].
 Our preliminary study indicates that gravity waves are continually propagating through the Arecibo thermosphere. Incoherent scatter radar measurements of photoelectron-enhanced plasma lines yield extremely accurate measurements of gravity wave imprints, but this technique is effective only seven hours a day. Gravity wave trails detected in the Arecibo radar power profiles allow the presence of thermospheric gravity waves to be monitored at all hours of the day.
 Some of the Arecibo data in the NCAR World Database may be usable for investigations of thermospheric gravity waves. We are currently acquiring calibration data to determine the proper processing algorithm for gravity wave trails observed during radar beam scans. If this effort is successful, power profiles recorded over the past 30 years at Arecibo will become available for gravity wave investigations. To broaden the database, we invite members of the aeronomy community to re-examine their Arecibo observations for the presence of gravity wave trails.
 Support from the NSF under grant ATM-9529392 is gratefully acknowledged by F.T.D. and J.H.E. J.D.M.'s effort is supported under NSF grant ATM-0108600. The Arecibo Observatory is part of the National Astronomy and Ionosphere Center, which is operated by Cornell University under cooperative agreement with the National Science Foundation.