F2-region atmospheric gravity waves due to high-power HF heating and subauroral polarization streams


Corresponding author: E. V. Mishin, Space Vehicles Directorate, Air Force Research Laboratory, 3550 Aberdeen Ave. SE, Kirtland AFB, NM 87117, USA. (evgeny.mishin@us.af.mil)


[1] We report the first evidence of atmospheric gravity waves (AGWs) generated in the F2 region by high-power HF heating and subauroral polarization streams. Data come from the CHAMP and GRACE spacecraft overflying the High-frequency Active Auroral Research Program (HAARP) heating facility. These observations facilitate a new method of studying the ionosphere-thermosphere coupling in a controlled fashion by using various HF-heating regimes. They also reveal the subauroralF2 region to be a significant source of substorm AGWs, in addition to the well-known auroralE region.

1. Introduction

[2] The energy released during geomagnetic (sub)storms is ultimately dissipated in the thermosphere. One of the major indicators of the thermospheric response is generation of atmospheric gravity waves. Since the neutral gas density N is much greater than the plasma density n, it is conventionally assumed that the thermosphere responds in ≥30 min. However, some observations reveal faster response times [e.g., Williams et al., 1993; Aruliah et al., 2005]. Millward et al.'s [1993]simulations show that a 0.2-V/m, 10-min saw-tooth electric pulse produces plasma upflowsUi ∼ 100 m/s and vertical winds Un ≈ Ui near the F2-layer peak at 300 km. In ∼5 min the thermosphere swelled by one atmosphere scale height. Then, the swelling propagated away and ceased, overall resembling substorm AGW surges [e.g.,Hocke and Schlegel, 1996].

[3] Subauroral polarization streams (SAPS), i.e., structured flows of enhanced plasma convection equatorward of the auroral boundary, quickly develop during substorms [e.g., Mishin and Burke, 2005]. Wang et al. [2011] found that the mass density at altitudes h ≈ 400 km enhances within SAPS on average by ≈10% with respect to that without SAPS. This indicates efficient thermospheric heating by ion neutral friction in the F2 region within SAPS. Understanding the development of the thermospheric response in space and time requires coordinated observations of neutral gas, ionospheric plasma, electromagnetic fields, and particle precipitations. This is a formidable task for dynamic localized natural phenomena.

[4] On the other hand, plasma upflows Ui∼ 100–300 m/s are routinely produced by high-power high frequency (HF) radio waves injected into theF2-region ionosphere [e.g.,Rietveld et al., 2003; Milikh et al., 2008, 2010; Kosch et al., 2010]. Given strong F2-region ionosphere perturbations, it seems worthwhile to explore the thermosphere's response to high-power HF injections. The objectives of this paper are twofold. First, we report on novel observations of HF-induced AGWs in theF2-region during the October 2008 and August 2011 heating campaigns at the High-frequency Active Auroral Research Program (HAARP) facility (http://www.haarp.alaska.edu/gen.html). We also found substorm SAPS-generated AGWs, in addition to the auroral source. Data come from the CHAllenging Minisatellite Payload (CHAMP) [Reigber et al., 2002] and twin Gravity Recovery And Climate Experiment (GRACE) [Tapley et al., 2004] satellites.

2. Observations

[5] During the campaigns, O-mode waves were transmitted toward the HAARP magnetic zenith (MZ) at full power 3.6 MW (the effective radiative power ERP = 450–650 MW) with either 0.5 Hz or 5 Hz 50% square modulation. TheF2-peak plasma frequency exceeded the heating frequencyf0by ≥0.5 MHz, ensuring HF beam-ionosphere interaction near 220–240 km [e.g.,Mishin et al., 2005]. There have been two CHAMP/HAARP (CH1 and CH2) and two GRACE/HAARP (GH1 and GH2) experiments in October 2008 and August 2011, respectively. Table 1 lists their dates, f0(MHz)/ERP (in MW), heater-on periods, UT-time Tmz of the satellite closest approach to MZ at the geographic longitude ≈214.1°E and latitude Latmz (see Table 1), and geomagnetic activity. The latter was checked using data from the WDC for Geomagnetism, Kyoto and the University of Alaska GIMA magnetometer array. We note exceptionally quiet conditions prior to and during the CH1 and GH2 experiments. To compare with CH1 and GH2, we selected overflights close in Tmz and distance from Latmz without HF heating during quiet (q) and disturbed (d) times. They are listed in Table 1 as Xqi and Xdi, where X stands for CHAMP or GRACE and i = 1, 2, 3.

Table 1. HF-Heating and Overflights Conditions
HF-on (UT)22:14 ÷ 3421:50 ÷ 70no16:05 ÷ 2504:10 ÷ 30nononono

[6] Thermospheric mass densities ρ = M ⋅ N (M is the average mass of neutral constituents) at CHAMP altitudes h≈ 330 were inferred from the STAR accelerometer data obtained at 0.1-Hz sampling rate (the Nyquist frequencyfN = 0.05 Hz) [e.g., Sutton et al., 2005; Sutton, 2009]. Figure 1 shows mass densities ρ, their trends ρtr, and relative residuals δρ/ρtr = ρ/ρtr − 1 vs. UT-Tmz during the CHAMP overflights CH1, Cq1, and CH2. Here ρtr is obtained by a simple polynomial fit within ±180 s about Tmz. Figure 2 shows filtered mass densities ρλvs. UT-Tmz, power spectral densities (PSD) of δρ/ρtrin the interval 57° ≤ GLat ≤ 67°, and GLat-λ color spectrograms. The auroral boundaries are found by inspecting the AFRL/DMSP database.

Figure 1.

Neutral mass densities ρ (solid lines) and their trends ρtr (dashed) during (a) CH1, (b) Cq1, and (c) CH2 and (d) their relative residuals δρ/ρtrvs. UT-Tmz.

Figure 2.

Filtered neutral mass densities ρλvs. UT-Tmz during (a) CH1 and (b) CH2 and Cq1, (c) power spectral densities (PSD) of δρ/ρtr, and spectral amplitudes of the time series along the orbit for (d) CH1 and (e) CH2. Color codes in linear scale for the wave spectrum are given to the right of the spectrograms. Vertical dashed and dash-dotted lines indicate Latmz and the auroral zone boundary, respectively. Black dashed lines encircle the regions of interest (see text).

[7] It is assumed that the satellite is sampling the spatial, rather than the temporal variations of the neutral density, so λ = vsat/f (vsat = 8 km/s) are wavelengths along the satellite track and f are apparent frequencies. This requires the actual wave period Tλ ≫ 1/f, which holds for λ < 2000 km at Tλ ∼ TBV. Here TBV ≈ 13 (18) min at h≈350 (500) km is the Brunt-Väisäla (BV) cutoff (the waves atTλ ≤ TBV are buoyancy waves). To obtain ρλ, we applied a passband filter preserving oscillations in the range 200≤ λ <2000 km. PSD and spectrograms are obtained by using a standard FFT of δρ/ρtr and complex morlet wavelet decomposition of the measured time series of ρ, respectively.

[8] The twin satellites GRACE A and Bwith the SuperSTAR accelerometer, which is an order of magnitude more precise than CHAMP/STAR, flew approximately 25 s apart in a near-circular orbit ath≈ 470 km. To maintain the line-of-sight between the satellites, hundreds of micro-thrusts per day are performed. Fortunately, during all listed overflights maneuvers had occurred either well before or after Tmz and thus not affected the results. We explore neutral mass densities ρinferred from 1-Hz accelerometer measurements and sampled at a frequency of 0.2-Hz (fN = 0.1 Hz) corresponding to the available satellite attitude data. We present only data from GRACE A, for GRACE B gives virtually identical results after accounting for the time delay. Figure 3 shows neutral mass densities ρ and relative residuals δρ/ρtrvs. UT-Tmz, and PSD of δρ/ρtrat 57° ≤ GLat ≤ 67° for all listed UT-morning overflights. GLat-λ color spectrograms are shown for the GH2, Gq3, and Gd1 overflights. These results have been obtained by the same methods as above. Note that the characteristic spatial scales for CHAMP are smaller than those for GRACE.

Figure 3.

(a) Neutral mass densities ρ and (b) their relative residuals δρ/ρtrvs. UT-Tmz during Gq1 (dashed lines), Gq2 (dash-dotted),Gq3, Gd1 (thin solid), and GH2 (solid), (c) power spectral densities of δρ/ρtr, and spectral amplitudes of the time series along the orbit for (d) GH2, (e) Gq3, and (f) Gd1. Vertical dashed and dash-dotted lines indicate Latmz and the auroral zone boundary, respectively. Black dashed lines encircle the regions of interest (see text).

3. Discussion and Conclusions

[9] The spectrograms indicate waves in and equatorward of the auroral region. We specify their source using the fact that short-scale AGWs decay faster than long-scale waves [e.g.,Hocke and Schlegel, 1996]. This comes with the corollary that near Latmz long scales dominate the wave spectrum arriving from the (distant) auroral source, while the short scales originate from a local source. First, we discuss the heating experiments. As SAPS are often observed near HAARP fairly long after substorms [e.g., Pedersen et al., 2007], we checked the AFRL/DMSP database. No SAPS-like flows have been found in the subauroral region near/during all quiet-time overflights, includingCH1 and GH2 (Figures 2d, 3d, and 3e). This fact allows us to explore the heating effect by comparing the quite-time wave distributions with and without HF heating.

[10] GLat-λ spectrograms help visualizing the wave spatial distribution. Note that CHAMP (GRACE) was moving southward (northward). Apparently, only the long waves λ ≥ 500 km (CH1) and ≥1500 km (GH2 and Gq3) are seen continuous between the auroral zone and ≤65° GLat. This agrees with PSD for Cq1 (Figure 2c) and Gq1-3 (Figure 3c) comprising mainly of the long waves. Quite the contrary, during the heating experiments the short waves λ ≈ 350 km (CH1) and 1000 km (GH2) dominate near Latmz (these regions are encircled by black dashed lines in Figures 2d and 3d). Therefore, we attribute them to the heating effect.

[11] A comprehensive model of HF-induced thermospheric perturbations has not yet been developed. High-power radio waves transfer energy to the ionospheric plasma via excited plasma waves [e.g.,Fejer, 1979]. Resulting enhanced plasma temperature and ion upflows streaming from the HF-heated spot are regularly observed. During theCH1-2 experiments, the HAARP digisonde was monitoring plasma below and above the heated layer using three probing frequencies fp, one below and two above f0. Reflections (echoes) of probing signals are placed on the skymap plane using their zenith and azimuth angles of arrival [e.g., Reinisch et al., 1998]. Inspection of the skymaps for CH1 shows that a minute after the heater was turned on, a tight cluster of echoes appeared in the F2 layer within −3°/+8° about MZ and remained during the heating.

[12] Figure 4 shows a Doppler skymap taken 30 s before Tmz for CH1 and raw spectra of the signal Doppler shift ΔfD ≃ fpvlos〉/2c measured at h = 220 to 255 km in one antenna channel. Here fp ≈ 3 MHz and 〈vlos〉 is the average line-of-sight speed. The spectra indicate irregular plasma upflows at speedsUi ≈ 80 to ≈100 m/s slightly increasing with h, as indicated by the dashed line [cf. Milikh et al., 2010, Figures 2 and 3]. As both the ion outflow and AGWs, coincident with the δρ/ρ-swelling (Figure 1d), are observed near Latmz, it seems reasonable to assume that the plasma and neutral perturbations are interrelated.

Figure 4.

(top) The HAARP digisonde Doppler skymap 30 s prior to Tmz for CH1 and (bottom) Doppler shift ΔfD ≈ f0vlos〉/2c for a given probing frequency at different altitudes between 220 and 255 km. The color is mapped to the Doppler frequency shift in Hz measured along the line of sight to the echo. Negative Doppler shifts indicate the upward motion. The dashed red line connects the medians of the spectra.

[13] The collisional heating of the neutral gas (mainly atomic oxygen) by O+ ions can be estimated as inline image. Here ΔTi is the increase in the ion temperature and νin ≈ 1.8⋅10−9 N s−1 is the bulk O+-O collision frequency [cf. Rees, 1989]. There are no means of measuring the ion temperature at HAARP. Therefore, we use the values of ΔTi ≤ 300 K that have been revealed by the EISCAT UHF radar during the Ui ≥ 100 m/s events caused by HF heating in the conditions close to that at HAARP [Rietveld et al., 2003; Kosch et al., 2010]. Taking n ≈ 3 ⋅ 105 cm−3 near the F2 peak at h ≃ 300 km, we roughly estimate the neutral temperature gain from the collisional energy exchange between O+ ions and atomic oxygen as ≃3 K/min. This gives ΔTn ≃ 30 K in a 10-min heating time or ΔTn/Tn ∼ 3%. Thus, the local swelling is anticipated to be ≤3%. This is consistent with the observed δρ/ρ and with Millward et al.'s [1993] δρ/ρ ∼ 1, which was calculated at the neutral heating rate ΔTnt ∼ 100 K/min.

[14] Finally, the above corollary is used to explore the disturbed-time AGWs. From the DMSP data close to the disturbed overflights, we found enhanced SAPS-like flows at subauroral latitudes. Here, as showFigures 2e and 3f, rather wide AGW spectra are enhanced (encircled by black dashed lines). Thus, we conclude that SAPS not only enhance neutral density [Wang et al., 2011] but also generate AGWs in the F2 region. The latter does not seem surprising, given spatial and temporal variations within the SAPS channel [e.g., Mishin and Burke, 2005]. The dominant wavelengths (encircled) seem to match the distance over which the waves were observed. That is, 450/1000 km is about 4°/9° in latitude. This suggests that there is an impulsive localized source. More observational and theoretical/modeling efforts are required to understand the underlying generation mechanism.

[15] In conclusion, the exact mechanisms behind the generation of HF- and SAPS-related atmospheric gravity waves remain to be determined. The experimental results presented here provide the groundwork for future investigations of the ionosphere-thermosphere coupling by means of HF heating in order to advance a much needed scientific understanding of the thermospheric response to SAPS-related phenomena.


[16] HAARP is operated jointly by the U.S. Air Force and U.S. Navy. Work at AFRL was supported by the Air Force Office of Scientific Research. GM was supported by DARPA via a subcontract N684228 with BAE Systems. The CHAMP satellite was sponsored by the Deutsche Forschungsanstalt für Luft und Raumfahrt (DLR) and operated by the DLR Space Operations Center. GRACE is a joint partnership between NASA and DLR.

[17] The Editor thanks two anonymous reviewers for assisting with the evaluation of this paper.