Volume 7, Issue 4 e2019EA000867
Research Article
Open Access

All-Sky Tracking of Sporadic urn:x-wiley:ess2:media:ess2506:ess2506-math-0002 Field-Aligned Irregularities as a Novel Probe of Thermospheric Winds

J. F. Helmboldt

Corresponding Author

J. F. Helmboldt

U.S. Naval Research Laboratory, Washington, DC, USA

Correspondence to: J. F. Helmboldt,

[email protected]

Search for more papers by this author
G. B. Taylor

G. B. Taylor

Department of Physics and Astronomy, University of New Mexico, Albuquerque, NM, USA

National Radio Astronomy Observatory, Charlottesville, VA, USA

Search for more papers by this author
First published: 20 April 2020
Citations: 2


This paper presents the results of an analysis of a unique observing campaign that tracks groups of irregularities associated with sporadic urn:x-wiley:ess2:media:ess2506:ess2506-math-0004 ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0005) via their coherent backscattering of a very high frequency transmitter of opportunity (an analog TV station). The unique all-sky imaging capability of the first station of the Long Wavelength Array used in this campaign, allowed for the identification and tracking of multiple groups of irregularities, or “clouds,” over the entire visible sky. This all-sky tracking yielded horizontal wind measurements from 75 distinct clouds observed within 18, 1-hr collections from May–September 2014 within the urn:x-wiley:ess2:media:ess2506:ess2506-math-0006 90- to 150-km altitude range. While unusually strong winds of 100–200 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0007 or more were observed, the winds generally agree with predictions from the updated Horizontal Wind Model. The exception to this was an offset in premidnight zonal winds, which the Long Wavelength Array observations indicated were systematically more westerly than predicted by the Horizontal Wind Model. We postulate that this may be due to a local excess in mountain waves, which is partially supported by tropospheric and stratospheric wind data from the National Oceanic and Atmospheric Administration North American Regional Reanalysis. In three out of the 18 collections, we also observe evidence of significant horizontal zonal wind shears (positive and negative) with magnitudes on the order of 1 m urn:x-wiley:ess2:media:ess2506:ess2506-math-0008 s urn:x-wiley:ess2:media:ess2506:ess2506-math-0009 km urn:x-wiley:ess2:media:ess2506:ess2506-math-0010.

Key Points

  • A novel method for all-sky tracking of radio frequency backscatter from sporadic urn:x-wiley:ess2:media:ess2506:ess2506-math-0011 irregularities is described
  • This presents a new way to probe lower thermospheric winds up to heights of about 150 km and over a large area
  • There were several instances of high winds ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0012 100–200 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0013) as well as a few occasions of horizontal zonal wind shears with magnitudes of urn:x-wiley:ess2:media:ess2506:ess2506-math-0014 1 m urn:x-wiley:ess2:media:ess2506:ess2506-math-0015 s urn:x-wiley:ess2:media:ess2506:ess2506-math-0016 km urn:x-wiley:ess2:media:ess2506:ess2506-math-0017

1 Introduction

One of the most dynamically variable portions of Earth's atmosphere is the lower thermosphere. The primary drivers of the wind profile within this region, especially at midlatitudes, are tidal pressure gradients related to large day/night temperature differences. Zonal wind shears associated with semidiurnal tides are a relatively common occurrence near urn:x-wiley:ess2:media:ess2506:ess2506-math-0018 110-km altitude. This general picture of lower thermospheric circulation has been directly observed over a number of years using a variety of techniques, including satellite-based measurements (e.g., Hedin et al., 1988), optical Fabry-Perot interferometers (e.g., Burnside & Tepley, 1989), incoherent scatter radars (e.g., Buonsanto et al., 1989), meteor radars (e.g., Hocking et al., 2001), and chemical releases (e.g., Larsen, 2002).

However, many of these same sensing methods have shown that there can be relatively large deviations within the wind profile from the typical, tidally driven behavior. For example, large peak winds in excess of 100 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0019 have been observed near urn:x-wiley:ess2:media:ess2506:ess2506-math-0020 100-km altitude as well as profiles with more irregular structure (see, e.g., Larsen, 2002). This can be the result of forcing from below by gravity/acoustic-gravity/acoustic waves. Gravity waves typically break below the thermosphere due to convective instabilities within the mesosphere, leading to a host of nonlinear effects that vastly complicate their disturbing impact on the lower thermosphere. In addition, the relatively high ion/neutral collision rate within this region, or the urn:x-wiley:ess2:media:ess2506:ess2506-math-0021 region in ionospheric parlance, leads to electric fields and currents that are almost entirely wind driven. In turn, electrobuoyancy forces within the lower ionosphere can feed back on the neutral component and distort it even further (see, e.g., Bernhardt, 2002; Cosgrove & Tsunoda, 2002). Electric fields associated with such ionospheric irregularities can interact and cause/enhance larger-amplitude disturbances at higher altitudes within the urn:x-wiley:ess2:media:ess2506:ess2506-math-0022 region. Thus, a thorough understanding of the complex dynamics of the lower thermosphere is also key to understanding structure formation within the larger ionosphere.

While the measurement techniques described above have provided decades of useful data for this purpose, they have some significant limitations. Methods based on both backscatter from meteor trails and optical airglow emission lines are limited to altitudes below urn:x-wiley:ess2:media:ess2506:ess2506-math-0023 100–110 km. This is especially true for nightside optical observations (Hocking et al., 2001; Killeen et al., 2006). Methods utilizing meteor trail reflections, satellite-based remote sensing, and/or incoherent scatter radars rely on the assumption that the horizontal wind is invariant over relatively large spatial and temporal scales (Hocking et al., 2001; Killeen et al., 2006; Zhang et al., 2003). Thus, localized and/or higher-altitude anomalies within the wind profile will be missed by such techniques. Chemical release experiments overcome these limitations but are limited in scope.

The research presented in this paper seeks to use another tracer for the neutral wind at these heights that can, to some degree, address the shortcomings discussed above. The tracers in question are field-aligned irregularities (FAIs) associated with midlatitude sporadic urn:x-wiley:ess2:media:ess2506:ess2506-math-0024 ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0025). At midlatitudes, the tidally driven zonal wind shears mentioned above lead to opposing, wind-driven electric fields that squeeze ions within the urn:x-wiley:ess2:media:ess2506:ess2506-math-0026 region into thin, dense layers. These layers take many minutes to form, requiring relatively long-lived metallic ions, usually Fe urn:x-wiley:ess2:media:ess2506:ess2506-math-0027 and Mg urn:x-wiley:ess2:media:ess2506:ess2506-math-0028, deposited by ablating meteors. (For a thorough review of these processes, see Haldoupis, 2012.) While often spoken about in terms of layers, urn:x-wiley:ess2:media:ess2506:ess2506-math-0029 typically consists of structures on scales of a few meters up to tens of kilometers. The bulk motions of these structures are dominated by the local neutral wind due to the fact that the ion/neutral collision frequency is much larger than the gyrofrequency at urn:x-wiley:ess2:media:ess2506:ess2506-math-0030 region altitudes. This makes urn:x-wiley:ess2:media:ess2506:ess2506-math-0031 a potentially useful probe of lower thermospheric winds.

Many authors have studied these structures using coherent backscatter radars aimed at FAIs typically associated with urn:x-wiley:ess2:media:ess2506:ess2506-math-0032 “clouds,” dense clumps of ions with sizes on the order of 10–50 km. Within such experiments, regions of backscatter can be observed drifting through the radar beam (see, e.g., Hysell et al., 2004, 2012; Larsen et al., 2007; Pan & Tsunoda, 1998). Additionally, when subjected to high-power radio waves, urn:x-wiley:ess2:media:ess2506:ess2506-math-0033 clouds emit enhanced airglow and can be tracked over the sky with optical all-sky cameras (see Bernhardt et al., 2003; Djuth et al., 1999; Kagan et al., 2000). The vertical structure of these urn:x-wiley:ess2:media:ess2506:ess2506-math-0034 clouds can also be probed with GPS radio occultations, but such observations yield little to no information about horizontal motions (see, e.g., Yue et al., 2015).

These experiments are somewhat limited by the fact that the motion of FAIs can be tracked with radars only within fixed beams or optically, but only when bombarded with high-power radio waves. For example, the 30-MHz radar used by Hysell et al. (2004) has a beam width of 10°. This paper details the results from an observing campaign that used a similar but distinct instrument. Here, an array of 256 bent dipole antennas was used with a transmitter of opportunity (an analog TV station) to monitor for very high frequency backscatter over the entire visible sky, not just within a single beam pointing. Thus, while the array has a relatively narrow beam (here, urn:x-wiley:ess2:media:ess2506:ess2506-math-0035 3°), it can be used to simultaneously form as many beams as necessary to cover the entire sky. Because of this, the motions of many groups of FAIs can be simultaneously tracked over a relatively large region, allowing for spatial variations in wind-driven drifts to be explored. Additionally, each group of FAIs can be tracked along a relatively long path (i.e., spanning several beams), reducing systematic errors in the measured drifts. The details of this observing campaign and the measurement methods used are described in section 2. Results are given in section 7 and discussed further in section 8.

2 Data Collection and Analysis

2.1 LWA1

All data used within this study were collected with the first station of the planned Long Wavelength Array (LWA), dubbed LWA1. LWA1 is an array of 256 bent, bowtie/dipole antennas confined to an approximately circular area with a diameter of roughly 100 m in a quasi-random configuration. It is located in the state of New Mexico at 34.070°N, 107.628°W and is currently operated as a stand-alone telescope.

The active electronics within the LWA1 antennas are optimized for 20–80 MHz, and observing is possible in the 10- to 88-MHz range. The array can be operated as a phased array with up to four electronically steered beams and 16 MHz of usable bandwidth. There are also two transient buffer modes that can record data from individual antennas. The narrowband mode (TBN) records data continuously at a sampling rate of 100 kHz, while the wide-band mode (TBW) can record the full output from the antennas (196-MHz sampling rate) in bursts of 61 ms every urn:x-wiley:ess2:media:ess2506:ess2506-math-0036 5 min. Both of these modes allow for all-sky images of radio emission to be generated by essentially beamforming after the fact. For more details of the LWA1 radio telescope, see Ellingson et al. (2013), Hicks et al. (2012), and Taylor et al. (2012).

2.2 Backscatter of Analog TV Signals

While primarily constructed as a telescope for targeting cosmic radio sources (e.g., pulsars and radio galaxies), LWA1 has been found to be a powerful probe of ionospheric structure when paired with transmitters of opportunity. Helmboldt et al. (2013) demonstrated that LWA1 and the WWV transmitters in Colorado can be used as a unique all-sky, bistatic ionospheric radar at 10, 15, and 20 MHz. At higher frequencies, reflections of TV signals off meteor trails (and aircraft) can be observed over the entire visible sky (see Helmboldt et al., 2014). In particular, the narrowband video carrier signal at 55.25 MHz from the TV station XEPM in Ciudad Juarez, Mexico provides a useful means to not only track backscattering sources over the whole sky, but to measure Doppler shifts as well. Because the video carrier is only amplitude modulated, it is extremely spectrally narrow ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0037 35 Hz wide) and thus these Doppler measurements require no demodulation. Helmboldt (2016) also showed that an arc of backscattering sources is often seen to the north within 55.25-MHz observations that is consistent with coherent backscatter from field-aligned irregularities within the urn:x-wiley:ess2:media:ess2506:ess2506-math-0038 region, presumably associated with urn:x-wiley:ess2:media:ess2506:ess2506-math-0039.

An imaging and analysis pipeline, geared toward studying meteor trails, was developed for 55.25-MHz TBN data and is described in detail by Helmboldt et al. (2014). In short, the pipeline low pass filters the TBN voltages to isolate a narrow, 195-Hz wide band. Every urn:x-wiley:ess2:media:ess2506:ess2506-math-0040 164 ms, the filtered voltages are Fourier transformed into 32 frequency channels. For each unique pair of antennas, or “baselines,” the voltages within these channels are cross correlated over an integration time of urn:x-wiley:ess2:media:ess2506:ess2506-math-0041 5 s. These cross correlations, or “visibilities,” are used to make an all-sky image cube of every urn:x-wiley:ess2:media:ess2506:ess2506-math-0042 5-s interval. Each image within this cube is a kind of “fisheye” view of the sky (orthographic projection) within which the station beam is nearly uniform everywhere. The image cube is collapsed into a single all-sky map of peak signal-to-noise ratio (S/N) by normalizing each channel by its measured urn:x-wiley:ess2:media:ess2506:ess2506-math-0043 noise and computing the maximum among all channels per all-sky pixel. Each S/N map is searched for significant (S/N  urn:x-wiley:ess2:media:ess2506:ess2506-math-0044 5) detections, the positions and peak intensities of which are cataloged. All of this is done separately for each linear polarization, namely, urn:x-wiley:ess2:media:ess2506:ess2506-math-0045 (north-south) and urn:x-wiley:ess2:media:ess2506:ess2506-math-0046 (east-west).

We note that while a Doppler shift is measured from each image cube for every detected source, these can be very heavily biased/inaccurate for relatively faint sources. This is because very bright sources (e.g., meteor trails and aircraft) will inevitably contaminate the Doppler profiles of much fainter sources via residual sidelobes and the wings of their Doppler profiles. In other words, the peak frequency channel for a relatively faint source may not represent its true Doppler shift. Instead, it will likely be associated with the part of its Doppler profile that is spectrally furthest from the nearest bright source while still appearing bright enough to detect. Additionally, the spectral width of the video carrier signal, 35 Hz, corresponds to a radial velocity resolution of nearly 100 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0047 at 55.25 MHz, which is larger than the typical wind-driven speed expected for urn:x-wiley:ess2:media:ess2506:ess2506-math-0048 region scatterers. For these reasons, the Doppler data for the candidate FAIs detected within the LWA1 image cubes were not used.

2.3 The 2014 Observing Campaign

Throughout 2014, an observing campaign was conducted with LWA1 in TBN mode at 55.25 MHz. The goal of this program was to use the distribution of meteor trail reflections on the sky to look for new meteor streams. It consisted of a series of 1-hr TBN collections, each separated by the previous one by 40 hr. This allowed enough time for each collection to be processed while covering the entire sky once every 5 days. Consequently, the collections were made within three time intervals, premidnight (03–04 UT), postmidnight (11–12 UT), and midday (19–20 UT).

As part of routine processing, an all-sky map of source counts was produced per polarization for each 1-hr collection. Within many of these maps, one or more arcs were apparent to the north that formed loci consistent with backscattering of the XEPM signal from FAIs at urn:x-wiley:ess2:media:ess2506:ess2506-math-0049 region heights. These were seen exclusively in the premidnight and postmidnight collections, in part due to increased interference from aircraft reflections during the midday collections. Close inspection of all source-count maps revealed the presence of these arcs during 12 premidnight collections and 6 postmidnight between May and September 2014. Data from the Digisondes in Boulder, Colorado, and Point Arguello, California, obtained through the digital ionogram database (DIDBase; Reinisch et al., 2004) confirmed the presence of urn:x-wiley:ess2:media:ess2506:ess2506-math-0050 in the region during or near the times of these collections. The derived urn:x-wiley:ess2:media:ess2506:ess2506-math-0051 properties from both Digisondes for all observing runs are summarized in Table 1. These imply that it is very likely that the observed arcs of sources correspond to FAIs associated with urn:x-wiley:ess2:media:ess2506:ess2506-math-0052.

Table 1. Digisonde Sporadic urn:x-wiley:ess2:media:ess2506:ess2506-math-0053 Parmeters
Boulder Point Arguello
foEs hEs foEs hEs
Date (MHz) (km) (MHz) (km)
2014-05-19 2.0 110 2.4 113
2014-06-03 3.0 140 5.3 115
2014-06-13 4.1 140 4.5 120
2014-06-18 4.0 140 3.2 109
2014-06-23 3.0 120 3.0 100
2014-06-28 2.0 125 5.6 106
2014-07-08 3.2 115 2.7 114
2014-07-18 2.4 115 3.6 113
2014-08-07 2.8 105 urn:x-wiley:ess2:media:ess2506:ess2506-math-0054 urn:x-wiley:ess2:media:ess2506:ess2506-math-0055
2014-08-17 urn:x-wiley:ess2:media:ess2506:ess2506-math-0056 urn:x-wiley:ess2:media:ess2506:ess2506-math-0057 2.3 94
2014-09-01 2.3 115 2.5 98
2014-09-06 2.3 105 2.5 113
2014-06-16 3.4 105 2.5 120
2014-07-01 4.4 110 2.3 116
2014-07-11 2.1 125 3.3 108
2014-08-05 1.9 115 2.5 125
2014-08-15 2.0 130 5.0 113
2014-08-20 2.1 115 2.4 116

To examine the nature of the sources found along these northern arcs, movies were made for each of these 1-hr collections using the peak S/N maps described in section 3 and Helmboldt et al. (2014). The movie for 13 June 2014 is included as supporting information (Movie S1), and similar movies for other collections can be made available upon request to the lead author. The movie shows the peak S/N per all-sky pixel separately for each polarization. White points are used to represent detected sources, which accumulate as the movie progresses to highlight the locations of the FAI arcs. A white dashed circle shows the approximate position of the horizon. One can see that the sources that appear along these arcs are rather distinct. They last longer than the fleeting meteor-trail reflections that occur all over the sky, but move much more slowly than the aircraft.

2.4 FAI Parameter Extraction

The 13 June movie qualitatively illustrates how one might use all-sky tracking of urn:x-wiley:ess2:media:ess2506:ess2506-math-0058 FAIs to probe the wind. In that case, the sources appearing along the FAI arcs generally drift westward with time, nominally indicating a westerly zonal wind. However, converting this apparent motion into quantified wind speeds requires several steps.

We should start with a theoretical background justifying our assumption that the motion of urn:x-wiley:ess2:media:ess2506:ess2506-math-0059 structures is wind dominated. Following Cosgrove et al. (2004), in the lower urn:x-wiley:ess2:media:ess2506:ess2506-math-0060 region, the electrons are magnetized, but the ions are collisional. Thus, the velocity vector for ion urn:x-wiley:ess2:media:ess2506:ess2506-math-0061 is given by
where urn:x-wiley:ess2:media:ess2506:ess2506-math-0063 ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0064 points eastward, urn:x-wiley:ess2:media:ess2506:ess2506-math-0065 northward, and urn:x-wiley:ess2:media:ess2506:ess2506-math-0066 vertical), urn:x-wiley:ess2:media:ess2506:ess2506-math-0067, urn:x-wiley:ess2:media:ess2506:ess2506-math-0068 is the ion-neutral collision rate, and urn:x-wiley:ess2:media:ess2506:ess2506-math-0069 is the cyclotron frequency. Generally, the neutral wind, urn:x-wiley:ess2:media:ess2506:ess2506-math-0070, is completely horizontal, such that urn:x-wiley:ess2:media:ess2506:ess2506-math-0071. Assuming that the electric field is predominantly wind driven, that is, urn:x-wiley:ess2:media:ess2506:ess2506-math-0072,
Here, urn:x-wiley:ess2:media:ess2506:ess2506-math-0075 was specified by the International Geomagnetic Reference Field for the appropriate dates. For an urn:x-wiley:ess2:media:ess2506:ess2506-math-0076 layer near LWA1 (geomagnetic latitude urn:x-wiley:ess2:media:ess2506:ess2506-math-0077 40°N) and an altitude of 110 km, urn:x-wiley:ess2:media:ess2506:ess2506-math-0078 (dominated by Fe urn:x-wiley:ess2:media:ess2506:ess2506-math-0079; Yokoyama et al., 2009) and urn:x-wiley:ess2:media:ess2506:ess2506-math-0080. In this case, the horizontal ion velocity only differs from the neutral wind by a few percent (3–4%), and the magnitude of the vertical motion is only 1–2% that of the horizontal speed. Thus, the measured horizontal motions of urn:x-wiley:ess2:media:ess2506:ess2506-math-0081 structures provide a reasonable estimate of the horizontal neutral wind, especially if the measurement uncertainties are on the order of a few percent. With higher precision measurements, a correction can be applied for non-wind-driven drift if reasonable estimates for urn:x-wiley:ess2:media:ess2506:ess2506-math-0082 and urn:x-wiley:ess2:media:ess2506:ess2506-math-0083 are known. We note that this assumes geomagnetically quite conditions. Significant deviations from this approximation have been observed during storm times (see, e.g., Chau & St.-Maurice, 2016). However, geomagnetic indices during the observations presented here showed no signs of increased activity (e.g., Kp  urn:x-wiley:ess2:media:ess2506:ess2506-math-0084  2.3 among all observing runs), and so this is not an issue for this particular study.

With this framework in hand, the apparent motion of FAIs on the sky can be used to probe winds near urn:x-wiley:ess2:media:ess2506:ess2506-math-0085 layers. The first step in this process is to isolate likely FAIs from the other types of sources within a particular collection. This was done by drawing a polygon region of interest around each arc visible within the combination of the source count images from both polarizations. This is illustrated in the top panel of Figure 1 for 13 June 2014 where the polygon boarders for three separate arcs (labeled Arc 1–3) are shown (in white). These separate arcs may correspond to distinct urn:x-wiley:ess2:media:ess2506:ess2506-math-0086 layers, likely at different heights.

Details are in the caption following the image
An example of the procedure for identifying groups of FAIs within LWA1 data. The image is an all-sky map of source counts (sum of both polarization) for a 1-hr observation (03–04 UT) on 13 June 2014. Polygons drawn by eye for three separate arcs of FAIs are shown in white. The remaining three panels show the FAI range (see section 5) as a function of time for sources within Arcs 1–3. Individual groups of FAIs identified by eye within these plots are highlighted with different colored polygons.
Next, the sources detected within all of the arcs are assigned three-dimensional (3-D) positions using their sky positions and the assumption that they are aligned along magnetic field lines. This was done by first computing the incident and scattered free-space wave vectors for each scatterer, urn:x-wiley:ess2:media:ess2506:ess2506-math-0087 and urn:x-wiley:ess2:media:ess2506:ess2506-math-0088, respectively. This was done for several assumed heights using the known locations of LWA1 and XEPM and the apparent sky position of the scatterer. Between these two, only urn:x-wiley:ess2:media:ess2506:ess2506-math-0089 depends on height; urn:x-wiley:ess2:media:ess2506:ess2506-math-0090 is taken directly from the all-sky map. Following Hysell and Chau (2001), the scattering vector is then simply

Optimum backscattering from an FAI occurs when urn:x-wiley:ess2:media:ess2506:ess2506-math-0092 with urn:x-wiley:ess2:media:ess2506:ess2506-math-0093 specified using the International Geomagnetic Reference Field. Linear interpolation was used to determine the height at which this condition was met for each scatterer. Combining this FAI height with the measured urn:x-wiley:ess2:media:ess2506:ess2506-math-0094 yields an approximate 3-D position.

These calculations were performed assuming a spherical Earth. Within this context, the height was calculated as urn:x-wiley:ess2:media:ess2506:ess2506-math-0095 where urn:x-wiley:ess2:media:ess2506:ess2506-math-0096 is the distance from the scatterer to the center of the Earth and urn:x-wiley:ess2:media:ess2506:ess2506-math-0097 is the radius of the Earth, assumed to be 6374.6 km near LWA1. The urn:x-wiley:ess2:media:ess2506:ess2506-math-0098 (east/west) and urn:x-wiley:ess2:media:ess2506:ess2506-math-0099 (north/south) positions were computed as arc lengths measured from the latitude and longitude of LWA1 (34.070° and −107.628°, respectively) at the height of the scatterer.

With urn:x-wiley:ess2:media:ess2506:ess2506-math-0101 positions determined, the sources within each polygon-isolated arc were further broken up into groups by plotting the FAI range ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0102) as a function of time. Within such a plot, sources appearing to move together can be identified and isolated by eye with another set of polygon boarders. This is illustrated in the lower panels of Figure 1 where the boundaries for groups identified within each arc are drawn with different colors. This practice also serves as a final step toward eliminated contamination by non-FAI sources (usually aircraft) and “stragglers,” those that do not appear associated with any comoving group. One should also note that within these and subsequent plots, the 3-D positions can appear somewhat quantized, or “striped.” This is due to the crude sky position determination per source within the Helmboldt et al. (2014) pipeline that simply adopts urn:x-wiley:ess2:media:ess2506:ess2506-math-0103 of the peak all-sky pixel rather than a more sophisticated centroid-finding method (e.g., center of mass and Gaussian fitting).

Figures 2-4 show the 3-D positions as a function of time for each of the 18, 1-hr collections where FAI arcs were found. For each collection, sources belonging to the same group are plotted with the same color; those not belonging to any group are plotted in black. One can see clear temporal gradients for many of the groups, while some are remarkably flat within one or more coordinates. These data are used to extract a horizontal velocity vector, urn:x-wiley:ess2:media:ess2506:ess2506-math-0104, using a linear fit per group. For each of these, the fit residuals are used to estimate urn:x-wiley:ess2:media:ess2506:ess2506-math-0105 errors.

Details are in the caption following the image
For the 18, 1-hr LWA1 observations where FAIs were present, the inferred east-west components of the positions (relative to LWA1) as functions of time. Each group of FAIs identified with the procedure illustrated in Figure 1 is shown with a different color. Black points represent those not assigned to any group.
Details are in the caption following the image
The same as Figure 2, but for the north-south components of the positions.
Details are in the caption following the image
The same as Figure 2, but for the vertical components of the positions.

One can also see temporal gradients within the plots of altitude versus time. At first glance, these may seem to contradict the assumption that vertical motions are negligible due to virtually nonexistent vertical winds and small urn:x-wiley:ess2:media:ess2506:ess2506-math-0106 vertical drifts (see equation 3). However, it seems more likely that this is a projection effect. For a urn:x-wiley:ess2:media:ess2506:ess2506-math-0107 structure, or “cloud,” not all FAIs within it will produce backscatter. LWA1 will only “see” those that lie on a surface that allows for Bragg scattering by FAIs, which locally resembles a plane. Thus, as a cloud moves horizontally, this plane intersects different parts of the cloud, which can give rise to an apparent vertical motion. In addition, it is also likely that our assumption of irregularities that are perfectly aligned with the magnetic field is not entirely accurate. Likewise, some scatterers may still be detected even when not viewed at the optimum aspect angle (i.e., when urn:x-wiley:ess2:media:ess2506:ess2506-math-0108). The limitations inherent to both these assumptions will lead to inaccuracies within the 3-D positions. In any case, the maximum and minimum altitude for each group gives a lower limit to the vertical extent of the cloud presumably associated with that group. In practice, this was done by measuring the difference between the upper and lower quartiles of the altitude distribution per group and is referred to here as urn:x-wiley:ess2:media:ess2506:ess2506-math-0109.

3 Results

The 2014 data collection and analysis described in section 2 yielded 75 district groups of FAIs, 60 within the premidnight time frame, and 15 postmidnight. For each of these, zonal and meridional winds, urn:x-wiley:ess2:media:ess2506:ess2506-math-0110 and urn:x-wiley:ess2:media:ess2506:ess2506-math-0111, and vertical extents, urn:x-wiley:ess2:media:ess2506:ess2506-math-0112 were measured. The winds are plotted as functions of altitude in Figures 5 and 6, separately for premidnight and postmidnight FAI groups. Within these plots, the horizontal error bars are urn:x-wiley:ess2:media:ess2506:ess2506-math-0113 errors and the vertical bars are urn:x-wiley:ess2:media:ess2506:ess2506-math-0114/2.

Details are in the caption following the image
Zonal wind speeds as a function of altitude measured for the FAI groups shown in Figures 2-4. As in Figures 2-4, these are divided into two categories, premidnight (03–04 UT; left) and postmidnight (11–12 UT; right). In both panels, predictions from the HWM14 model are given for each observation with horizontal error bars representing the range in predicted values for altitudes within urn:x-wiley:ess2:media:ess2506:ess2506-math-0115/2 of the measured altitude (see section 5). The ranges in vertical profiles from the TIDI mission for midlatitudes (25–45°) over the same time period are also shown.
Details are in the caption following the image
The same as Figure 5, but for meridional wind speeds.

We note that there are three observing runs (3 and 13 June and 15 August) where FAI groups were detected at altitudes urn:x-wiley:ess2:media:ess2506:ess2506-math-0116 130 km, which are higher than typical midlatitude urn:x-wiley:ess2:media:ess2506:ess2506-math-0117. However, as the Digisonde parameters in Table 1 indicate, there were also unusually high detections of urn:x-wiley:ess2:media:ess2506:ess2506-math-0118 over Boulder with hEs values between 130 and 140 km. Additionally, inspection of the individual ionograms from the Boulder Digisonde revealed the apparent formation of dense urn:x-wiley:ess2:media:ess2506:ess2506-math-0119 structures (reflection frequencies urn:x-wiley:ess2:media:ess2506:ess2506-math-0120 3–4 MHz) up to virtual heights of 150 km. Thus, while unusual, these high-altitude detections are actually consistent with concurrent radar data.

For comparison, we have plotted our measured wind values with those computed using the latest version of the empirical Horizontal Wind Model (HWM14; Drob et al., 2015). For these, the wind was computed for the date and time of each collection and at two altitudes, urn:x-wiley:ess2:media:ess2506:ess2506-math-0121 and urn:x-wiley:ess2:media:ess2506:ess2506-math-0122. Each model value is plotted in pink in Figures 5 and 6 with horizontal error bars indicating the difference in the wind between the two altitudes.

Also, for comparison, wind profiles were obtained from the TIMED Doppler Interferometer (TIDI) from the time period spanned by our data collections, specifically day of the year 139–249. Since there were few if any instances of overlap in time between our LWA1 collections and TIDI measurements within the same region, we have used these TIDI data to provide a statistical description of typical wind profiles below urn:x-wiley:ess2:media:ess2506:ess2506-math-0123 100-km altitude at midlatitudes both premidnight and postmidnight. In particular, we used all TIDI observations in the latitude range 25–45° and local times 20–24 for premidnight and 00–04 for postmidnight. These data were combined and included in Figures 5 and 6 as median profiles (red curves), upper and lower quartiles (blue curves), and 10th and 90th percentiles (green curves). The entire span is shaded in gray in each plot.

From these plots, it is apparent that there is relatively good agreement between the measured winds and those predicted by HWM14, especially between about 100- and 150-km altitude. There are several instances of unusually strong winds of 100 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0124 or more implied by the LWA1 data that are not captured by the empirical HWM14 model, which is to be expected. Notably, there are some rather large westward winds in the premidnight time frame, exceeding 200 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0125. Below 100 km, both the measured and HWM14-predicted winds are reasonably consistent with the midlatitude TIDI profiles, except for the premidnight zonal winds. In this case, the measured winds are systematically more westward for the LWA1 data than either the TIDI profiles or the HWM14 predictions. There is a similar offset among the few observations above urn:x-wiley:ess2:media:ess2506:ess2506-math-0126 150 km in the premidnight time frame where the meridional HWM14 and LWA1 winds are in general agreement, but the LWA1-determined zonal winds tend to be more westward.

These results are reinforced by the comparison of histograms for urn:x-wiley:ess2:media:ess2506:ess2506-math-0127 and urn:x-wiley:ess2:media:ess2506:ess2506-math-0128 between the LWA1 data and the HWM14 predictions shown in Figure 7. Given the nature of the HMW14 model, one would expect the histograms for its predictions to be more sharply peaked than observations since the model is incapable of predicting anomalously high winds in either direction. However, the locations of the histogram maxima should roughly agree. For the meridional winds and the postmidnight zonal winds, that is precisely what is seen within Figure 7. However, the premidnight zonal wind distribution peaks about 50 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0129 lower than the corresponding HWM14 histogram, at −25 versus 25 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0131.

Details are in the caption following the image
Histograms for (left column) zonal wind speed, (middle column) meridional wind speed, and (right column) vertical thickness. Separate histograms are shown for (to row) premidnight and (bottom row) postmidnight time periods (black) and for HWM14 model estimates (red).

Figure 7 also shows that the vertical extents of the groups of FAIs are urn:x-wiley:ess2:media:ess2506:ess2506-math-0132 7 km and are as large as nearly 40 km. Note that urn:x-wiley:ess2:media:ess2506:ess2506-math-0133 is essentially a full width at half maximum and likely a lower limit at that, implying the full extent is typically urn:x-wiley:ess2:media:ess2506:ess2506-math-0134 20 km. This is reasonably consistent with the properties of dense urn:x-wiley:ess2:media:ess2506:ess2506-math-0135 clouds observed by several other authors, often in association with so-called quasiperiodic echoes (e.g., Bernhardt, 2002; Bowman, 1989; Pan & Tsunoda, 1998; Hysell et al., 2012; Hysell & Burcham, 1999; Larsen, 2000). In fact, using a different set of LWA1 observations in combination with data from the Boulder, Colorado, Digisonde, Helmboldt (2016) found that the northern arc of sources associated with urn:x-wiley:ess2:media:ess2506:ess2506-math-0136 FAIs appeared most prominently when urn:x-wiley:ess2:media:ess2506:ess2506-math-0137 layers with large peak densities (foEs urn:x-wiley:ess2:media:ess2506:ess2506-math-0138 4 MHz) were present.

It should also be noted that as Figures 2 and 3 show, spatially separate groups of FAIs observed at the same time were generally seen to move in the same direction with similar speeds. The exceptions to this are the blue color-coded groups from 23 June and 8 and 11 July. For 23 June and 11 July, the zonal drift is eastward while the other groups' drifts are westerly. The opposite is true for 8 July. In all three cases, they are at roughly the same altitudes as other groups ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0139 100–110 km). They are all about 100–200 km east of the other observed groups and for the 8 and 11 July cases, they are also urn:x-wiley:ess2:media:ess2506:ess2506-math-0140 50 km further south. These indicate the occasional presence of horizontal zonal wind shears, both positive and negative, with magnitudes urn:x-wiley:ess2:media:ess2506:ess2506-math-0141 0.60–1.5 m urn:x-wiley:ess2:media:ess2506:ess2506-math-0142 s urn:x-wiley:ess2:media:ess2506:ess2506-math-0143  km urn:x-wiley:ess2:media:ess2506:ess2506-math-0144. Note that this is roughly an order of magnitude smaller than the vertical zonal wind shears typically associated with urn:x-wiley:ess2:media:ess2506:ess2506-math-0145 (see, e.g., Yokoyama et al., 2009).

4 Discussion

The results presented here clearly demonstrate both the novelty and potential utility of using passive all-sky tracking of FAIs within urn:x-wiley:ess2:media:ess2506:ess2506-math-0146 layers as a tracer of urn:x-wiley:ess2:media:ess2506:ess2506-math-0147 region winds. While this can also be done optically, this requires a high-power active system to excite airglow from within the urn:x-wiley:ess2:media:ess2506:ess2506-math-0148 structures with radio waves, which is only observable at night (e.g., Bernhardt et al., 2003). The heights at which urn:x-wiley:ess2:media:ess2506:ess2506-math-0149 FAIs are found tend to be urn:x-wiley:ess2:media:ess2506:ess2506-math-0150 100–120 km but can reach as high as urn:x-wiley:ess2:media:ess2506:ess2506-math-0151 150 km. Thus, while monitoring of urn:x-wiley:ess2:media:ess2506:ess2506-math-0152 FAIs cannot deliver instantaneous wind profiles like satellite-based remote sensing and/or meteor radars can, it can provide measurements of wind vectors at heights that are inaccessible with these other methods. The results of our LWA1 campaign show that this is especially useful in three areas.

The first of these is the detection and characterization of horizontal variations in the wind. As noted in section 7, out of the eighteen 1-hr collections presented, three (20  urn:x-wiley:ess2:media:ess2506:ess2506-math-0153  10%) had instances of horizontal wind shears with amplitudes on the order of 1 m urn:x-wiley:ess2:media:ess2506:ess2506-math-0154 s urn:x-wiley:ess2:media:ess2506:ess2506-math-0155 km urn:x-wiley:ess2:media:ess2506:ess2506-math-0156. The effect of such wind shears would be averaged out within methods that require the assumption that the wind is fixed over a relatively large area (e.g., all-sky meteor radars) and would be missed by radar-based methods that monitor a relatively limited area. The all-sky mapping enabled by LWA1's transient buffer modes allows for the best of both worlds by essentially monitoring a large area with many narrow beams. We note that relatively new multi-static, multifrequency meteor radar techniques also enable wind profiles to be measure without the assumption of large scale uniformity (Stober & Chau, 2015). The method presented here is quite complementary to these since it can do the same, but at potentially higher altitudes.

The second area of significant impact is the measurement of localized and relatively large winds that can only be captured by tracking an isolated tracer over a short period of time. The effect of these localized anomalies will be greatly diminished within profiles from traditional all-sky meteor radars that are generated by fitting a single vertical profile to the range/Doppler data from meteors over the entire sky. Our study has found several such instances of high winds, especially in the premidnight collection periods. These are mostly consistent with results from chemical release studies of winds within a similar altitude region compiled by Larsen (2002). These results indicated that the winds between 80 and 140 km were typically around urn:x-wiley:ess2:media:ess2506:ess2506-math-0157 50 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0158, but with large deviations up to urn:x-wiley:ess2:media:ess2506:ess2506-math-0159 150–200 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0160. These large deviations were usually found within urn:x-wiley:ess2:media:ess2506:ess2506-math-0161 100- to 120-km altitude, similar to what was found here (see Figures 5 and 6).

The third area involves regional/local offsets from climatological winds. Apart from the anomalously high winds discussed above, the results presented here generally agree quite well with the HWM14 model, especially in the postmidnight time frame. However, as noted in the previous section, the premidnight zonal winds are a notable exception. While the LWA1 and HWM14 zonal winds match fairly well near 110-km altitude where they are near 0, there is a systematic offset below and above this altitude. The LWA1-measured winds are significantly more westward, even in relation to the median TIDI winds below 100 km. This may suggest a relatively isolated regional/local effect that is not sufficiently captured by either the HWM14 model or the midlatitude TIDI winds (recall that all longitudes were combined for the comparison TIDI profiles; see section 7).

Given the location of LWA1, a likely culprit is modification of the impact of the semidiurnal tide via gravity waves. The western/southwestern United States is quite mountainous, and orographic gravity waves are a common occurrence. Several authors have noted that such waves can potentially significantly alter the morphology and/or vertical location of tidally driven features within the lower thermosphere wind profile, and vice versa (see, e.g., England et al., 2006; Fritts & Alexander, 2003 and references therein).

However, the LWA1 data were acquired primarily in summer months when tropospheric/stratospheric wind profiles often make vertical propagation of mountain waves beyond the troposphere difficult. Therefore, to assess the plausibility of the mountain wave hypothesis, lower atmospheric wind profile maps were obtained from the National Oceanic and Atmospheric Administration North American Regional Reanalysis for the collection dates. Mean wind profile maps were computed separately for the premidnight and postmidnight collections. To facilitate a more direct comparison, only the six premidnight collections that were closest in time to the six postmidnight collections were used. The mean surface winds are displayed in the left panels of Figure 8. While relatively strong, easterly winds ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0162 6–10 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0163) are present over the mountainous west/southwest in both maps, these are much more widespread in the premidnight map, covering the entire Rocky Mountain range, as well as all of Utah and Arizona.

Details are in the caption following the image
Mean zonal surface winds from NARR for (top left) premidnight and (bottom left) postmidnight data collections. The corresponding stratospheric Scorer parameters are shown in the panels in the right column. For a more direct comparison, only the six premidnight collections times (out of 12 total) that were closest in time to the six postmidnight collections were used.

While this indicates the generation of lee waves was plausible, their escape from the troposphere is far from guaranteed. To assess this, the Scorer parameter, urn:x-wiley:ess2:media:ess2506:ess2506-math-0164, was computed at the 300-hPa pressure level ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0165 buoyancy frequency, urn:x-wiley:ess2:media:ess2506:ess2506-math-0166 horizontal wind). For a mountain wave to survive ascent, the Scorer parameter must remain urn:x-wiley:ess2:media:ess2506:ess2506-math-0167 ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0168 horizontal wavenumber). One can see that the Scorer parameter in the stratosphere was quite low over most of the mountains of the western United States as expected for summertime. However, there was a region over Arizona/New Mexico where the Scorer parameter was significantly larger during the premidnight collections only. This, coupled with relatively strong winds blowing directly toward New Mexico, likely led to increased mountain wave activity in the middle-to-upper atmosphere during the pre-midnight time frame. This could be the cause of not only the offset from both the HWM14 and median TIDI profiles but also the larger amount of scatter among the winds measured premidnight.

In summary, the novel method demonstrated here and enabled by the all-sky imaging capability of LWA1 provides a unique means for exploring lower thermospheric winds. While it cannot yield full, instantaneous wind profiles, it complements other methods by providing a probe of both large, localized winds and regional, horizontal variations/shears. As an example, the summer 2014 study presented here shows several localized instances of wind magnitudes in excess of 100 m s urn:x-wiley:ess2:media:ess2506:ess2506-math-0169, and a systematic westerly offset in the premidnight time frame that appears to be unique to the southwestern United States, possibly due to increased mountain wave activity. This study also found evidence for occasional ( urn:x-wiley:ess2:media:ess2506:ess2506-math-0170 10–30% of collections) horizontal wind shears at 100- to 110-km altitude with magnitudes on the order of 1 m urn:x-wiley:ess2:media:ess2506:ess2506-math-0171 s urn:x-wiley:ess2:media:ess2506:ess2506-math-0172 km urn:x-wiley:ess2:media:ess2506:ess2506-math-0173.


LWA1 data are proprietary but can be made available upon request to the lead author. Digisonde data and ionograms were obtained through the Global Ionosphere Radio Observatory at http://giro.uml.edu website. NOAA NARR wind data were obtained from ftp://ftp.cdc.noaa.gov website. Geomagnetic indices were obtained through the pyglow python package (currently available at https://github.com/timduly4/pyglow). The authors would like to thank S. Cockrell and S. Bello for assistance in LWA1 data processing. Basic research at the Naval Research Laboratory is supported by 6.1 base funding. Construction of the LWA has been supported by the Office of Naval Research under Contract N00014-07-C-0147. Support for operations and continuing development of the LWA1 is provided by the National Science Foundation under Grants AST-1139963 and AST-1139974 of the University Radio Observatory program.