Simultaneous radio interferometer and optical observations of ionospheric structure at the Very Large Array



[1] Radio astronomers are searching the cosmos for new scientific discoveries at increasingly lower radio frequencies and with larger antenna arrays, but their observations of the sky are blurred by the dynamic ionosphere. At the same time, ionospheric scientists are seeking to understand, at increasingly higher spatial and temporal resolutions, the dynamics that drive the ionosphere and its effects on technological systems. Advancements in radio astronomy at the Very Large Array (VLA) are leading to advancements in ionospheric physics and vice versa. We review some of the ionospheric observations made by the VLA at low frequency. Results from a 2003 summer campaign at the VLA are discussed, during which an all-sky optical camera was used to monitor ionospheric structure during VLA 74-MHz operations. The camera and additional off-site sensors, including ionosondes and incoherent scatter radar, were used to identify the dominant, summer nighttime ionospheric phenomena contributing to VLA signal distortion. Knowledge of the specific phenomena, including their spatial and temporal characteristics, can be used to improve low-frequency, astronomical imaging. Similarly, the VLA observations can be used to investigate ionospheric phenomena in great detail, leading to an improved understanding of ionospheric physics. Key to these findings is the identification of specific ionospheric phenomena using support sensors. Implications for the development of the Long Wavelength Array are discussed.

1. Introduction

[2] Radio astronomy measurements can be strongly impacted by atmospheric effects that, below ∼1400 MHz, are dominated by the Earth's ionosphere. Variations in total electron content (TEC) over different antennas commonly disrupt the phase coherence of large radio interferometers like the Very Large Array (VLA), located near Socorro, New Mexico, which has a maximum antenna separation ∼35 km. Consequently, the data must be corrected [Kassim et al., 2007] to recover the cosmic source visibility (amplitude and phase) to permit high angular resolution imaging and accurate radio flux density measurements of cosmic background emitters. At the same time, these corrections can provide a powerful probe of the structure of the regional ionosphere between the interferometer and the distant radio sources. The next generation large, low-frequency radio interferometers (e.g., the LWA: and LOFAR: will be much larger (>100 km) and more sensitive (e.g., MWA: than the VLA and will provide exciting opportunities for ionospheric research on regional scales.

[3] Horizontal gradients in the ionosphere introduce differential phase offsets between the antennas of a radio interferometer. Radio interferometers can be incredibly sensitive to these gradients in total electron content (TEC): a TEC difference of 0.01 total electron content units (TECU, 1014 electrons/m2) between two antennas is equivalent to a phase offset of 66° at 74 MHz. For accurate imaging, phase differences between antennas of the VLA should be ∼1° or less, so that TEC variations of ∼0.0001 TECU are detectable. If the differential phases are not completely removed, the phase variations lead to refractive distortions or refractive wander of the apparent positions of sources in the VLA images. For the 74-MHz VLA, large-scale ionosphere-induced refraction is on the order of an arc minute at night and a few arc minutes shortly after sunrise (N. Kassim, Opening a new window on the electromagnetic spectrum: The Low Frequency Array (LOFAR), paper presented at RF Ionospheric Interactions Workshop, National Science Foundation, Santa Fe, N.M., 28 April to 1 May, 2002). At 1-min intervals, refractive wander shows considerable variability. Refractive wander leads to an apparent loss in brightness of the source due to smearing over a larger region of the image. As an illustration, measurements of the bright radio source Virgo A have shown a decrease in apparent flux density by a factor of 3, from 300 Jy (1 Jansky (Jy) = 10−26 W m−2 Hz−1) to 100 Jy, owing to increased refractive wander from nighttime to a couple of hours after sunrise. Short-term (1 h or less) variability in the apparent flux density on the order of 50 Jy has been attributed to ionospheric structure passing through the line of sight. This refractive wander is not identical between radio sources owing to the presence of small-scale ionospheric structures; therefore, differential wander is an important consideration when imaging multiple sources. Differential wander leads to variations in position and peak intensity of the sources and it can also affect the sidelobe structure of surrounding sources [Kassim et al., 2007].

[4] Kassim et al. [2007] describe two methods that have been developed to model and remove differential ionospheric phases. The more commonly used method is a data-adaptive technique known as self-calibration. All antennas in the array are pointed to a sufficiently strong and isolated source, so that a unique radio signal can be observed. The measured relative phase between antennas consists of the intrinsic structure of the source, differential ionosphere between the antennas, and instrumental noise. For sufficiently low frequency, the troposphere is ignored. A least squares solution is used to estimate the differential ionospheric phases, which minimizes the differences between a model of the source and the measurements [Cornwell and Fomalont, 1999]. For an array consisting of N antennas, there are N(N-1)/2 measurements between antennas and N-1 unknown differential ionospheric phases relative to a common antenna. For N > 3, the number of measurements exceeds the number of unknowns; for the VLA (N = 27), the problem is extremely overdetermined. Consequently, the N-1 unknown differential ionospheric phases can be solved to high precision and are a very precise measure of ionospheric structure passing between the antennas.

[5] Structure in the ionosphere is normally attributed to the dominant ionospheric layer called the F region or F layer (altitude 150–1000 km). This layer is produced during the daytime primarily by solar extreme-ultraviolet photoionization. Both the ions and electrons in the F region plasma move subject to the Earth's magnetic field, so the plasma can exhibit field-aligned structures. Interactions between F region plasma and the Earth's magnetic field and upper atmosphere can create a variety of structures on a variety of size scales including the equatorial anomaly, storm-enhanced density, traveling ionospheric disturbances (TID), and plasma bubbles [e.g., Kelley, 1989; Foster and Rideout, 2007]. R. Perley and G. S. Bust (Probing the ionosphere with the Very Large Array, paper presented at XXVIIth General Assembly, Int. Union of Radio Sci., Maastricht, Netherlands, 17–24 August 2002) identified ionospheric structures on several different size scales present in 74-MHz VLA measurements of radio source Virgo A. The main features included: refractive “wedges” (linear TEC gradients) at midnight and dawn, medium-scale traveling ionospheric disturbances, small-scale (∼10 km) structures, and a period of ionospheric quiescence before dawn; all present in one continuous 9-h-long observation. Previous ionospheric studies using the VLA and other radio interferometers have indicated the presence of traveling ionospheric disturbances (TIDs) of various scales and smaller-scale structures due to the plasmasphere [Mercier, 1986; Jacobson and Erickson, 1992a, 1992b, 1993; Kassim et al., 2007].

[6] Additional ionospheric structures are present in the E layer (altitude 90–150 km) and the valley between the E and F layers. The E layer is a secondary ionospheric layer – total electron content is nominally an order of magnitude or two lower than the F layer. Here solar EUV and soft X rays drive photoionization, and most of the layer disappears after sunset owing to the fast recombination rates at these altitudes. However, vertically thin (1–3 km) but dense (a few 1011 electrons/m3) sporadic E clouds occur at E layer altitudes and often persist throughout the night [Mathews, 1998]. Intermediate layers or tidal ion layers, composed primarily of metallic ions from meteor ablation, have been observed to descend from F layer altitudes down into the E layer. These lower layers are more susceptible to structuring from gravity waves in the upper atmosphere than the F layer. However, they are also electromagnetically coupled to the F layer, so structure in the lower layer has the potential to map along magnetic field lines to the upper layer, and vice versa [Mathews et al., 2001].

[7] Jacobson and Erickson [1993] observed electron density structures in the plasmasphere using 333-MHz and 307.5-MHz VLA data collections. Originally interpreted as field-aligned plasma structures frozen onto the Earth's magnetic field lines, these plasmaspheric structures produced small temporal scale oscillations (1–3 min) in VLA data with magnitudes of ∼0.01 TECU (∼1014 electrons/m2) across a 20-km baseline and were observed primarily along southward lines of sight parallel to the Earth's magnetic field. The plasma had an apparent magnetically eastward velocity due to Earth's corotation of the trapped plasma. The plasma was inferred to reside between 2000- and 10,000-km altitude on magnetic L shells of L = 2–3. Hoogeveen and Jacobson [1997] subsequently demonstrated that the plasma was not actually “frozen on” the field lines but was often convected westward or eastward across the field lines. They argued that the restricted range of L shells observed was a limitation of the instrumentation and observations rather than the physics of the plasmasphere.

[8] A key question, both from the standpoint of development of radio astronomical calibration techniques as well as for ionospheric physics, is establishing the connection between the ionospheric structure detected by astronomical interferometers and the structures typically detected in ionospheric observations. This work focuses on the identification of nighttime ionospheric structures affecting low-frequency radio interferometer operations. The Very Large Array (VLA) radio interferometer (centered at 34°04′43.497″N, 107°37′05.819″W) was used for this study. Self-calibration was used to estimate the differential ionospheric phase between elements of the array and the central antenna element. The goal was to identify and characterize ionospheric structures affecting VLA measurements and to eventually develop mitigation algorithms for correcting the VLA data. For this study, an all-sky optical camera [Makela et al., 2004] was used to monitor F region ionospheric structure simultaneously during VLA 74-MHz operations. Additional off-site sensors, including ionosondes and incoherent scatter radar, were used in this study to identify the summer nighttime ionospheric phenomena contributing to VLA signal distortion.

2. Simultaneous Radio and Optical Observations

[9] In August 2003, simultaneous observations were made using the VLA and an optical camera. Program AC677 consisted of three 8-h data collects at 74 MHz in the A configuration. An all-sky camera, on loan from Cornell University, was deployed at the VLA in order to collect simultaneous ionospheric and mesospheric information. The camera employed four filters: 630.0 nm, sensitive to F region electron density at low F region heights; 777.4 nm, sensitive to F region electron density; 557.7 nm, sensitive to mesospheric neutral density; and, oxygen hydroxyl (∼20 nm broadband filter, centered on 865 nm) sensitive to mesospheric neutral density. The objective of the simultaneous observations was to identify nighttime ionospheric structures affecting low-frequency VLA operations.

[10] Figure 1 shows 7 h of data collected from the VLA while tracking Cygnus A on 25 August 2003. Between 0500 and 0600 UT the VLA was not tracking Cygnus A. The differential total electron content between each antenna and the center of the array is shown for each arm of the VLA. In the early evening (0200–0500 UT) all the north arm data are positive, indicating a positive gradient along the north arm of the array. This positive gradient is modulated by 10–20 min waves. The data in this period show a negative gradient along the east arm and essentially no gradient along the west arm. This gradient (or evening wedge) is relatively constant (ignoring the superimposed 10–20 min waves) and appears to run from the northwest to southeast. The optical camera confirms the presence of this gradient over a much larger spatial dimension. Figure 2 shows the intensity of the 630.0-nm camera data along two widely separated points in the sky: one 200 km north of the array and the other 200 km east of the array. The camera data begin shortly after sunset and show that the northwest-to-southeast gradient remains in place for a couple of hours before disappearing near 0600 UT. One possible explanation for this stable gradient is the orientation of the dusk terminator. The dusk terminator is aligned southwest to northeast for the VLA latitude during late August. As the terminator passes over the area solar production of ionospheric plasma ceases in the southeast before it ceases in the northwest, establishing the observed gradient. Once the terminator has passed, the ionosphere simply decays in place, maintaining the same gradient orientation until some other process overtakes it.

Figure 1.

Differential ionosphere (TECU) between each antenna and the center antenna for the VLA tracking Cygnus A on 25 August 2003: (a) north arm, (b) east arm, and (c) west arm.

Figure 2.

Intensity of the 630.0-nm camera data along a 200-km northward and a 200-km eastward line of sight from the VLA.

[11] The camera data in Figure 2 also show that the ionosphere at the bottomside of the F region (∼250-km altitude) decays sharply after sunset to a minimum level near local midnight (∼0700 UT) and then increases slightly near 0800 UT. Since production does not occur at night, this increase near 0800 UT suggests that plasma was transported to lower F region altitudes during this time. Near midnight the thermospheric winds concentrate neutrals at low latitudes creating the Midnight Temperature Maximum [Faivre et al., 2006]. This causes the meridional winds to reverse and flow poleward rather than equatorward, which in turn pushes plasma down the magnetic field lines to lower altitudes. Coincident with the increase in plasma at low F region altitudes observed by the camera, a weak large-scale structure was observed faintly in the camera field of view aligned northwest to southeast and propagating northeastward. The effects of this F region wave are observed in the VLA data in Figure 1 from 0700 to 0900 UT. The west arm sees the largest component of the wave, first positive and then negative gradients indicating the passage of the wave over the VLA field of view. The north arm shows a similar effect, although reversed in sign; and the east arm shows minimal effect. This is consistent with the optical data showing a northwest to southeast wave propagating northeastward. The details of this large-scale wave are masked to some degree by the continued presence of 10–20 min and even smaller waves.

[12] In order to examine these small-scale waves, Figure 3 shows an enlargement of the west arm data from Figure 1. A wide range of wave periods are evident, ranging from 1 to 20 min with magnitudes in the 0.002–0.1 TECU range over the length of the west arm (∼20 km). What are the sources for these structures? Jacobson and Erickson [1993] argued that structures with periods less than ∼6 min are not from the F layer because gravity wave periods smaller than this do not reach F layer heights. This has been confirmed by the recent modeling work of Vadas [2007]. However, short-period gravity waves can reach E layer altitudes where they affect the formation of sporadic E clouds [Mathews, 1998].

Figure 3.

Enlargement of west arm differential ionosphere (TECU) between each antenna and the center antenna of the VLA tracking Cygnus A on 25 August 2003.

[13] To investigate whether sporadic E clouds could be detected by the VLA, samples of sporadic E clouds from high vertical resolution incoherent scatter radar data were obtained from Arecibo ( The VLA analogous 20-km baseline differential TEC was estimated from radar data using a 50 m/s horizontal velocity for the sporadic E cloud. Figure 4 shows that background levels of sporadic E (<2 MHz) produce a variety of temporal-scale waves over a 20-km baseline with magnitudes in the 0.001–0.05 TECU range. When sporadic E rises above background levels to 3 MHz, the magnitude of the wave increases to ∼0.02 TECU. Although not shown here, another radar sample showed that when sporadic E rose to 5 MHz, the magnitude of the wave increased to ∼0.08 TECU. These results indicate that sporadic E can indeed be detected by the VLA and may account for much of the structure observed during the campaign.

Figure 4.

Sample incoherent scatter radar data from summer night over Arecibo in Atlantic Standard Time (AST) showing (a) peak electron density of the sporadic E layer in units of 105/cm3, (b) vertical TEC of the layer, and (c) estimated differential TEC over a 20-km baseline antenna baseline assuming a sporadic E horizontal velocity of 50 m/s.

[14] While direct observation of sporadic E over the VLA was not available during the campaign, climatology indicates that sporadic E occurs most frequently in the summer months and covers large geographic regions [Smith, 1957]. Additionally, ionosondes in west Texas and California at similar latitudes to the VLA showed persistent presence of sporadic E during the night of 25 August 2003. During the 8-h VLA data collection, both ionosondes observed fairly persistent sporadic E in the 2- to 5-MHz range with a median value around 2.5 MHz. This is a good indicator that sporadic E was likely present in the VLA field of view and could contribute significantly to the structure observed by the VLA.

[15] The optical camera provides further evidence of significant gravity wave activity during the observational period. Figure 5 shows mesospheric neutral density waves as observed on the all-sky camera using the 557.7-nm filter. These waves are the result of gravity waves at 85- to 95-km altitude. Note the complexity of the gravity waves. At 0609 UT the camera shows a pattern of narrow waves (∼30 km) elongated in the southwest-to-northeast direction. Two hours later the wave pattern shows more complexity and even smaller wave scales (∼5 km). It is likely that these complex gravity waves reach E layer altitudes, where they would contribute to the formation of sporadic E clouds.

Figure 5.

Mesospheric neutral density waves (85- to 95-km altitude) observed at 557.7 nm by the all-sky camera and projected to geographic coordinates for (a) 0609 UT on 25 August 2003 and (b) 0809 UT on 25 August 2003.

3. Implications

[16] The results of simultaneous radio and optical observations indicate that both F layer and E layer structures contribute to the ionospheric variations observed on the VLA at 74 MHz. The VLA and the all-sky camera observed an F layer evening wedge (or gradient) after sunset and a propagating F layer structure associated with the midnight temperature maximum after local midnight. On top of these large- and medium-scale features were small-scale waves (periods 1–20 min, magnitudes 0.002–0.1 TECU). While 15- to 20-min waves in the VLA data, under magnetically quiet conditions, are generally attributed to gravity waves in the F layer, observation and simulation of E layer structure indicates that sporadic E clouds contribute significantly to the small-scale structure observed by the VLA. Moderately intense sporadic E clouds (3–5 MHz) produce waves with temporal scales and magnitudes (0.02–0.08 TECU) similar to those observed in the F layer. Moreover, background levels of sporadic E (∼2 MHz) produce waves with a variety of small temporal scales and magnitudes (0.002–0.01 TECU), which may account for much of the smaller-scale features (1–10 min) observed on the VLA.

[17] The high sensitivity and spatial resolution of the VLA provide unique opportunities to investigate large and small-scale ionospheric structure with a precision an order of magnitude better than GPS. This opens up the possibility for the effect of gravity waves on the E layer and F layer to be studied in great detail. Additionally, the VLA can be divided into sub arrays and monitor multiple celestial radio sources making it possible to map ionospheric structure across the sky. Future arrays with many more antennae over much larger distances, like the Long Wavelength Array, promise to provide enough geometric diversity to not only provide altitude information about the ionosphere but to image the ionosphere.

[18] For low-frequency radio astronomy, ionospheric structure is a significant source of error. Knowledge of the particular ionospheric structure or structures affecting the array can be used to aid radio astronomy ionospheric mitigation techniques. This can be accomplished in three fundamental ways: through climatology, morphology, and weather. The climatology of ionospheric structure allows the scheduling of low-frequency radio astronomy to minimize the impact of the ionosphere. As an example, gravity wave activity is much more likely during summer than winter. As a result, ionospheric structure is typically less during winter months than summer months. Time of day is also important as the ionosphere is more dense during the day than at night. Figure 6 gives an example showing how summer nights are significantly more disturbed by the ionosphere than winter nights. In this example, the summer night shows significant structure (>0.05 TECU or 360° at 74 MHz over 20-km baselines) with periods ranging from 5 min to an hour or two. The winter night shows structure but the magnitude is significantly less (<0.05 TECU). Another significant climatological factor is the 11-year solar cycle. The ionosphere is more dense, and magnetic storms are more prevalent during maximum years of the cycle.

Figure 6.

Comparison of ionospheric effects on two VLA 74-MHz data collections from (a) summer night of 25 August 2003 tracking Cygnus A and (b) winter night of 19 January 2001 tracking Virgo A.

[19] The morphology of ionospheric structure–how it moves and evolves–can also be used to develop improved ionospheric mitigation algorithms. These algorithms would take advantage of the physical characteristics of ionospheric structure: size, shape, direction of motion, and altitude.

[20] Finally, monitoring the ionospheric weather can also aid radio astronomy by allowing operators to dynamically adjust operations to minimize impacts or to take advantage of particularly quiet conditions. While some of this capability can be provided directly by monitoring the radio astronomy signals, independent ionospheric sensors are needed to identify ionospheric phenomena and predict their behavior. A Digisonde or a bistatic HF radar could provide identification of specific ionospheric structures effecting radio astronomy signals. Real-time knowledge of ionospheric weather features and short-term forecasting of those features would allow operators to take advantage of opportune ionospheric weather conditions.


[21] The authors wish to thank Cornell University for the use of their all-sky camera.