PSFR and ISR scans for the six selected observation days are presented in Plates 1 through 6. At the top of each Plate the PSFR data is displayed as a grayscale image. The gray level corresponds to the intensity of the received signal with darker pixels representing stronger signals and a roughly 35 dB difference between black and white pixels. The horizontal axis is universal time (UT) and the vertical axis is frequency. Midnight magnetic local time corresponds to 0217 UT at ground level. Both the time and frequency scales vary between plates. Horizontal dark lines in the grayscale are fixed-frequency anthropogenic transmissions such as short-wave radio, AM band, and navigational transmissions.
Vertical black lines superposed on the grayscale plots indicate the start and stop times of ISR scans. The electron density profiles measured during these scans are shown below the grayscale plots as false color plots. Electron densities for which the estimated error is less than 50% are shown, but the uncertainty, which depends primarily on the density, is typically less than 10% at F region altitudes. Letters (A, B, C, etc.) indicate which color density plot corresponds to which period in the grayscale plot. All electron density color plots have been rendered on grids with identical scales: the vertical axis is altitude 75-500 km, the horizontal axis is ground range (magnetic north) from the facility, -500 - 500 km, and the color bar displays the logarithm of the electron density from 1.010 cm to 3.210 cm. In Plate 4, all the ISR scans are elevation scans (through the zenith) which are rendered on meridional grids. In all other plates (1, 2, 3, 5, 6), composite (mixed elevation/azimuth) scans were conducted, in which alternating scan planes are offset to the east and west in a manner described by Weber et al. . In these cases, each composite scan has been geometrically projected onto the meridional plane. A small curved arrow near the origin of the contour plot indicates the direction of the scan. The vertical geometric projection of the azimuthal angle of the radar beam is shown in the lower left corner of each panel with the progression of the scan indicated by colors ranging from green (start of scan) to red (end of scan).
Plate 1 shows an example of both 2 and 3 auroral roar events observed on August 3, 1995. Observations of 3 roar, first reported by Weatherwax et al., , are much less common than 2roar. Segments B and C of the grayscale show the 3 roar event starting at 0110 UT near 4 MHz. In the middle of segment C a radio emission can be seen at a slightly lower frequency, but this probably represents a different type of emission called an MF Burst [Weatherwax et al., 1994; LaBelle et al., 1997]. Segments D and E show a related, but uncorrelated, 2 roar emission near 2.8 MHz starting around 0125 UT and present during nearly all of segment E. For this experiment, the ISR performed composite scans (moving in both azimuth and elevation) such that planes parallel to the magnetic meridian, but offset by 20 to the east and again to the west, are probed. The scan rate varied, by design, to give constant ground tracking. The processed data integration of 15 s results in a horizontal spatial resolution at 275 km altitude of 23 km.
The corresponding ISR scans show the F region ionosphere (H = 270 km, N = 1.9610 cm) during these two events to be mostly laminar without significant electron density structure ( / 200 km). Discrete, weak E layer arcs, signified by patches of enhanced plasma density between 100 and 200 km due to increased ionization from energetic auroral electrons in the upward current region are observed throughout this period. An E region arc appears at the equatorward edge of scan C, but the F region ionosphere north of the arc is notably smooth. No significant horizontal gradients are observed, and there are no auroral ionospheric cavities, discussed by Doe et al., , observed during this event (such cavities would be visible as a region of low plasma density extending through the F region due to evacuation in the downward current region).
Plate 2 shows a particularly intense 2 auroral roar event observed on August 15, 1995, beginning around 0222 UT, toward the end of scan C. The emission is particularly strong during radar scan D and trails off during the following scan. After an interlude of nearly 28 min with no roar observations, another strong emission is observed starting at 0301 UT (segment F). The ISR operating mode was identical to that of the first experiment, and the resulting resolution is the same as that of Plate 1.
The series of ISR scans for this event indicate that the field of view is at the edge of the polar cap boundary with an E region boundary arc, an auroral ionospheric cavity, and a structured F region ionosphere. Significant horizontal ionization gradients are observed in the polar cap north of the arc. For example, the density of the F region at H = 270 km varies from = 1.0610 cm ( = 2.92 MHz) in the middle of an ionization patch to = 4.3210 cm ( = 1.87 MHz) in the middle of a cavity over a scale length of approximately 60 km. The ionospheric cavity begins to form slightly equatorward of the radar in scan C and is apparent in scans D and E. The composite scans provide an estimate of the zonal extent of the ionospheric cavity (~250 km) and its duration (~20 min). Scan F, corresponding to the second strong roar, also suggests significant horizontal F region gradients.
Plate 3 shows an example of 2 roar observed on November 1, 1995, beginning near 1927 UT and continuing intermittently for nearly 30 min. During this time the center frequency of the roar rises from ~2.65 MHz to ~2.75 MHz. The ISR operating mode was identical to that of the first two experiments, and the resulting resolution is the same as that of Plates 1 and 2.
This period is characterized by the intermittent observation of a strong E region arc (1938-1959 UT) and significant F region horizontal gradients to the north. A shallow depletion is observed immediately poleward of the arc in scan C. This feature mimics the morphology of a typical auroral ionospheric cavity but cannot be unambiguously classified without prior observation of the characteristic ambient F region electron density, which is difficult to determine due to the intrinsic `patchiness' during this period. During scan B, just prior to panel C, there appears to be a depletion in the F region with minimum density at the core of this depleted region on the order of = 6.210 cm ( = 2.23 MHz), ~55% below N. It is unlikely that the depletion shown in panel B is related to a downward field-aligned current filament, the classic auroral ionospheric cavity formation agent, as there is little evidence for an associated upward field-aligned current signature such as an E region arc. This density depletion may be, however, a fossil cavity which has advected into the ISR field of view. Depletions of this scale were observed continuously from 1907 through 1935 UT (not shown).
Plate 4 shows a 2 auroral roar event observed on March 12, 1996, beginning near 0057 UT, weakening for ~15 min, and returning in an intense burst near 0116 UT. During this event, the ISR was performing a series of elevation scans through the zenith, covering more latitude than in the earlier experiments. These scans were done at a constant angular rate (giving variable ground tracking). The horizontal spatial resolution at 275 km altitude varied from 137 km at the ends of the scans to 44 km when the scan went overhead. For these scans, most of the interesting ionospheric structure lies within 45 of zenith, a region with an average horizontal spatial resolution at 275 km altitude of 60 km.
An auroral ionospheric cavity was observed to form in ISR scans C and D, and it can be observed during two subsequent scans E and F. Scan C (starting at 0104:33 UT) shows a classic cavity with an F region depleted by 45% to a value of = 5.6210 cm ( = 2.13 MHz). Prior to 0104:33 UT the E region arc is too far south of the radar to adequately bound the equatorward edge of the ionospheric cavity, but the depleted F region to the south of zenith is most likely coincident with the cavity. The F region ionosphere is patchy during this entire period, and the scans include the boundary of the auroral zone/polar cap, similar to the scenario for August 15, 1995 (Plate 2).
Plate 5 shows a 2 auroral roar event observed on March 29, 1996, near 2.7 MHz starting near 2340 UT (scan C) in the presence of severe radio frequency interference. A separate emission, called MF-burst, can also be seen during this time at a lower frequency (~2.5 MHz). The ISR operating mode was identical to that of the first three experiments, and the resulting resolution is the same as that of Plates 1, 2, and 3. The sequence of ISR scans from panel B to panel E has similar salient features as described for the August 15, 1995, and the March 12, 1996, events (Plates 2 and 4): a well-structured polar cap F region ionosphere, a dense E region auroral arc, and an F region auroral ionospheric cavity between them. The latitudinal extent of the cavity appears to grow somewhat larger (~120 km) in scan C, but it tracks the poleward boundary of the aurora in all six scans. For the scan associated with the auroral roar emissions (panel C), the F region is depleted by approximately 65% to a value of = 4.4810 cm ( = 1.55 MHz).
Plate 6 shows a 2 auroral roar event observed on April 9, 1997, starting near 0015 UT and occurring intermittently through radar scans labeled A-F. The ISR mode was similar to that of Plate 1 (constant ground tracking composite scans), but the radar covered more latitude and scanned at an overall faster rate, and the scans were offset to the east and west by 25. The resulting horizontal spatial resolution at 275 km altitude is 36 km. The absolute electron density in the F region overhead is higher than in most of the examples shown here, with the exception of the August 3 event (Plate 1), which this event resembles in the respect. However, unlike the August 3 event, significant horizontal F region gradients are present and persist during all of the ISR scans. For instance, several regions are evident where the density is depleted by ~55% to a minimum value of = 5.7510 cm ( = 2.15 MHz). No F region arc is evident in these scans and hence no ionospheric cavity. There is evidence for a broad dip in the electron density to the south which persists for over 20 min during scans A through F.
The 6 days of auroral roar observations shown in this paper are characterized by a variety of ionospheric conditions summarized in Table 1. Five out of six auroral roar events can be related to a depleted F region ionosphere, characterized by patchiness and relatively large horizontal density gradients or relatively small gradient scale lengths ( / 120 km measured with 23-137 km resolution), and three of these are associated with classic auroral ionospheric cavities at the poleward edge of the auroral zone. The remaining event occurs when the F region ionosphere is unstructured and laminar.
The existing theories of auroral roar generation can be classified as either direct or indirect depending on whether electromagnetic (EM) radiation is stimulated in the source region and propagates directly to the ground in the O or X mode, or electrostatic (ES) Z mode waves are generated and mode convert to an escaping O or X mode. Several of these theories require the presence of density cavities, possibly auroral ionospheric cavities, or enhancements in the source region. The direct O/X mode maser theory [Weatherwax et al., 1995; Yoon et al., 1996] requires significant density cavities to provide a region for EM waves to reflect back and forth at the cavity walls, passing multiple times through the growth region within the cavity until they escape and are observed on the ground. Growth rate calculations predict that the X mode waves dominate [e.g., Yoon et al., 1996], but recent observations of auroral roar polarization as the O mode [Shepherd et al., 1997] make this an unlikely candidate for the generation of auroral roar emissions. More recent growth rate calculations by Yoon et al.,  confirm previous calculations of a positive growth rate for the upper hybrid (UH) or Z mode [e.g.,Kaufmann, 1980] and show that the Z mode growth rate far exceeds that of the O or X mode for / near and 3. The Z mode is trapped in the ionosphere and a mode conversion mechanism is required to produce escaping O mode radiation. One such conversion mechanism is the Ellis Window which requires an inhomogeneous electron density in order for the Z mode waves to refract and achieve the conditions required for conversion [Yoon et al., 1998]. Another conversion mechanism requires a density cavity for growth of the fundamental Z mode waves and to provide reflection at the cavity walls for coalescence of Z mode waves into an escaping EM mode [Willes et al., 1998]. The escaping radiation must be in the O mode for this theory to still be considered a candidate.
The typical auroral ionospheric cavities observed in Plates 2, 4, and 5 are ~50-70 km in latitudinal extent, are ~45-65% depleted relative to ambient densities, and appear to be adequate for the requirements of these theories. The presence of an ionospheric cavity also implies that horizontal gradients in the F region ionosphere also exist and could possibly be the location of wave mode conversion. It is tempting to conclude that these clear cases of ionospheric cavities, present during the simultaneous observation of auroral roar, support one of the previously mentioned theories. There is, however, one case that shows no signs of cavities. For this laminar case (Plate 1), it is unlikely that the spatial coverage of the instrument might simply fail to coincide with the source region since a laminar F region persists for many scans, and presumably the source region would eventually move through the path of the radar scanning magnetically North/South. Another possible scenario is that the temporal resolution (~5 min per scan) is too slow or the spatial resolution (~23 km) too large to capture a dynamic source region. In many of the events, auroral roar can be seen to change on a timescale of several seconds and high time resolution measurements show auroral roar is often composed of many narrow filaments that persist for only a few milliseconds and possibly less Shepherd:98.
Significant horizontal F region gradients appear in five of the six examples studied. Such gradients are an intrinsic feature of the source region in the Z mode Ellis window theory [Yoon et al., 1998a], the Z mode coalescence theory Willes et al., 1998], and other theories for the mode conversion portion of an indirect generation mechanism including scattering on small-scale (5%) density fluctuations [Bell and Ngo, 1990] and wave-wave interactions [e.g., Weatherwax et. al., 1995]. Measured gradient scale lengths are summarized in Table 1. In order to characterize these gradients, we used the ISR density profiles from individual radar scans to compute normalized horizontal gradient scale lengths (/) at six selected F region altitudes: 250, 260, 270, 280, 290, and 300 km. The spatial resolution of the data is the limiting scale of the ISR measurements at these altitudes, which varies from ~23 to 137 km and is summarized in Table 1. Both the maximum and average of the rectified and normalized gradient scale lengths over the entire horizontal range of the ISR scans are given in Table 1. The latter indicates an average value for the detectable F region electron density gradients, while the former measures the magnitude of the largest detected gradient. Table 1 illustrates the laminar nature of the August 3 event. The minimum normalized gradient scale length for this event is large (215 km) compared to the other events indicating that no significant horizontal gradients are present. Of course, no information is available about possible features during this event with scales less than ~30 km, which are probably also present in the F region ionosphere during auroral roar emissions and could be significant to their generation.
Auroral roar may be much more common in the ionosphere than it is at ground level [e.g., Weatherwax et. al., 1995]. The canonical explanation is that auroral arcs are most likely responsible for generating the emissions, but when they move directly overhead a ground station, they also screen the emissions from the ground due to the increased ionization raising the plasma frequency cutoff in the E region above the emission frequency. Conversely, the observed presence of auroral ionospheric cavities may serve to locally reduce the plasma frequency and allow HF ducting from the ionosphere to the ground. Plate 5 shows a sequence of ISR scans which support the screening hypothesis. The early scans (A through C) show a somewhat structured F region developing. Observation of auroral roar occurs primarily in segment C. An E region arc develops in scan A at 2318:13 UT, strengthens, and moves overhead in scan D near 2345:14 UT just at the termination of the auroral roar observation. ISR data from the August 15, November 1, and March 12 events (some ISR scans not shown) suggest a similar effect.