Measurements of neutral gas density in the thermosphere at the base of the Earth's south magnetic cusp from the Streak mission are reported and discussed. In contrast to recent reports of enhanced density in the cusp, these measurements show the density to be depleted relative to the surrounding region. The difference is interpreted as an altitude effect. This observation and interpretation lead to new constraints on the physical mechanisms that could explain cusp upwelling and are inconsistent with a Joule heating mechanism. A mechanism based on heating by soft cusp particle precipitation is put forth, and model calculations are used to show how it explains the relative depletions observed by Streak as well as the enhancements at higher altitudes reported previously.
 The thermosphere, along with the comingled ionosphere, is the transition region from the dense atmosphere below to space above. The region is far from homogeneous and is characterized by transitions among regimes in numerous physical parameters. Thermospheric structure is apparent across a wide range of scales and is produced by a large variety of phenomena having origins within the thermosphere-ionosphere system as well as being the result of interactions with the external environment. This paper discusses structure associated with the Earth's magnetospheric cusp, a portion of the thermosphere subject to considerable external influence.
 The existence of mesoscale structure in the thermosphere has been underscored by recent measurements from satellite experiments. Structures associated with the equatorial ionization anomaly have been investigated using measurements from the Streak [Clemmons, 2006] and CHAMP [Liu et al., 2007] missions, as have features associated with the magnetospheric cusp [Lühr et al., 2004; Liu et al., 2005] and the south Atlantic magnetic anomaly [Clemmons, 2006]. These phenomena indicate strong influence by the Earth's magnetic field, and thus the interaction of the thermosphere with charged particles, in determining mesoscale thermospheric structure. They are also regions where localized forcing phenomena produce strong, observable effects, so they are amenable to study using straightforward techniques.
 Mesoscale structure in the thermosphere has a considerable history in the literature and has been reviewed recently by Moe and Moe . However, it is the recent measurements from the CHAMP mission that have caused new interest. They are based on state-of-the-art experimental technique through the use of a sensitive accelerometer hosted by a carefully-constructed spacecraft. CHAMP's high duty cycle and longevity have led to several years of systematic thermosphere measurements.
 CHAMP's cusp measurements are summarized here based on the work of Lühr et al.  and Liu et al. . Significant enhancements in the thermospheric mass density in and near the cusp were seen. The enhancements are prominent when comparing with either the density in nearby regions or the values predicted by the MSISE90 [Hedin, 1991] model. The measurements covered the altitude range 390–460 km and were associated to the cusp by observation location and co-location with intense fine-scale field-aligned currents measured by CHAMP's magnetometer. Enhancements were observed for geomagnetically quiet and moderately-disturbed conditions for both cusps. A year of data showed the enhancement to average about 30% over MSIS, although individual measurements of 100% were also observed. The most likely explanation was considered to be upwelling of neutral gas due to Joule heating associated with the observed field-aligned currents.
 The present work describes observations complementary to those of CHAMP. They were made during ten months in 2005–2006 by an ionization-gauge based sensor, called the IGS, carried by the Streak mission. Streak utilized a Sun-synchronous high-inclination (96.3°) circular orbit. The orbit plane was in the dawn-dusk meridian and the initial equatorial altitude was 325 km. Streak carried propulsion to re-boost its orbit occasionally, and measurements were made down to 123 km before it reentered. Thus Streak sampled a lower altitude region than has CHAMP. Streak visited the southern polar cusp region quite often, and measurements returned from these passes are presented here.
 The IGS was mounted on the leading surface of the satellite to measure the pressure of gas rammed into it. Similar techniques were used previously [e.g., Newton et al., 1965; Champion et al., 1970; Rice et al., 1973]. The IGS was a miniaturized, modernized version of similar instruments that were flown on some of the earliest space missions [e.g., Spencer and Dow, 1954] used to explore the structure of the thermosphere. It was based on a commercial Bayard-Alpert gauge [Bayard and Alpert, 1950] having wide intrinsic dynamic range and utilized a custom controller and wide-dynamic range electrometer. In contrast to the accelerometer on CHAMP, which is sensitive to both number density and mean molecular weight, the IGS was sensitive only to number density. The IGS and the Streak mission are described by Clemmons et al. .
Figure 1 shows ten-second averaged pressure measurements from the IGS for a single day of orbits. Instrumental noise contributes about 0.01% to the uncertainty in these samples. Following the work with CHAMP [Liu et al., 2007], the data are plotted relative to the predictions of MSIS. The plot shows several features, including the rather coherent structure near the magnetic equator at dusk reported by Clemmons  and Liu et al.  and variability in the southern dayside region. Significantly, the large (several tens of percent) localized enhancements in the southern cusp (near noon magnetic local time (MLT) and 76° magnetic latitude (MLAT)) reported by Lühr et al.  are not seen in this plot. This absence is representative of all of the IGS southern dayside passes, of which there were 2625. However, comparing the measurements to MSIS predictions is more conducive to examining large-scale differences between them than it is to finding more subtle mesoscale features. Therefore a different approach was taken to seek a cusp effect.
 In order to focus on mesoscale features, an MSIS-independent technique based on simple filtering was employed. High- and low-frequency components were derived using a filter having a cutoff of 600 s. This value corresponds to a spatial extent of about 4600 km and was chosen because it separates large-scale features from mesoscale features quite well. The high-pass portion was divided by the low-pass portion to form a filtered dataset, and Figure 2 contains the data from Figure 1 thus treated. Figure 2 shows considerable variability at high latitudes, both in the northern nightside auroral region and near the southern cusp. Variations during these passes exhibit no coherent trend, i.e., nearby measurements have a wide scatter in magnitude and direction of variation from pass to pass. Again, this result is representative of the IGS measurements and likely the signature of waves or other disturbances. It is also noteworthy that on this day some of the largest variations appear well poleward of the cusp and are reminiscent of the waves in the polar cap reported by Johnson et al. . This discussion makes it clear that the level of perturbation due to any stable feature in the cusp must be much below the several percent variations seen in Figure 2.
 To determine whether the IGS detected any stable structure in the cusp, all of the IGS data were filtered as above, then averaged into bins of 0.1 hr MLT by 1° MLAT. In the region of interest, taken to be between 09 and 15 MLT and below −60° MLAT, nearly 105 ten-second averages representing a total of over 275 hr of observation time were included. Figure 3a shows the distribution of measurements, and Figure 3b shows the resulting picture of the density structure. The cusp is taken to be the region between about 72° and 76° MLAT and between 11 and 13 MLT, an identification that corresponds well with the statistical position of the low-altitude cusp identified by Newell and Meng . It is apparent that the density in the cusp was depleted relative to adjacent areas by 1–2%.
 The IGS measurements of a weak density depletion in the cusp differ markedly from CHAMP's observations of a strong density enhancement. The difference is likely due to the different altitude regimes of the two data sets. The CHAMP measurements were made about 150 km, or about three scale heights, above the IGS measurements. This difference is significant for a distinguishing characteristic of the cusp, namely the precipitation of soft particles. These particles generally lose most of their energy to collisions with neutral species before they reach altitudes below about 250 km. Therefore they can heat the thermosphere above this altitude sufficiently to cause the upwelling CHAMP observes, but cannot heat lower regions such that significant upwelling would be observed by the IGS. In this scenario the relative depletion seen by the IGS is interpreted as enhanced density in adjacent regions. In these areas the precipitation is relatively harder [e.g., Newell and Meng, 1988], so the particles penetrate more deeply to cause upwelling from lower altitudes. This idea is explored below.
 Further exploration of the altitude dependence of the cusp structure was performed by dividing the IGS data equally into four altitude ranges as shown in Figure 4. The bins were enlarged to 0.2 hr by 2°. An altitude effect is clear, showing that the magnitude of the cusp depletion (or adjacent enhancement) increases over the altitude range of the IGS measurements.
 The electron transport model of Strickland et al.  has been used extensively for auroral analysis [e.g., Hecht et al., 2006] and forms the basis for analyzing the effects of precipitating particles here. A model thermosphere based on average conditions prevailing during the mission was set up using MSIS. Ap was taken to be 10 and F10.7 was 100 – this interval was quite inactive geomagnetically. Vernal equinox conditions were used to represent average insolation during the mission. Precipitation parameters were derived from the average characteristics given by Newell and Meng . Cusp precipitation was modeled as a Gaussian energy distribution with a characteristic energy of 100 eV and an energy flux of 1.6 mW/m2. Parameters for the region adjacent to the cusp were 300 eV and 2.3 mW/m2. For reference a beam representing moderately strong aurora was also modeled – it was taken to be 1 keV and 10 mW/m2.
 The results from the model runs are compared in Figure 5a, which shows how energy is deposited for the three precipitation types. It is clear that the 100 eV population heats effectively at high altitudes, while the higher energy populations are more effective heating at low altitudes.
 In order to examine the effects of heating, the theory of Hays et al.  was employed to estimate the vertical winds produced by the energy deposition rates in Figure 5a. These estimates are plotted in Figure 5b. It is clear that at high altitudes the 100 eV population is an effective agent for strong upwelling. The 300 eV population also causes significant upwelling, but only about one-third as much. Thus these calculations are consistent with the CHAMP results, which show the strongest upwelling in the central cusp region with weaker upwelling in nearby regions. In contrast, at lower altitudes, the 300 eV population is much more effective at producing upwelling than the 100 eV population. Thus this scenario is also consistent with the Streak results showing enhancement in the cusp-adjacent regions relative to the cusp itself.
 Further runs were performed with different parameters driving the atmospheric model to determine the sensitivity to atmospheric changes. The results were qualitatively the same as those of Figure 5, but the phenomena moved in altitude. Increasing F10.7 by 30 units, or using summer solstice conditions, had the primary effect of raising the curves by about 25–30 km. Similarly, lowering F10.7 or using winter solstice conditions moved the curves down by about 25–30 km. Changing Ap had little effect. The conclusion is that the results above and their interpretation are robust to atmospheric changes, but that altitude changes of the order of one scale height are to be expected.
4. Discussion and Conclusion
 Although the above interpretation plausibly explains both the CHAMP and Streak observations qualitatively, Figure 5b shows that the winds in the cusp are predicted to be stronger than those adjacent to it above about 260 km. It would then seem to follow that Streak should have observed a cusp enhancement above about 260 km rather than the relative depletions of Figures 4b–4d. The key to reconciling this idea is to realize that Figure 5b shows the upwelling winds, not the densities. Density at a given altitude is a function also of the duration of the heating, the extent of the heating below that altitude, and the density below the altitude. Thus density change is an integrated effect of conditions below that altitude. Although calculating the magnitude of the integrated effect is beyond the present scope, its consideration brings the above interpretation back into the plausible range. Further credence is lent to this explanation by revisiting Figures 4b–4d and noting the slight asymmetry of the depleted regions with respect to the noon meridian. The asymmetry is toward later hours as would be expected by the integrated effect a structured, approximately fixed, heat source would have on the co-rotating thermosphere.
 A few other points should be made. First, the Streak results present a major challenge to the Joule heating explanation. Joule heating is strongest below 200 km and peaks substantially below that altitude, so this explanation predicts that Streak would have observed a density enhancement. Amplifying this point is that the nightside auroral zone plays host to strong Joule heating, yet CHAMP observes weaker enhancement there than in the cusp. The blue curve in Figure 5b shows the results of a heating source that extends to low altitudes. Despite the large energy input rate, depositing the energy in a dense medium makes for a less efficient generator of high-altitude upwelling. Second, the enhancements (or depletions) will always appear larger on CHAMP than on Streak because upwelling gas has not only higher number density, but also is richer in heavier species. CHAMP responds to mass density while the IGS responds to number density. Third, this work predicts that as the CHAMP orbit decays, the cusp enhancement will fade.
 In conclusion, the present work has shown that observations of upwelling in and near the cusp made by CHAMP and Streak differ primarily because different altitude regimes were sampled. Direct heating by soft cusp and near-cusp precipitation can explain both the enhanced densities at high altitudes as well as the relative depletions at low altitudes.
 Support of the Streak program by the Defense Advanced Research Projects Agency is gratefully acknowledged. A portion of this work was supported under The Aerospace Corporation's Independent Research and Development Program.