G. A. Germany, G. K. Parks, M. J. Brittnacher, J. F. Spann, J. Cumnock, D. Lummerzheim, F. Rich, and P. G. Richards
AGU Monograph, "Encounter Between Global Observations and Models in the ISTP Era", Jim Horwitz, Dennis Gallagher, and Bill Peterson, editors, 1998.
Abstract | Introduction | Technique | Data | Discussion | Acknowledgments | References | Figure Captions | Plate 1 | Plate 2 | Figure 1
The GGS POLAR satellite, with an apogee distance of 9 Earth radii, provides an excellent platform for extended viewing of the northern auroral zone. Global FUV auroral images from the Ultraviolet Imager onboard the POLAR satellite can be used as quantitative remote diagnostics of the auroral regions, yielding estimates of incident energy characteristics. UVI images have been used previously to study total hemispheric input into auroral regions and local time variations. The same analysis on a higher temporal and spatial scale can be challenging in the presence of dynamic auroral forms which change significantly between images. Here, such an event is analyzed and compared with in situ DMSP observations of the same event. The difficulties, and potential remedies, associated with such a study are discussed.
The principal science objective of the Ultraviolet Imager (UVI) is to provide global information on the flow of energy between the Earth's magnetosphere and its ionosphere [Torr et. al., 1995]. This is accomplished by using far ultraviolet (FUV) auroral images provided by UVI to estimate incident energy flux and average energy of the precipitating auroral particles causing the observed emissions. UVI images have been used previously to study total hemispheric input into auroral regions and local time variations [e.g. Brittnacher et al., 1997; Doe et al., 1997; Lummerzheim et al., 1997] but only limited effort has been expended to analyze UVI image data at high spatial and temporal resolution.
The Ultraviolet Imager is capable of temporal resolutions on the order of 1 second for specialized operating modes. In practice, however, the normal imaging resolution is 36.8 seconds. The spatial resolution varies with orbital altitude from a perigee resolution of about 5 km per pixel to 40 km per pixel at apogee. Since auroral forms can significantly change in brightness on a time scale shorter than the UVI temporal resolution, and since at least two image frames are required for average energy estimation, analysis of dynamic auroral forms can be particularly difficult - thus the relative lack of such studies to date. Here, such an event is analyzed and compared with in situ DMSP observations of the same event.
The technique for using FUV emissions as remote diagnostics depends on energy-dependent loss mechanisms, principally absorption by O2 and has been discussed elsewhere [e.g. Strickland et al., 1983; Germany et. al., 1994a,b; 1990]. The principal emissions within the UVI bandpass (125.0 - 200.0 nm) are atomic oxygen emissions and molecular N2 Lyman-Birge-Hopfield (LBH) emissions. Atomic nitrogen lines also appear. The LBH emissions are of particular interest since they are due solely to electron impact excitation. Thus, in the absence of dayside photoelectrons, LBH emissions are direct diagnostics of the incident auroral flux. The O2 Schumann-Runge absorption continuum peaks within the UVI bandpass, decreasing with longer wavelength. Auroral emissions viewed from space in this spectral region of strong absorption will thus exhibit some losses, provided the incident auroral energy is high enough to reach lower altitudes where O2 density is greatest. Therefore the short wavelength emissions exhibit an energy-dependent loss mechanism that can be used as a remote diagnostic of the mean incident energy.
The energy characteristics of the precipitating auroral particles are described by stating the flux of incident particles, their mean energy, and how their energy is distributed about that mean. The incident flux is typically specified as an energy flux in mW/m2. The energy distribution about the mean is typically given as either a Gaussian, a Maxwellian, or some other similar distribution. Ideally all three parameters would be derivable from UVI images.
Two of the requisite three parameters, incident energy flux and average energy, are provided from the intensity of the LBHl images and from the ratio of LBHl and LBHs (or 1356) images. The LBH emissions can be divided into two regions: one at shorter wavelengths with significant losses due to O2 absorption (LBHs) and longer wavelength emissions with less loss (LBHl). Germany et al. [1994b] studied the possibility that FUV auroral images could be used to specify the incident auroral energy distributions, i.e. the third parameter necessary to completely specify the energy characteristics of the incident auroral particles. They found that the magnitude of changes in column intensities with choice of energy distribution is less than about 30% for a number of different distributions, including Gaussian and Maxwellian. The dependence of modeled FUV auroral emissions on energy distribution is thus much less than their dependence on average energy or energy flux. The determination of these two quantities from UVI images should therefore be relatively unambiguous, since such determinations will not be dependent on an exact knowledge of the incident energy spectrum. Since the estimated errors in the analysis of UVI data (see below) are larger than the variations found by Germany et al. [1994b], the sensitivity of this type of analysis is generally not enough to discriminate between different energy distributions based on UVI images. Nevertheless, UVI images provide the two most significant diagnostics - energy flux and average energy.
Modeling of expected emissions can be used to estimate incident energy flux from LBHl intensities. Auroral intensities are modeled with a two-stream energy deposition code [Richards and Torr, 1990]. Emission cross sections of N2 LBH are those of Ajello and Shemansky [1985] with the downward scaling of Ajello et al. [1991]. The average energy can be estimated from the ratio of either OI 1356 or LBHs to LBHl. The ratio is necessary to normalize against changes in incident energy flux. Since the emissions in the LBHs and LBHl filter passbands originate from the same specie, the ratio is nearly independent of compositional changes with season or over a solar cycle [Germany et al., 1990].
In the ideal case, analysis of FUV auroral emissions would be performed with only two of the many LBH emission bands - one at the wavelength of peak absorption and the other at a longer wavelength where O2 absorption is negligible. The longer wavelength emission would therefore be essentially independent of average energy and solely dependent on energy flux. This is the case modeled previously by Germany et. al. [1994a,b; 1990]. In practice, however, the UVI LBH filters must necessarily include multiple LBH bands and therefore contain a range of loss factors. Thus the LBHl emission is not totally independent of average energy and shows a weak dependence with average energy that is not indicated in the previous work by Germany et. al. [1994a,b; 1990]. For example, the LBHl emission intensity produced by a fixed electron energy flux input decreases by about 10% for a change in average energy from 5 to 10 keV. Because this effect is relatively small, however, it is not included in the analysis shown below.
For the period examined here UVI was in an operational mode that employed only the two LBH filters. The discussion below is therefore limited to the analysis of the LBHs/LBHl ratio.
Quantitative photometric analysis of the type attempted here requires a proper knowledge of all error i.e., non-auroral, sources. In particular, auroral images must have instrumental and airglow backgrounds removed before analyzing. Airglow can be estimated by ab initio calculations. However, this requires the use of an airglow model that is tailored to meet the specific viewing requirements of the observation, e.g. changes in line of sight enhancements across the image. For the work presented here airglow is approximated by binning the airglow pixels by solar zenith angle, excluding auroral emissions, using a technique similar to that of Lummerzheim et al. [1997]. From these binned pixels an airglow surface is approximated and then removed from the original image.
A few caveats must be kept in mind when reviewing energy products derived from UVI images. First of all, the ratioed images used to determine the average energy are not temporally coincident since the imager has only a single optical path allowing only a single filter to be used at a time. Temporal and spatial changes in the auroral morphology between image frames can lead to incorrect energy estimations. This is especially evident in highly dynamic conditions in which the local intensities may change significantly between image times. Also, since the energy estimation utilizes ratios of images it is sensitive to uncertainties in both images. The results presented here have internal quality checks to remove these artifacts from the final products. For the work presented here image processing (binning, smoothing, ratioing, etc.) introduces about 3% uncertainty and the Poisson uncertainty is about 5%. Assuming an accuracy of 25% for the instrumental calibration, 22% for the LBH cross sections [Ajello and Shemansky, 1985], and 25% for all other model uncertainties provides a total uncertainty of about 45%.
Data analyzed is from May 19, 1996 (96140) with the POLAR spacecraft near an apogee distance of 9 Earth radii. Between 21:41 and 21:46 UT the DMSP F12 satellite passed through the northern auroral region at an altitude of about 800 km. Plate 1 shows a sequence of UVI images taken between 21:42 and 21:44 UT coincident with the DMSP overflight. The DMSP F12 satellite traveled northward through the nightside oval over Greenland. The UVI operations during this time used pairs of images in the following sequence: 2 LBHs, 2 LBHl, 2 LBHs, shutter. The first of each pair of images is an 18 second exposure instead of the normal 36 second exposures. (This is a UVI operational requirement to prevent image smearing during movement of the filter wheel.) The images in Plate 1 are the three 36 second exposures available for this period and are spaced about 1.25 minutes apart. These are the images used in the analysis below.
At 21:15 UT, about 30 minutes before the first LBHs image in Plate 1, the poleward boundary began the intensification that is clearly evident in the image. Activity propogates westward almost 90 degrees in longitude during the 1.5 minutes shown in Plate 1.
The two images in Plate 2 are energy maps derived from LBHl image and the second LBHs image beginning at 21:43:20 UT and 21:44:33 UT, respectively. The images are displayed in MLAT-MLT coordinates where MLAT is computed for an apex magnetic coordinate system and MLT is the corresponding local time. The average energy map (Eavg) is calculated from the ratio of the LBHl and LBHs images and has a time resolution set by the time encompassed by the two images (roughly 1 minute). The energy flux map (Eflux) is calculated from the LBHl image alone and therefore has a higher time resolution than the average energy map. The ground track of the DMSP overflight is shown.
The energy flux map shows concentrated high energy fluxes in a broad arc (in local time) on the poleward boundary of the oval. A secondary arc of lower energy is seen on the equatorward boundary which appears to be the remnant of previous activity beginning at 20:20 UT. The average energy map shows two bands of higher energy electrons colocated with the higher energy fluxes. A trough of lower energy electrons exists between the two arcs on the dusk sector and on the equatorward boundary near midnight. Little information is displayed for the dayside since the signal strength after dayglow removal was deemed too low for analysis.
We wish to compare the in situ DMSP observations with those derived from analysis of the UVI images. This is done by plotting DMSP energy parameters along its ground track and then superimposing values derived from UVI images for the same spatial locations. Note that since the two instruments have different time resolutions the comparisons along the ground will be coincident for only a limited time corresponding to the UVI integration time. If we assume that the overall auroral morphology is roughly stable over the 5 minute DMSP overflight we can compare values from the two instruments along the entire overflight using only the LBHs and LBHl images used to compute Plate 2. However, as shown in Plate 1, the morphology is clearly dynmic and changes signifcantly during this period. Nevertheless, for the purpose of discussion, the data from DMSP and the energy derived from UVI are superimposed in Figure 1.
Figure 1 shows comparisons between energy flux and average energy as determined from DMSP measurements and UVI image analysis along the DMSP ground track. In situ observations of incident energy flux and average energy are shown as solid lines. Crosses and boxes represent values from UVI energy maps corresponding to the ground tracks. Vertical lines correspond to the beginning of the UVI LBHl or LBHs integrations (36.8 seconds), as appropriate. The data marked by crosses is obtained from analysis of the first two images in Plate 1 while boxed data are derived using the final two images in Plate 1. Dashed lines represent the assumed 45% total uncertainty discussed above.
The derived energy flux from the first pair of UVI images (top of Figure 1) show generally good agreement for the equatorward part of the oval (UT less than 21:44 in the figure). On the poleward boundary the in situ DMSP observations show two spikes of energy flux that are significantly higher than is derived from the UVI images. Clearly the UVI data doesn't have the same resolution as teh DMSP observations and tends to smooth all the poleward structure into a single broad arc with its peak near the first (equatorward) brightening seem by DMSP but seemingly not detecting the second (poleward) brightening. Recall, however, that the poleward boundary is brightening and expanding throughout this event. The UVI data is properly coincident only at the time of the LBHl integration which ends by 21:44 UT. Thus, by the time the DSMP satellite reached the northern boundary the incident energy flux had increased over what was seen in the UVI image taken over a minute previously.
This interpretation is further borne out by the average energy comparisons shown in the bottom panel of Figure 1. The average energies derived using the first LBHs images are in general agreement for the equatorward portion of the oval but disagree strongly for the poleward portion. The image ratio used in this calculation is LBHl divided by LBHs. Since the LBHs preceded the LBHl image, a monotonic increase in poleward brightness as seen here results in an overestimate of the average energy. The boxed values derived from the second pair of images shows significantly better agreement, though UVI still overestimates the in situ average energy observations, presumably because of dynamic changes between the two image frames used in the analysis. An analysis employing the 18 second integration between the two could help further refine the analysis.
Thus Figure 1 shows both the power and the potential limitations of using FUV auroral images as remote diagnostics of the aurora. The analysis is most limited in highly dynamic situations in which the aurora changes significantly between two UVI images. Plate 2, however, best illustrates the potential of analysis based on Ultraviolet Imager data. The images allow quantitative estimates across extended regions of the auroral zone and are not restricted to a single ground track. Thus imaging data should be viewed as complementary to the various available in situ and ground based observations.
In this paper, UVI images have been used to estimate the magnitude of the incident energy flux over the entire auroral zone. The inferred energy fluxes generally agree in magnitude and morphology with selected DMSP overflights. The fluxes inferred from UVI images do not exhibit the same spatial or temporal resolution as the in situ measurements but offer a global perspective unattainable from single satellite passes. Average energy maps are also constructed from ratioed images. Care must be taken in interpreting such maps since they are subject to error from temporal and spatial changes as illustrated in Figure 1. Continued comparison with in situ, ground based, and other imaging observations is needed to build confidence in this approach. Such studies are currently underway.
This work was supported, in part, under U. Washington contract 256730 to the University of Alabama in Hunstville and NASA grant NAG5-3170 to the University of Washington. Work at the University of Alaska was supported by NASA grant NAG5-1097. A. Richmond kindly supplied the database upon which the apex coordinates are based. The authors gratefully acknowledge the many helpful comments of the reviewers.
Ajello, J. M., and D. E. Shemansky, A reexamination of important N2 cross sections by electron impact with application to the dayglow: The Lyman-Birge-Hopfield band system and
NI (119.99nm), J. Geophys. Res., 90, 9845, 1985.
Ajello, J. M., D. E. Shemansky, and G. K. James, Cross sections for production of H(2p, 2s, 1s) by electron collisional dissociation of H2, Astrophys. J., 371, 422, 1991.
Brittnacher, M., R. Elsen, G. Parks, L. Chen, G. Germany, and J. Spann, A dayside auroral energy deposition case study using the Polar Ultraviolet Imager, Geophys. Res. Lett., 24, 991 1997.
Doe, R. A., J. D. Kelly, D. Lummerzheim, G. K. Parks, M. J. Brittnacher, G. Germany, and J. Spann, Initial comparison of POLAR UVI and Sondrestrom IS radar estimates for auroral electron energy flux, Geophys. Rev. Lett., 24, 999, 1997.
Germany, G. A., D. G. Torr, P. G. Richards, M. R. Torr, and S. John, Determination of ionospheric conductivities from FUV auroral emissions, J. Geophys. Res., 99, 23297, 1994a.
Germany, G. A., M. R. Torr, D. G. Torr, P. G. Richards, Use of FUV auroral emissions as diagnostic indicators, J. Geophys. Res., 99, 383, 1994b.
Germany, G. A., M. R. Torr, P. G. Richards, and D. G. Torr, The dependence of modeled OI 1356 and N2 LBH auroral emissions on the neutral atmosphere, J Geophys. Res., 95, 7725, 1990.
Lummerzheim, D., M. Brittnacher, D. Evans, G. A. Germany, G. K. Parks, M. H. Rees, and J. F. Spann, High time resolution study of the hemispheric energy flux carried by energetic electrons into the ionosphere during the May 19/20 auroral activity, Geophys. Rev. Lett., 14, 987, 1997.
Richards, P. G., and D. G. Torr, Auroral modeling of the 3371 A emission rate: Dependence on characteristic electron energy, J. Geophys. Res., 95, 10337, 1990.
Strickland, D. J., J. R. Jasperse, and J. A. Whalen, Dependence of auroral FUV emissions on the incident electron spectrum and neutral atmosphere, J. Geophys. Res., 88, 8051, 1983.
Torr, M. R., D. G. Torr, M. Zukic, R. B. Johnson, J. Ajello,
P. Banks, K. Clark, K. Cole, C. Keffer, G. Parks, B. Tsurutani, J. Spann, A far ultraviolet imager for the International Solar-Terrestrial physics mission, Space Sci. Rev., 71, 329, 1995.
Figure 1. Comparisons between energy flux and average energy as determined from DMSP measurements and UVI image analysis. In situ observations of incident energy flux and average energy are shown as solid lines. Crosses and boxes represent values from UVI energy maps corresponding to the ground tracks. (Crosses are derived using the first two images in Plate 1; boxes from the second pair of images.) Vertical lines correspond to the beginning of each of the UVI LBHl or LBHs integrations, as appropriate. Each integration is 37 seconds. Dashed lines represent an estimated 45% uncertainty level (shown, for clarity, for only the crosses).
Plate 1. Sequence of UVI images taken between 21:42 and 21:44 UT on May 19, 1996 coincident with a DMSP F12 overflight. The DMSP satellite traveled northward through the nightside oval over Greenland during this period.
Plate 2. Maps of incident energy flux (Eflux) and average energy (Eavg). The ground track of the DMSP satellite is shown near 21:30 MLT.
