Volume 114, Issue A9
Magnetospheric Physics
Free Access

Diffuse, monoenergetic, and broadband aurora: The global precipitation budget

P. T. Newell

P. T. Newell

Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA

Search for more papers by this author
T. Sotirelis

T. Sotirelis

Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA

Search for more papers by this author
S. Wing

S. Wing

Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA

Search for more papers by this author
First published: 15 September 2009
Citations: 378

Abstract

[1] We have developed an auroral precipitation model which separately categorizes the discrete aurora and both the electron and ion diffuse aurora. The discrete aurora includes acceleration by two distinct physical mechanisms, namely, quasi-static electric fields, producing monoenergetic peaks, and dispersive Alfvén waves, producing broadband electron acceleration. The new model is not merely finer in magnetic latitude (MLAT) and magnetic local time (MLT) resolution than previous models but is parameterized by solar wind driving instead of Kp and is based on functional fits to the solar wind coupling function which best predicts auroral power. Each of the four auroral types in each MLAT and MLT bin is separately fitted, a departure from the traditional compilation of a handful of discrete models, each assigned to represent a Kp (or other activity index) range. The variation of any of these four types of aurora at any local time can be predicted on the basis of the specific solar wind history of an epoch. This approach permits perhaps the first comprehensive comparison of the hemispheric contribution of each type of aurora. It turns out that the diffuse aurora is surprisingly dominant, constituting 84% of the energy flux into the ionosphere during conditions of low solar wind driving (63% electrons, 21% ions). The diffuse aurora is far from quiescent, tripling in power dissipation from our low to high solar wind–driving conditions. Even under the latter condition, the diffuse aurora contains 71% of the hemispheric energy flux (57% electrons, 14% ions). The monoenergetic aurora contributes more energy flux (10% quiet, 15% active) than does broadband acceleration signatures (6% quiet, 13% active). However, the broadband aurora rises fastest with activity, increasing by a factor of 8.0 from low to high driving. Moreover, this most dynamic auroral type contributes very high number fluxes, even exceeding monoenergetic aurora under active conditions (28% of hemispheric precipitation versus 21%). Thus, dynamic ionospheric heating and ion outflow is likely heavily affected by the wave aurora. Although energy flux peaks on the nightside, number flux peaks on the dayside. The cusp, as previously reported, is much better defined by ions than electrons. Hence, the ion number flux peak is confined, corresponding to the cusp, while the region with high electron number flux is broad (a cleft, corresponding to the boundary layers, including the closed low-latitude boundary layer).

1. Introduction

1.1. Overview

[2] Auroral precipitation includes electrons, which may be accelerated, and ions, which seldom are. There are two types of auroral electron acceleration commonly observed. The more commonly discussed is acceleration over a narrow range of energies, often called a monoenergetic peak, an auroral form which is associated with quasi-static electric fields aligned with the geomagnetic field by multiple methods of observation and deduction [Hallinan and Davis, 1970; Evans, 1974; Haerendel et al., 1976; Mozer et al., 1980; Kletzing et al., 1983]. These are often embedded within large-scale structures, tens or even hundreds of kilometers in latitudinal extent, usually called “inverted Vs” from the shape some make on spectrograms. Despite the nomenclature, most rise sharply to a relatively constant accelerating potential and hence have large gradients at the edges [Newell, 2000]. The other type of electron acceleration observed occurs over a wide range of energies [Johnstone and Winningham, 1982; Lotko, 1986]. We will call this “broadband” acceleration. It is believed to be caused by dispersive Alfvén waves (DAWs) [Ergun et al., 1998; Chaston et al., 2003], and although for years thought to be of relatively minor consequence, it has been more recently demonstrated to be significant [Wygant et al., 2000; Chaston et al., 2007]. Of course electrons and ions do not need acceleration to precipitate. When the meandering motion of charge particles leads to contact with the upper atmosphere, the resulting precipitation constitutes the diffuse aurora. We will study here both the electron and ion components of the diffuse aurora.

[3] There have been several statistical studies of monoenergetic aurora [Lin and Hoffman, 1979; Newell et al., 1996a], along with their associated electric fields and numerous ancillary effects, such as upflowing ion conics [Øieroset et al., 1999] and auroral kilometric radiation. Broadband, or wave aurora, had been primarily studied in a few individual cases, until a recent series of papers by Chaston et al. [2003, 2004, 2007] helped elucidate several basic facts. Note that these studies were based on the electromagnetic signature (although they have firmly established that broadband electron signatures correspond to DAWs). Similarly, there have been several previous models of total electron precipitation [Hardy et al., 1987; Sotirelis and Newell, 2000] and even a few ion models [Hardy et al., 1989; Newell et al., 2005].

[4] We present here, for the first time, an empirical model of the diffuse aurora, that is, of auroral precipitation with acceleration events explicitly removed. That permits the first quantitative comparison between these four types of auroral precipitation. One prominent result is that the diffuse aurora is much more important than often realized, constituting more than three quarters of the energy budget, and is more dynamic, rising rapidly with increasing solar wind driving (though monoenergetic and especially broadband aurora show even more dramatic increases).

[5] Our approach differs from previous work in several ways beyond separately categorizing each type of auroral precipitation. By binning by an empirical function thought to be proportional to the dayside merging rate, dΦMP/dt = v4/3BT2/3sin8/3(θ/2) (where BT = (By2 + Bz2)1/2 and θ is the interplanetary magnetic field (IMF) clock angle), and thus solar wind driving, we greatly improve the ability to predict auroral power [Newell et al., 2007a]. Our fitting is functional (least squares regression) rather than a handful of large bins, as has been done previously, which further improves accuracy (and avoids large jumps between nearly adjacent conditions). The magnetic local time (MLT) and magnetic latitude (MLAT) resolution is greater than in previous models, which is possible because of the larger database. In the near future, we will introduce the effects of UV insolation and separation by hemisphere, among other enhancements.

1.2. Historical Background

[6] Early sounding rocket shots through active auroral displays often found that most of the energy flux appeared over a narrow range of energies, typically one or two electrostatic analyzer channels [Evans, 1968, 1974]. This was interpreted to mean that a field-aligned accelerating potential equal to that monoenergetic peak energy exists above the aurora. Detailed analysis of the curling of auroral forms clearly implied space charge buildup and hence field-aligned electric fields [Hallinan and Davis, 1970]. Subsequently, the observation of such field-aligned potentials has been verified in a variety of ways, including Ba+ releases (allowing the charged ion motion to be tracked) [Haerendel et al., 1976] and by satellite electric field probes [Mozer et al., 1980; Kletzing et al., 1983; Marklund et al., 1994].

[7] Monoenergetic aurora is also often called inverted V [Frank and Ackerson, 1971] precipitation, although the majority of such acceleration events do not have gradual potential ramps up and down as that name implies [Newell, 2000]. More typically, sharp gradients, and hence large electric fields, exist at the edges, with relatively constant potentials (and small electric fields) within the large-scale auroral events.

[8] Broadband electron acceleration, although intermittently discussed in the literature [Johnstone and Winningham, 1982; Temerin et al., 1994] for nearly as long as monoenergetic events, was for many years considered to be far less important. As the properties of quasi-static electric fields were observed and monitored, broadband acceleration was associated with the edges of inverted V events [Burch, 1991], although some authors reported that broadband acceleration can be distinct from inverted Vs [Newell, 2000].

[9] Wygant et al. [2000] used polar observations to show that strong Poynting flux into the ionosphere can exist associated with Alfvén waves, directly above an active aurora display. Meanwhile, Ergun et al. [1998] showed that broadband electron acceleration was associated with kinetic Alfvén waves in many instances. In a series of articles, Chaston et al. [2003, 2004, 2007] have established many of the basic facts about the wave aurora. They have shown convincingly that DAWs correspond to broadband electron acceleration, that DAWs have the greatest energy flux into the dusk-to-midnight oval (although the frequency of events is highest on the dayside), and that their frequency rises rapidly with increasing geomagnetic activity.

[10] As for diffuse aurora, Wing and Newell [1998] showed for ions, and Kletzing et al. [2003] with electrons, that the diffuse precipitation is primarily a kappa distribution (as is true of the source plasma sheet population [Christon et al., 1991]). That is, it is Maxwellian, but with a power law tail at high energies. The primary question for the diffuse aurora is how the loss cone is continually refilled, given the relative shortness of the bounce time (seconds for electrons, minutes or tens of minutes for ions). The answer is better understood for ions, for which pitch angle scattering crossing curved magnetic fields is crucial. This is because the ion gyroradius can be comparable to the radius of curvature of the magnetic field lines. On the nightside, pitch angle scattering occurs crossing the current sheet [Sergeev et al., 1983] and precipitation ceases after the curvature of the field lines is reduced moving earthward [Newell et al., 1998b]. On the dayside, curvature effects associated with the magnetopause and boundary layers can cause ion precipitation [Sergeev et al., 1997].

[11] By contrast, the electron diffuse aurora is scattered mostly by waves, especially broadband electrostatic waves. On the dayside, where such wave activity is less intense, there can be trapped plasma sheet energy electrons which do not precipitate. Solar wind pressure pulses can lead to the observation of an electron diffuse aurora equatorward of previously existing dayside precipitation, presumably because of induced pitch angle scattering [Liou et al., 2002].

2. Data and Methodology

2.1. DMSP Particle Data Set

[12] Data from the SSJ/4 electrostatic analyzers on the DMSP series satellites (F6 through F12 and a small amount of F13) from 1 January 1988 through 31 December 1998 were used to compile cusp boundaries. These 11 years, besides covering all phases of a solar cycle, in general contain higher-quality data than some later years, during which increased operational longevity produced deterioration of sensitivity and hence the reliability of boundary identification in several detectors. During recent years, some SSJ/4 detectors have been launched with the low-energy ion head performing suboptimally, which has also reduced the quality of the ion precipitation data. For these reasons, our study concentrated on an even 11 years (one solar cycle) worth of good quality data.

[13] The DMSP satellites are in Sun-synchronous nearly circular polar orbits at about 845 km altitude, with orbital inclinations of 98.7°. The orbits of the DMSP satellites are such that the least covered regions are postnoon and especially postmidnight, except at high magnetic latitudes. The years chosen here do, however, cover all local times with the limitation that the postmidnight region is covered only from 1992 to 1998. The SSJ/4 instrumental package included on all these flights uses curved plate electrostatic analyzers to measure electrons and ions with one complete spectrum each obtained per second [Hardy et al., 1984]. The satellites are three-axis stabilized; and the detector apertures always point toward local zenith. At the latitudes of interest in this paper, this means that only highly field aligned particles well within the atmospheric loss clock are observed.

2.2. Methodology and Algorithms

[14] Figure 1 shows two sample spectrograms, illustrating the distinction between the two types of discrete aurora. In both parts of Figure 1, places where the electron energy flux rises above about 108 eV/cm2 s sr eV are accelerated (a more formal identification methodology follows in sections 2.2.1 and 2.2.2). In Figure 1a, monoenergetic events dominate, as readily seen by the narrow range of acceleration. In the auroral crossing shown in Figure 1b, broadband acceleration dominates the total energy flux. Although a small monoenergetic event does exist at 1404:15 UT (a so-called inverted V), most of the acceleration seen in Figure 1b involves a number of energy channels, typically from the lowest channel up to a cutoff level, with relatively flat energy flux between.

Details are in the caption following the image
(a) A spectrogram of DMSP F7 particle data showing a monoenergetic aurora dominated crossings of the nightside oval. (b) An example in which broadband acceleration is dominant.

[15] To better illustrate the two types of electron acceleration, Figure 2 shows four individual spectra. The top plots are selected from Figure 1a, and represent monoenergetic acceleration events. One or two channels spikes above 108 eV/cm2 s sr eV, and those channels stand out far above the neighboring channels. The bottom plots of spectra in Figure 2 are from the crossing shown in Figure 1b and illustrate broadband acceleration. Here, multiple energy channels are accelerated to roughly comparable values. Typically in a broadband event, the acceleration covers every channel up to a particular energy, and there is a relatively rapid dropoff above that energy. These criteria are made more quantitative in sections 2.2.12.2.4.

Details are in the caption following the image
The top plots show monoenergetic spectra from the DMSP pass shown in Figure 1a. The bottom plots show two broadband acceleration spectra from Figure 1b.

2.2.1. Identifying Monoenergetic Events

[16] Our algorithm first identifies the differential energy flux peak and subsequently looks at the drops one and two energy channels above and below the peak. If the differential energy flux drops to 30% or less of the peak within these two steps (at energies both above and below the peak), then the event is considered monoenergetic. A Maxwellian provides a useful comparison:
equation image
The peak occurs at E = 2kT, and the neighboring energy channels measured by DMSP have an energy flux of about 80% of the peak (the SSJ/4 detector steps by a factor 1.43 between energy channels). To be flagged as monoenergetic, the drop must be significantly greater than a Maxwellian, specifically to below 30% of the peak, moving to both higher and lower energies.

[17] The second major condition for monoenergetic acceleration events is that the differential energy flux must be above 1.0 × 108 eV/cm2 sr eV at the peak channel. Our experience is that the diffuse aurora never reaches this threshold and that such spectra stand out on spectrograms, apart from neighboring spectra. The value was chosen to be the smallest threshold which seemed to exclude homogeneous aurora (compare the discussion of diffuse aurora, or central plasma sheet, by Newell et al. [1996b]).

[18] A few additional conditions apply. If either the average energy is below 80 eV or the differential energy flux peak is below 100 eV, the spectrum is not considered “accelerated” no matter how intense the peak is. Such events may be spacecraft charging, although probably some very low acceleration potentials are excluded by this rule (but relatively little energy flux). Finally, if the peak energy flux is at the highest energy channel, then a sharp drop off is required only at lower energies.

2.2.2. Identifying Broadband Acceleration

[19] If three or more energy channels have dJE/dE > 2.0 × 108 eV/(cm2 s sr eV), an event is considered broadband, with some caveats. The first such exception is that a spectra satisfying a monoenergetic peak condition is not considered broadband. Another condition is that the average energy be greater than 80 eV and that the acceleration must extend to 140 eV (that is, there must be one or more channels with dJE/dE > 2.0 × 108 eV/(cm2 s sr eV) at or above 140 eV).

[20] An awkward circumstance, at least for our purposes, exists in the cusp, where very high energy fluxes at relatively low energies can exist unaccelerated. It turns out that in a significant minority number of cases, apparently Maxwellian spectra in the cusp can, just barely, fit the above criteria. Therefore, an additional condition was imposed requiring that within 9.5 < MLT < 14.5, the broadband acceleration must reach 300 eV. This probably does exclude modest instances of weak broadband acceleration. In practice, such events are difficult to reliably distinguish within the intense cusp fluxes based on the particle data alone. Note that the energy flux associated with such weak events is not much different from the overall cusp at any rate; hence, the impact of such exclusions on the hemispheric budget is not large. (It is almost a tautology to state that events that differ little from the cusp in which they are embedded make little difference.)

[21] It is possible for a spectrum to fit both the monoenergetic criteria and the broadband criteria [e.g., McFadden et al., 1998, Figure 1]. The full resolution of the conflict may be somewhat tedious for most readers. Basically, consideration is given to stricter conditions for both monoenergetic and broadband. If the flux drops to 10% of the peak (instead of 30%) within two energy channels from the peak, a spectra is considered monoenergetic regardless of whether the broadband condition is satisfied. Conversely, if stricter broadband criteria are met (four or more accelerated channels, i.e., with dJE/dE > 2.0 × 108 eV/cm2 s sr eV) but only weak monoenergetic conditions described in the section 2.2.1 are satisfied, then the spectra is broadband. Otherwise, meeting both sets of criteria will result in a spectrum being flagged as monoenergetic.

2.2.3. Diffuse Aurora

[22] Any spectrum not flagged as either broadband or monoenergetic is counted as diffuse, so all spectra are counted in one way or another. It has been established that both ions and electrons have a kappa distribution in the magnetotail [Christon et al., 1991; Wing and Newell, 1998; Kletzing et al., 2003] and in diffuse precipitation. Dayside populations also have a suprathermal population, although not necessarily a kappa power law (e.g., halo or strahl populations). Although broadband and monoenergetic spectra generally drop sharply above the accelerating range, diffuse aurora may at times have a significant fraction of its energy flux outside the DMSP detector upper limit (30 keV). This is particularly true for ions, especially plasma sheet ions within 10–20 RE, where temperatures can be 5–10 keV (recall the energy flux peaks at twice the temperature). The diffuse spectra were thus extrapolated to 50 keV, with the additional energy flux based on a simple Maxwellian fit. This upper limit is slightly more than one additional energy channel beyond the measurements. Extrapolation added about 3% to the diffuse electron aurora, and somewhat more to the ions (although the ions could quite reasonably have been extrapolated to 100 keV, as done by Hardy et al. [1989]).

2.2.4. Ion Aurora

[23] The ion aurora is constructed using every spectra, regardless of how the electron component is classified. Therefore, no probabilities or categories are associated with the ion aurora. Otherwise, it is treated much like the electron diffuse aurora. In particular, spectra are extrapolated from the measured limit of 30 keV up to 50 keV, under the assumption that the spectra is Maxwellian, with a temperature equal to half the differential energy flux peak. This approach certainly underestimates the total energy flux associated with ions, particularly in the equatorward portion of the dusk-to-midnight sector. It would probably require measurements up to about 100 keV to minimize such underestimation. Nonetheless, on the basis of our measurements only to 30 keV, we do not feel comfortable extrapolating to high energies.

[24] Because the spacecraft often charges up to −28 eV (when the solar batteries are being charged, in sunlight), a test is made for artificially high count rates in the lowest channel. If this is the case, then the channel is simply zeroed to prevent spurious results in the spectral moments.

2.3. Model Construction

[25] The grid size is 0.25 h MLT by 0.25° MLAT, with the latter covering 60–90°. This resolution is higher than used in any previously presented precipitation model, representing the larger amount of data available. Altogether, 9632 satellite days were used, from 1988 to 1998 (at least two satellites operational at a time, with occasional periods of as many as four).

[26] Each type of aurora is given a separate linear regression fit to the solar wind parameter which best predicts hemispheric global power, dΦMP/dt = v4/3BT2/3sin8/3(θ/2) [Newell et al., 2007a]. Thus, there are 4 (types of aurora) times 96 (MLT bins) times 120 (MLAT bins) which equals 46,080 individual regression fits constituting the model. This is necessary because each type of aurora responds differently to solar wind driving (e.g., wave aurora is most sensitive, diffuse aurora is the least) and because at different latitudes and different local times responses can be quite different (witness the difference between dayside and nightside response to activity level, for example). Each of these 46,080 linear regressions are of the form
equation image
Within each MLAT-MLT bin there are also three linear regression fits predicting the probability of observing each of three types of electron aurora, also as a function of dΦMP/dt. Thus, the energy flux of, say, monoenergetic aurora at a given MLAT-MLT bin is based upon the product of the fitted intensity of monoenergetic aurora, when it is present, with the probability of observing monoenergetic aurora (fraction that exhibit a monoenergetic peak). Thus, for each of the four types of aurora
equation image
The probability of observing the ion diffuse aurora is, however, taken to be unity everywhere.

[27] Note that there is no postmidnight region coverage from 1988 to 1991, although acceptable coverage does occur in 1992 through 1998. The solar wind driving over the course of a single year, or even a few consecutive years, can be significantly different from the long-term averages, and hence the “all conditions” values given below, which have no solar wind sorting, are susceptible to aliasing. However, even quiet and active years have a wide range of solar wind conditions, and the process of making functional fits to the auroral power yields results that, by definition, are normalized to the solar wind conditions. Thus, the low and high solar wind–driving models presented here should not be impacted. This aliasing of the unnormalized (all conditions) data is one reason that model is not presented graphically, and the results are not extensively discussed. However, the all conditions model does nonetheless fall between the low and high solar wind driving, as one would hope.

[28] We assume isotropy for all types of aurora, multiplying the directional energy (number) flux by π to get energy (number) flux. For both the electron and ion diffuse aurora, this is a reasonably good assumption. Both broadband and monoenergetic aurora tend to be field aligned. The loss cone at DMSP altitude is quite large (typically about 56° from the vertical is still within the loss cone even at 50° MLAT), and even the discrete aurora tends to be relatively isotropic within the loss cone. Nonetheless, we probably modestly overestimate the accelerated aurora by assuming isotropy. An advantage of our simple approach (multiplying directional fluxes by π) is that one can easily revert to the more directly measured directional fluxes if desired.

[29] Finally, hemispheric power is calculated by simply multiplying the surface area of each MLAT-MLT bin by the auroral power averaged over that bin, and summing over bins (separately, for each type of aurora). Although Northern and Southern Hemisphere data are individually tabulated, all the results presented in this paper come from combining the two hemispheres.

[30] Number flux is calculated in much the same way. An entirely new set of 46,080 individual linear regression fits (equalling 4 types of aurora times 96 MLT bins times 120 MLAT bins) are made of the same form as the auroral power fits. The same probability regression fits are reused, however, to calculate total number flux within each bin attributable to each type of aurora.

3. Comparative Energy Flux for the Four Auroral Precipitation Types

[31] Figure 3 shows the hemispheric precipitating energy flux averaged over one solar cycle for monoenergetic auroral events under (a) conditions of low solar wind driving (which was chosen to be dΦMP/dt = 0.25 〈dΦMP/dt〉, that is, one quarter the average strength of solar wind driving), and (b) moderately high solar wind driving (dΦMP/dt = 1.5 〈dΦMP/dt〉). For reference, averaged over one solar cycle, 〈dΦMP/dt〉 = 4421(km/s)4/3(nT)2/3. Thus, our active conditions are only moderately active (50% above average). By staying well within the normal range of activity, any questions such as saturation at high solar wind driving or other possible nonlinear effects are evaded.

Details are in the caption following the image
Monoenergetic hemispheric energy flux for conditions of (a) low and (b) high solar wind driving.

[32] Most of the energy flux precipitates into the dusk-midnight sector, in keeping with earlier surveys of so-called inverted Vs [Lin and Hoffman, 1979] and of monoenergetic electron acceleration events [Newell et al., 1996a]. Much weaker secondary peaks occur prenoon (during active conditions) and postnoon (for quiet conditions). The band of monoenergetic acceleration events corresponds to the statistical Ijima-Potemra R1 system out of the ionosphere [cf. Newell et al., 1996a, Figure 2]. Between low and high levels of activity, the precipitating energy flux from monoenergetic events rises by a factor of 5.3 (from 1.1 to 5.8 GW).

[33] Figure 4 shows the results for wave aurora in the same format as Figure 3. The peak energy flux once again lies in the premidnight sector. Chaston et al. [2007] recently demonstrated that the DAW Poynting flux into the ionosphere peaks premidnight. In going from low to high solar wind–driving conditions, the energy flux into the ionosphere rises by an impressive factor of 8.0 (from 0.6 to 4.8 GW). This is by far the largest rise of any of the four types of auroral precipitation. Although wave aurora is overall the smallest contributor to the hemispheric precipitation energy budget, during times of increased solar wind driving, it rises to become comparable to the monoenergetic aurora or ion aurora (although still well below the diffuse electron aurora).

Details are in the caption following the image
Broadband acceleration hemispheric energy flux for (a) low and (b) high solar wind driving.

[34] Figure 5 shows the pattern for the electron diffuse aurora, which proves to be the dominant contributor to the global precipitation budget. Because electrons from the nightside plasma sheet E × B convect eastward, the diffuse aurora is more intense postmidnight and into the morning hours. Diffuse auroral precipitation is often relatively insignificant from postnoon through dusk. Although originally thought of as relatively insensitive to geomagnetic activity [cf. Winningham et al., 1975] diffuse precipitation rises with solar wind driving nearly as rapidly as monoenergetic aurora (from 6.8 to 20.2 GW for the cases considered here, a factor of 3.0). This fairly strong dependence on solar wind driving may well be indirect. For example, dipolarizations in the magnetotail can raise electron energies (through conservation of the adiabatic invariant as the electrons move to regions of larger B) and clearly require enhanced dayside merging as a precursor but are not necessarily externally driven.

Details are in the caption following the image
Diffuse aurora hemispheric energy flux for (a) low and (b) high solar wind driving.

[35] Figure 6 shows the ion energy flux for low and high solar wind driving. The ions aurora extends toward the west (dusk), perhaps because ions above a few keV energy drift westward (the curvature and gradient drifts exceed the E × B convection drifts). There is a maximum postmidnight which is more pronounced for active conditions than quiet conditions. Although the DMSP data do not cover the postmidnight sector prior to 1992, Newell et al. [2005] reported that this could not explain the ion maximum postmidnight. A significant advantage of the approach used in this paper is the explicit fit to solar wind conditions, so it is not logical that variations in a yearly average solar wind input could produce the postmidnight maximum. Instead, it is more likely that the presence of diverging electric fields postmidnight [Marklund et al., 1994, 1997] accounts for some or all of the postmidnight ion precipitation maximum. Although the equivalent of monoenergetic acceleration events for ions is rare or possibly nonexistent (we ourselves have searched for candidate events without success, and there seem to be no reports in the literature), it is still reasonable that ion energy fluxes precipitating into the postmidnight region are enhanced by the properly directed electric fields (and retarded premidnight).

Details are in the caption following the image
Ion hemispheric energy flux for (a) low and (b) high solar wind driving.

[36] Table 1 summarizes the energy budget for the various types of aurora. Combined, the electron and ion diffuse aurora, averaged over all conditions, accounts for 77% of all aurora. The combined contribution of electron acceleration events is just 22%, with monoenergetic events somewhat exceeding wave aurora. It is worth noting that the dayside precipitating energy flux clearly is far less responsive to solar wind driving than is the nightside, as is evident from Figures 37. This may seem puzzling at first, since the dayside boundary regions are more intimately interacting with the solar wind than is the nightside. However, as the discussion will make clear, this result fits fairly well with current thinking on the formation of the magnetopause frontside boundary layers.

Details are in the caption following the image
Monoenergetic aurora hemispheric number flux for (a) low and (b) high solar wind driving.
Table 1. Energy Flux Hemispheric Contributions of the Four Auroral Precipitation Types and as a Percentage of the Combined Totala
Low Solar Wind Driving High Solar Wind Driving All Conditions High/Low Ratio
Diffuse ions 2.3 (21%) 4.9 (14%) 3.4 (16%) 2.1
Diffuse electrons 6.8 (63%) 20.2 (57%) 12.6 (61%) 3.0
Monoenergetic 1.1 (10%) 5.8 (15%) 3.3 (16%) 5.3
Broadband 0.6 (6%) 4.8 (13%) 1.5 (6%) 8.0
  • a The combined diffuse precipitation, averaged over all conditions, is 77%. Wave aurora contributes the least but is the most dynamic (responsive to increasing solar wind driving). (The low and high fits remove some aliasing effects likely present in the all conditions data). Contributions are measured in GW.

4. Comparative Number Flux (and Probability) for Four Auroral Precipitation Types

[37] On the basis of the energy flux plots, one would suppose that the nightside was the main region of interest, particularly the dusk-to-midnight sector. The situation is very different when number flux is considered. Ion outflows from the magnetosphere are dependent on soft particle precipitation, since the energy is primarily deposited into the F region, while keV energy electrons deposit their energy in the E region.

[38] Figure 7 shows the number flux input into the ionosphere associated with monoenergetic precipitation events for conditions of low and high solar wind driving. On the basis of number flux, there are three regions of activity, the first being the premidnight region which is so prominent in energy flux, and the other two flanking noon (or, geophysically speaking, flanking the cusp). The afternoon maximum is more pronounced in monoenergetic precipitation events, but a clear morning side peak occurs also. Overall, the dayside is more active than the nightside, particularly for conditions of low solar wind driving. The hemispheric precipitating number flux associated with monoenergetic events rises from 1.1 × 1025 el/s for quiet conditions to 2.3 × 1025 el/s for strong solar wind driving. This factor of 2.1 increase is smaller than that associated with energy flux, because number flux is more associated with the dayside, which changes its precipitation characteristics less dramatically than does the nightside (which dominates energy flux).

[39] As Figure 8 shows, wave aurora, while sharing prenoon and postnoon peaks, actually differs from monoenergetic events. The number flux for wave aurora peaks prenoon, where wave aurora events are extremely common. For quiet conditions, the wave aurora is more common prenoon than postnoon, or even premidnight.

Details are in the caption following the image
Broadband aurora hemispheric number flux for (a) low and (b) high solar wind driving.

[40] Figure 9 shows the number flux for diffuse electrons, under quiet and active conditions. The nightside, and indeed much of the auroral oval, shows a dramatic increase in number flux for active conditions. However, the dayside boundary layers are actually more pronounced for quiet conditions. This result is less surprising than it may first appear, as the discussion in section 5 will show. Perhaps the most striking thing about Figure 9 is the extent to which the dayside dominates the number flux, in clear contrast to precipitating energy flux.

Details are in the caption following the image
Diffuse aurora hemispheric number flux for (a) low and (b) high solar wind driving.

[41] Figure 10 shows the ion number flux under low and high solar wind–driving conditions. The cusp, and to a lesser extent the neighboring boundary regions, dominate the ion number flux, although there is the usual nightside increase for more active solar wind driving. The nightside number flux peak occurs at the same local time as the nightside energy flux peak, postmidnight.

Details are in the caption following the image
Ion aurora for (a) low and (b) high solar wind driving.

[42] Figure 11 shows the probability that a given spectrum exhibits monoenergetic acceleration. This fraction, or probability, is calculated by dividing the number of spectra identified as monoenergetic by the total number of spectra observed, regardless of their characteristics. The plot for monoenergetic aurora under low solar wind driving shows the classic northward auroral oval which has been variously termed “tear drop,” “pear shape,” or “horse collar” [Lassen and Danielsen, 1978; Hones et al., 1989]. The contracted central region void of aurora presumably represents the lesser amount of flux opened under low solar wind driving. Conversely, for high solar wind driving, the oval is larger and more circular. Our experience in looking at individual spectrograms is that the likelihood of observing monoenergetic aurora somewhere along the poleward portion of the oval is close to unity. As Figure 11 shows, more than 20% of all individual spectra in the poleward and dusk region show monoenergetic acceleration.

Details are in the caption following the image
Probability of observing monoenergetic aurora of any intensity for (a) low and (b) high solar wind driving (0.2 means 20% of spectra exhibit monoenergetic peaks).

[43] Figure 12 shows the probability of seeing broadband acceleration spectra. Note that the scale peak has dropped to 0.10 (10%), half that of monoenergetic aurora. The wave aurora shows a strong prenoon-postnoon asymmetry, with events most frequent in the late morning. Various types of transient ionospheric phenomena are known to peak in the morning or late morning sectors, such as some types of high-latitude auroral transients common for northward IMFs [Yahnin et al., 1997; Vorobjev et al., 1999].

Details are in the caption following the image
Probability of observing broadband acceleration for (a) low and (b) high solar wind driving (0.10 = 10% of the spectra have broadband acceleration.)

[44] Table 2 summarizes the hemispheric number flux for all four types of aurora. Number flux is not quite as responsive to solar wind driving as is energy flux, though all types of aurora do experience significant increases. The relative responsiveness of the four types of aurora is the same as four energy flux, however. Thus, number flux increases by a factor of 1.7 for both electron and ion diffuse aurora, by a factor of 2.0 for monoenergetic aurora, and by a factor of 3.0 for broadband aurora.

Table 2. Number Flux Hemispheric Contributions of the Four Auroral Precipitation Types and as a Percentage of the Total Number of Precipitating Particlesa
Low Solar Wind Driving High Solar Wind Driving All Conditions High/Low Ratio
Diffuse ions 2.4 × 1024 (5%) 4.1 × 1024 (4%) 3.1 × 1024 (4%) 1.7
Diffuse electrons 3.2 × 1025 (60%) 5.4 × 1025 (48%) 4.1 × 1025 (55%) 1.7
Monoenergetic 1.1 × 1025 (21%) 2.3 × 1025 (21%) 1.6 × 1025 (21%) 2.1
Broadband 7.6 × 1024 (14%) 3.1 × 1025 (28%) 1.5 × 1025 (20%) 4.1
  • a Particles are of all types and species. Number flux is less dynamic than energy flux, with broadband acceleration as the most dynamic. (All conditions data are subject to aliasing effects which are removed by the high and low fitted data.) Contributions are measured in particles/s.

5. Discussion

5.1. Dayside Domination of Number Flux and Nightside Domination of Energy Flux

[45] It is only a slight oversimplification to state that the number flux precipitating into the ionosphere is dominated by the dayside boundary regions, while energy flux is dominated by the nightside, particularly premidnight where substorms typically originate. This suggests particles enter the magnetosphere preferentially on the frontside through the cusp and various boundary layers and are energized by the magnetotail. Of particular interest is the fact that the number flux dominance of the dayside is even more pronounced for low solar wind driving (which implies northward IMF conditions). There are good reasons why this should be so.

[46] High-altitude observations indicate that the low-latitude boundary layer (LLBL) is thicker for northward IMF than southward [Mitchell et al., 1987; Hasegawa et al., 2004]; indeed, the LLBL is sometimes absent altogether for strongly southward IMF [Eastman et al., 1996]. This is probably true in part because for southward IMF, merging is more active all along the frontside magnetopause, stripping away the boundary layers [cf. Newell et al., 1997]. However, frontside merging generally involves only modest acceleration of magnetosheath particles which are significantly colder than is typical of the magnetotail. By contrast, merging and dipolarization can accelerate particles to higher temperatures in the magnetotail, creating a hotter diffuse aurora. Discrete auroras are also more intense on the nightside, for quite a distinct reason. UV insolation reduces auroral acceleration [Newell et al., 1996a, 1998a]; indeed, the largest accelerations occur in deep plasma density cavities. It is also likely that stronger and more transient field-aligned currents driven by magnetotail merging and other depolarization processes contribute to these larger accelerating potentials on the nightside. (The dayside currents, particularly around 1500 MLT, are, however, more consistently present.)

[47] It is the dayside boundary layers, with their copious fluxes of low-energy electrons, which lead to the dayside dominance of ion conics [Øieroset et al., 1999] flowing out of the heated F region. By contrast, higher energy and more field-aligned ion beams are more typical of the larger acceleration energies present in the nightside oval.

5.2. Contribution of Broadband Acceleration (Wave Aurora)

[48] Wave aurora is energetically insignificant for quiet conditions. However, it is the most dynamic of the four types of aurora, by far, and may be significant energetically for active conditions. For example, if we extrapolate our model to dΦMP/dt to a few times 〈dΦMP/dt〉, wave aurora exceeds monoenergetic aurora in energy flux, although it is hard to judge the validity of making such an extrapolation.

[49] Moreover, the number flux contribution from broadband acceleration can be highly significant for active conditions, easily exceeding monoenergetic aurora even at modest activity levels (well within the range in which our data set is amply sampled). Theories in which ionospheric outflows or magnetotail mass loading affects substorm development are probably best referred to the wave aurora, with its highly dynamic response and high number fluxes. This is even more the case when one considers that the lower average energies associated with broadband aurora mean that more of the energy deposition will be within the F layer.

[50] It was recently demonstrated in some detail by Chaston et al. [2005] that surface waves on the magnetopause can couple into the ionosphere, specifically in the form of broadband acceleration and committant wave aurora. The high frequency of broadband acceleration in the late morning (compare Figure 12, both quiet and active conditions) may thus relate to this effect. Ground-based observations of various ionospheric transients often observe a morning or prenoon peak. For example, Yahnin et al. [1997] and Vorobjev et al. [1999] have reported that high-latitude auroral transients observed for northward IMF on the dayside frequently occur in the late morning sector. It is certainly tempting to associate these auroral transients with surface wave excitations on the magnetopause surface, according to the coupling mechanism outlined by Chaston et al. [2005].

[51] In Figure 4b, an irregular ring of broadband acceleration around 65° MLAT is observable. Most reports in the literature describe broadband acceleration as occurring in the poleward portion of the oval. Although it is true that the few spectra at the poleward edge of the evening auroral oval often show broadband acceleration, our experience is that signatures can be seen throughout the auroral oval (for example, we hand examined a large number of events in the study reported by Newell [2000]). Shiokawa et al. [1997] have reported instances of broadband acceleration occurring at low-latitudes during geomagnetic storms. There are not enough storms to account for the low-latitude ring in Figure 4b. Nonetheless, the events are real and relatively common.

[52] Figure 13 illustrates one of the low-latitude (meaning around 65° or so, thus the equatorward portion of the auroral oval) broadband events. Like most such, it does not occur during a storm or even a particularly active time. Cases can be found all around the auroral oval, and the subject clearly deserves detailed study at some future date. The events we have examined, on the basis of a random print out from the data processing, are consistently real, not noise.

Details are in the caption following the image
(top) A spectrogram showing an instance flagged as having broadband acceleration at relatively low latitude. (bottom) The spectrum marked by the arrow, which does indeed show broadband acceleration (to 300 eV).

5.3. Cusp (and Double Cusp)

[53] Newell and Meng [1988] showed that the cusp differs from the boundary layers more dramatically in the ion signatures than in the electron signatures. Specifically, the region with more magnetosheath like ion energies (a spectral flux peak at or below 3000 eV) was more concentrated near noon and had much higher (and hence more magnetosheathlike) fluxes. Distinctions between the electron cusp population (spectral flux peak below 200 eV) and the boundary layers are more gradual, with no clear break between the populations (see the figures of Newell and Meng [1988] for details).

[54] Figure 10, which showed the ion number flux under low and active conditions does indeed show the near-noon cusp as distinct from the boundary layers much better than does the diffuse electron precipitation. Whereas the boundary layers extend for several hours on either side of noon, the region of intense ion number fluxes is more concentrated.

[55] Of particular interest in the “double cusp” signature which forms in the ion number flux plot for active conditions. The double cusp was first introduced by Wing et al. [2001] in case studies. This phenomenon was explained as occurring when merging is ongoing simultaneously at high and low latitudes at the magnetopause. Newell et al. [2007b] demonstrated that the double cusp occurs overwhelmingly for conditions of high solar wind driving (typically strongly southward IMF). In Figure 10b, which is for a high magnetopause merging rate, one can see what appears to be a statistical double cusp. If so, the higher latitude peak near noon (centered about 80° MLAT) is due to merging just below the magnetic cusp, while the lower latitude peak (about 78° MLAT) is due to merging near the subsolar point.

6. Summary and Conclusions

[56] Hemispheric precipitating energy flux is dominated by the diffuse aurora, that is, by the unaccelerated meanderings of ions and especially electrons in the magnetosphere. The contribution of the diffuse aurora is 84% of the total 10.8 GW precipitating under conditions of low activity (dΦMP/dt ≤ 0.25〈dΦMP/dt〉), while even under conditions of moderately high activity (≥1.5〈dΦMP/dt〉) 71% of all precipitating energy flux is diffuse (35.7 GW total, 25.1 GW diffuse). The diffuse aurora is not quiescent but rises with increasing solar wind driving (by a factor of 3 from our low to high conditions), albeit not quite as fast as the discrete aurora. Thus, the least glamorous and arguably least studied form of auroral precipitation is the most energetically important. Indeed, our simple assumption of isotropy for all forms of precipitation possibly slightly understates the importance of diffuse aurora, since the accelerated aurora is more likely to be field-aligned (DMSP measurements are all within the loss cone).

[57] Monoenergetic aurora is the energetically more significant type of discrete aurora, rising from 10% of the hemispheric budget under low solar wind driving to 15% for moderately active. Broadband acceleration, associated with wave aurora, is relatively energetically unimportant for quiet conditions (6% of the total) but rises to 13% for active conditions. Because of its far greater responsiveness to solar wind driving, broadband aurora becomes comparable to monoenergetic aurora as activity rises. Moreover, broadband aurora contributes more to the hemispheric number flux under active conditions (28%) than does anything but the diffuse electron aurora (48%). This suggests that it is the wave aurora, with its copious flux of somewhat softer electrons, which is driving ionospheric ion outflows, at least during geomagnetically active times. No other type of aurora contributes such high energy fluxes below 1 keV.

[58] All types of aurora precipitate more energy flux into the nightside than dayside since plasma is heated in the magnetotail and since auroral density cavities on the nightside permit greater acceleration of discrete aurora. However, number flux is actually higher on the dayside, especially in less active conditions, where the dayside boundary layers lead to copious quantities of relatively low energy precipitation. The very high number fluxes associated with the dayside boundary layers certainly suggest that they are the dominant means of supplying magnetospheric plasma.

[59] The precipitation model here has several advantages over its predecessors, in addition to the larger database. Instead of using a handful of activity levels with large discontinuous jumps between, we have a regression fit to the solar wind function which best predicts auroral power. The separate categorization of each type of aurora allows each independent solar wind variability; indeed, each type of aurora in each MLT-MLAT bin varies independently. This approach is promising both for better empirical representation of global precipitation and as a research tool, as the first results here suggest.

Acknowledgments

[60] This work was supported by NSF grants ATM-0741344, ANT-0738055, and ATM-0802708 to the Johns Hopkins University. J. King, R. Lepping, and N. Papatashvilli of NSSDC at NASA Goddard provided the solar wind and IMF data.

[61] Wolfgang Baumjohann thanks Gerhard Haerendel and another reviewer for their assistance in evaluating this paper.