The causes of the hardest electron precipitation ... - Wiley Online Library

Journal of Geophysical Research: Space Physics RESEARCH ARTICLE 10.1002/2016JA022346

Special Section: Energetic Electron Loss and its Impacts on the Atmosphere

The causes of the hardest electron precipitation events seen with SAMPEX David M. Smith1 , Eric P. Casavant1,2 , Max D. Comess1,3 , Xinqing Liang1 , Gregory S. Bowers1 , Richard S. Selesnick4 , Lasse B. N. Clausen5 , Robyn M. Millan6 , and John G. Sample7 1 1Physics Department and Santa Cruz Institute for Particle Physics, University of California, Santa Cruz, California, USA,

Key Points: • Hard e- precipitation (folding energy >500 keV) has two populations: nearer dusk, with harder spectra, and nearer midnight, with softer spectra • The harder population shows evidence of being due to EMIC wave scattering, while the softer may be due to current sheet scattering • Concurrent proton precipitation does not necessarily help distinguish the two mechanisms

Correspondence to: D. M. Smith, [email protected]

Citation: Smith, D. M., E. P. Casavant, M. D. Comess, X. Liang, G. S. Bowers, R. S. Selesnick, L. B. N.Clausen, R. M. Millan, and J. G. Sample (2016), The causes of the hardest electron precipitation events seen with SAMPEX, J. Geophys. Res. Space Physics, 121, 8600–8613, doi:10.1002/2016JA022346.

Received 6 FEB 2016 Accepted 30 AUG 2016 Accepted article online 6 SEP 2016 Published online 22 SEP 2016

©2016. American Geophysical Union. All Rights Reserved.

SMITH ET AL.

2 FlexStr8 Corporation, El Segundo, California, USA, 3 SpaceX Corporation, Hawthorne, California, USA, 4 Space Vehicles

Directorate, Air Force Research Laboratory, Kirtland Air Force Base, New Mexico, USA, 5 Department of Physics, University of Oslo, Oslo, Norway, 6 Department of Physics and Astronomy, Dartmouth College, Hanover, New Hampshire, USA, 7 Department of Physics, Montana State University, Bozeman, Montana, USA

Abstract We studied the geomagnetic, plasmaspheric, and solar wind context of relativistic electron precipitation (REP) events seen with the Solar, Anomalous, and Magnetospheric Particle Explorer (SAMPEX), Proton Electron Telescope (PET) to derive an exponential folding energy E0 for each event. Events with E0 < 400 keV peak near midnight, and with increasing E0 , the peak magnetic local time (MLT) moves earlier but never peaks as early as the MLT distribution of electromagnetic ion cyclotron (EMIC) waves in the outer belt, and a distinct component near midnight remains. Events with E0 > 750 keV near dusk (1400 < MLT < 2000) show correlations with solar wind dynamic pressure and proton density, AE index, negative Dst index, and an extended plasmasphere, all supporting an EMIC wave interpretation. Events with 500 keV < E0 < 600 keV near midnight (MLT 2200–0200) do not show these correlations. Comparing these two samples to all events with E0 > 500 keV (“hard REP”), we estimate that roughly 45% of the whole population has the distributions of geomagnetic and solar wind parameters associated with EMIC waves, while 55% does not. We hypothesize that the latter events may be caused by current sheet scattering (CSS), which can be mistaken for EMIC wave scattering in that both simultaneously precipitate MeV electrons and keV protons. Since a large number of MeV electrons are lost in the near-midnight hard REP events, and in the large number of E0 < 400 keV events that show no dusk-like peak at all, we conclude that CSS should be studied further as a possibly important loss channel for MeV electrons.

1. Introduction 1.1. Relativistic Electron Precipitation and Electromagnetic Ion Cyclotron Waves A subset of relativistic electron precipitation (REP) events from the outer belt with particularly hard spectra (e-folding energies E0 on the order of 500 keV) has been identified for many years from satellite data [Vampola, 1971; Imhof et al., 1986] and balloon data [Foat et al., 1998; Lorentzen et al., 2000; Millan et al., 2002, 2007; Li et al., 2014]. From the very beginning [Thorne and Kennel, 1971], such events were predicted to take place when electromagnetic ion cyclotron (EMIC) waves powered by a ring current ion anisotropy resonantly scatter the most relativistic end of the trapped electron population. These hardest REP events have been called “duskside REP” [Millan and Thorne, 2007] or DREP due to their concentration in the dusk-to-midnight sector of magnetic local time (MLT). This local time range has been described as corresponding to the nightward edge of the dusk plasmaspheric bulge, where protons injected from the tail might first encounter higher plasma densities and generate EMIC waves. In order to open the discussion up to include the hardest precipitation events regardless of their MLT, we will use the more general term “hard REP” for events with significant precipitation >1 MeV and E0 > 500 keV or thereabouts (when that parameter is available). There has been a wide range of direct experimental evidence supporting the hard REP/EMIC connection. Rodger et al. [2008] on four occasions found similar waves in concert with ground-based measurements of atmospheric ionization due to precipitation. Yuan et al. [2013], in a case study of the geomagnetic storm of 8–9 May 2001, found simultaneous ring current proton and relativistic electron precipitation within a plasmaspheric plume. Other studies have simultaneously made direct measurements of EMIC waves from the ground along with simultaneously precipitating ring current protons and relativistic electrons with the SAMPEX ELECTRON PRECIPITATION

8600

Journal of Geophysical Research: Space Physics

10.1002/2016JA022346

Polar-orbiting Operational Environmental Satellites (POES)-17, including several case studies [Miyoshi et al., 2008; Rodger et al., 2015; Clilverd et al., 2015] and a large-scale systematic investigation that included an important control for the chances of accidental detection of EMIC waves alongside an unrelated precipitation event [Hendry et al., 2016]. Carson et al. [2013] and Wang et al. [2014] used a large sample of POES events to search for coincidences between 52 keV proton precipitation and >800 keV electron precipitation, finding that these events concentrate in the dusk-to-midnight sector, echoing an earlier result by Imhof et al. [1986] using multiple satellites; these surveys did not measure EMIC waves. Two recent studies made use of precipitation data from the Balloon Array for RBSP Relativistic Electron Precipitation (BARREL). Li et al. [2014] not only found EMIC waves colocated with a hard REP event, they simulated pitch angle diffusion for the particular conditions at the time and found agreement with the intensity and spectrum of the precipitation. L. W. Blum et al. [2015] presented two more BARREL hard REP events coincident with EMIC wave activity. One challenge for the EMIC model for hard REP has been meeting the resonance condition for the electrons; in most published models [Thorne and Kennel, 1971; Li et al., 2013], electrons below resonance are not scattered. A relatively low minimum electron resonant energy depends on a high plasma density and/or a low magnetic field, meaning that the duskside plasmaspheric bulge, or even better, plumes or other local density enhancements at higher L, are the most likely regions for resonance, but “relatively low” generally still means 1 MeV or higher. Li et al. [2014] conducted a full simulation of pitch angle diffusion based on the measured electron spectrum, plasma density (100 el cm−3 ), EMIC wave spectrum, and magnetic field corresponding to their BARREL hard REP event and found that the expected precipitating electron spectrum peaked at 1.2 MeV and fell off rapidly below that—implying an even faster falloff of the scattering efficiency, since the trapped population being acted upon has a falling spectrum with energy. In a broader study, Meredith et al. [2003] calculated electron minimum resonant energies for 800 detections of EMIC waves by the Combined Release and Radiation Effects Satellite (CRRES) at L values across the outer belt. The most common calculated resonant energies were around 4 MeV, with only a very small fraction (about 3%) falling below 1 MeV and none below 600 keV. But hard REP electrons are seen at energies below 1 MeV in several of the satellite and balloon observations cited above. In Comess et al. [2013] and in further analysis of data from the Proton Electron Telescope (PET) on the Solar, Anomalous, and Magnetospheric Particle Explorer (SAMPEX), we find no evidence of a cutoff or turnover in the spectrum down to 400 keV. Yahnin et al. [2014] found that even those hard REP events that simultaneously precipitate ring current ions, supposedly the EMIC fingerprint, also precipitate subrelativistic (> 30 keV) electrons that should not meet the EMIC resonance condition. The event studied by Clilverd et al. [2015] and connected to direct EMIC wave measurements had a notably low lower cutoff energy of 280 keV and was attributed by the authors to a nonresonant class of EMIC wave interaction. A second challenge to the EMIC interpretation, less often discussed, has been the poor agreement between the local time distributions of EMIC waves and relativistic precipitation. The population studies of the most relativistic precipitation events show them distributed from roughly 18–24 h in MLT [Imhof et al., 1986; Comess et al., 2013; Carson et al., 2013; Wang et al., 2014]. In contrast, helium band EMIC waves observed by GOES in geosynchronous orbit [Clausen et al., 2011] and by satellites that cover values of L across the outer belt [Anderson et al., 1992; Meredith et al., 2003; Halford et al., 2010; Min et al., 2012; Meredith et al., 2014; Allen et al., 2015] peak earlier in the afternoon sector, at MLT of roughly 13–18 h. This holds true at solar minimum [Clausen et al., 2011] and solar maximum [Meredith et al., 2014], across the range of values of AE [Meredith et al., 2014], and regardless of wave intensity [Meredith et al., 2014]. Hydrogen band waves, which are less common, peak even earlier (13–15 h) [Meredith et al., 2014]. Usanova et al. [2013] found that even when studying EMIC waves specifically chosen to be within plasmaspheric plumes, the wave occurrence distribution peaked in the early afternoon despite the occurrence rate of the plumes themselves peaking in the premidnight hours. It seems likely that the case studies in which EMIC waves were measured in conjunction with relativistic precipitation [e.g., Miyoshi et al., 2008; Rodger et al., 2008; Li et al., 2014; L. W. Blum et al., 2015] really do represent a causal relationship, since EMIC waves are not ubiquitous; this conclusion has been greatly strengthened by the statistical study, with control cases, by Hendry et al. [2016]. But due to the mismatch of the MLT range of relativistic precipitation seen in larger statistical studies and that of typical EMIC waves, we follow Yahnin et al. [2014] in suggesting that some statistical studies may be combining true cases of EMIC-driven precipitation with precipitation driven by other mechanisms. SMITH ET AL.

SAMPEX ELECTRON PRECIPITATION

8601


10.1002/2016JA022346

1.2. Current Sheet Scattering The pitch angle distribution of electrons measured at the outer edge of the outer belt at the tail is often isotropic, meaning there is pitch angle scattering and loss there, both at quiet and active times [Imhof et al., 1977; West et al., 1978; Imhof et al., 1979; Imhof , 1988; Lee et al., 2006]. The L shell of isotropy increases with decreasing energy. This is predicted if the scattering is due to the loss of the first adiabatic invariant when the cyclotron radius of the electron approaches the (relatively small) radius of curvature of the field as it crosses the current sheet: in a stretched tail, this radius is actually smaller at higher L and can scatter down to lower electron energies [Sergeev and Tsyganenko, 1982]. This phenomenon has been called current sheet or plasma sheet scattering, field line curvature scattering, or 𝜇 scattering. The simultaneous precipitation of ring current (keV) ions with MeV electrons has been considered a signature of EMIC-driven precipitation [Imhof et al., 1986; Yuan et al., 2013; Carson et al., 2013; Wang et al., 2014] and has been observed directly along with the EMIC waves themselves on at least one occasion [Miyoshi et al., 2008]. But due to the mismatch in the MLT distributions of EMIC waves and relativistic precipitation cited above, we explore the possibility that current sheet scattering could produce a similar sort of joint electron/proton precipitation. Indeed, such joint precipitation at the trapping boundary has been recognized and interpreted as related to current sheet scattering by Imhof [1988]. All that is needed for 𝜇 scattering to take place is for the particle’s gyroradius to approach the field line curvature; protons and electrons of the same gyroradius should therefore be able to scatter together. This mechanism has been suggested as the dominant loss mechanism for both energetic ring current protons [Sergeev et al., 1983] and auroral protons [Gvozdevskij and Sergeev, 1995]. The cyclotron radius of a charged particle in the relativistic case is RC =

𝛾𝛽mc2 sin 𝛼 qB

(1)

where m is the particle mass, 𝛼 the pitch angle, q the particle charge, and B the magnetic field strength. The proton kinetic energy (nonrelativistic; Ep = 12 mp vp2 ) giving the same RC as a relativistic electron of energy Ee = (𝛾 − 1)mc2 is therefore 2 m 1 Ee Ep = Ee e + . (2) mp 2 mp c 2 So, for example, the gyroradius of an electron of 2 MeV corresponds to that of a proton of 3.2 keV and that of an electron of 5 MeV corresponds to that of a proton of 16 keV, with the correspondences being independent of pitch angle. The corresponding proton energies here are very similar to those for the same electron energies for the cases of EMIC wave scattering shown in Figure 11 of Meredith et al. [2003]. Bearing in mind that this is not a resonance, but rather a floor—with more energetic particles being easier to scatter—we suggest that the presence of ring current protons being precipitated alongside relativistic electrons [Imhof et al., 1986; Yuan et al., 2013; Carson et al., 2013; Wang et al., 2014] does not necessarily favor the EMIC hypothesis over current sheet scattering; the measurement of the waves in addition is also necessary to make the connection conclusive. But a separation of hard REP events by precipitation mechanism has been attempted even in the absence of EMIC wave measurements. Yahnin et al. [2014] provide an intriguing analysis of precipitation seen in POES that divides REP events (>700 keV electrons) into three classes based on the presence of protons and the position of the precipitation relative to the outer trapping boundary. They associate 31% of their REP events with current sheet scattering, these being associated with proton precipitation and taking place near the outer trapping boundary. They associate only 13% with the EMIC mechanism, based on having simultaneous proton precipitation but being well equatorward of the outer trapping boundary. The rest of the events, which show no proton precipitation, they associate with other wave-particle interactions. But they also note that even their small class of EMIC candidates shows precipitation of subrelativistic (> 30 keV) electrons that is difficult to explain with EMIC waves alone. Sandanger et al. [2007] used observations of proton anisotropy—thought to be the driver of EMIC waves—along with cospatial proton precipitation to identify REP events as EMIC driven. They also found occasions of proton precipitation with isotropy at higher magnetic latitudes that they identified with current sheet scattering. These observations took place in the long-lived recovery phase of a single geomagnetic storm. SMITH ET AL.


8602


10.1002/2016JA022346

1.3. Scope of This Work Here we take a closer look at the sample of hard REP we previously extracted from the SAMPEX database [Comess et al., 2013]. We look for indirect evidence that this sample may have multiple causes, rather than being due entirely to EMIC waves. Unlike authors using POES data [Carson et al., 2013; Wang et al., 2014; Yahnin et al., 2014], we do not have access to data on proton precipitation for the full time period selected in this study, but as we have shown above, and as was discussed by Yahnin et al. [2014], this does not distinguish automatically between EMIC wave scattering and CSS. Instead, we examine the local time distribution of precipitation as a function of the electron spectral index derived from SAMPEX/PET. We hypothesize that events that are both harder and closer to dusk are more likely to be EMIC driven, while those softer and closer to midnight are more likely to be due to another mechanism, possibly CSS. To test this hypothesis, we first show that these two characteristics (spectral index and local time) are indeed correlated in the correct sense and then relate the harder and more duskward population, but not the softer population near midnight, to higher solar wind dynamic pressure, higher solar wind proton density, higher AE index, more disturbed Dst, and a more extended plasmasphere at the time of precipitation, all parameters that should favor EMIC wave occurrence but not necessarily CSS. We use data from the Extreme Ultraviolet Imager (EUV) on the Imager for Magnetopause-to-Aurora Global Exploration (IMAGE) satellite to examine the plasmasphere during a subset of our hard REP events when these data are available.

2. SAMPEX Observations 2.1. Data Set SAMPEX flew in low-Earth, polar orbit from 1992 to 2012. The set of precipitation events we have extracted [Comess et al., 2013] includes data from 1992 to 2004. We used The Heavy Ion Large Area Telescope (HILT) [Klecker et al., 1993] to identify outer-belt precipitation events that include >1 MeV electrons and PET [Cook et al., 1993] to perform spectroscopy on these events in the range from 0.5 to 6 MeV and extract E0 (the folding energy of an exponentially decaying model of the spectrum) as a parameter of spectral hardness. Note that all events that can be captured by the HILT trigger, which gave us events with E0 between about 0.1 and 1.5 MeV, would be considered “hard” relative to most precipitation events detected by balloon instruments via their atmospheric bremsstrahlung emission; the latter usually have e-folding energies of only tens of keV [e.g. Smith et al., 1995]. The selection and processing of SAMPEX data are described in Comess et al. [2013] and in even more detail in Comess [2011]. We selected only data taken in the north Atlantic, a zone in which the magnetic conjugate point to the spacecraft is within the atmosphere (1 MeV electrons) in each satellite pass under the outer belt in the bounce loss region as long as there is any peak significantly above the background noise. Multiple peaks are summed only when they overlap significantly (do not return to near background between events). The e-folding energy E0 from the PET spectra is obtained by a fitting method [Selesnick, 1993; Mewaldt et al., 2005] that produces a spectral model that is constrained to be smooth but does not have a fixed functional form. The values of the model in the highest and lowest energy bins are used to derive E0 . The smoothness constraint of the fitting algorithm assures that this method does not give a result that is highly dependent on the large error bar of the highest spectral bin. In Comess et al. [2013] we showed that the population of “hard REP,” those with E0 > 0.5 MeV, occurred almost exclusively between 15 and 01 h MLT; this range is similar to what was reported by the POES studies [Carson et al., 2013; Wang et al., 2014]. 2.2. MLT Distributions of Hard REP and EMIC Waves In the first part of our analysis, we make further use than in Comess et al. [2013] of the precipitation spectral index E0 , as being the measurement that most distinguishes our SAMPEX data set from POES. Figure 1 shows the MLT distribution of hard REP events with folding energies in the bands 400–500, 500–600, 600–700, and >700 keV, with the softer REP events (E0 < 400 keV) and the MLT distribution of EMIC waves at geosynchronous orbit [Clausen et al., 2011] shown for comparison. The SAMPEX distributions are corrected for varying exposure with local time [Comess et al., 2013]. Because SAMPEX cycled rapidly and regularly through MLT over the 13 years in the data set, there is no bias toward particular geomagnetic conditions in any range of local time. SMITH ET AL.


8603


10.1002/2016JA022346

For E0 < 400 keV, the distribution is symmetrical around the noon-midnight line; within that broad band of e-folding energies, the noon peak is dominant for E0 < 200 keV and the midnight peak for E0 > 200 keV. Episodes of relativistic microbursts (defined as per Comess et al. [2013]) have been removed from these samples, but they are relatively small in number. Their MLT distribution is shown as the dotted black line in Figure 1, Figure 1. MLT distributions of nonmicroburst REP events as a function of and this dawnside-dominated distrithe range of e-folding energies E0 (solid colored lines), of microburst bution is similar to previous SAMPEX events (dotted line), and of EMIC waves at geosynchronous orbit observed with GOES 10, 11, and 12 (dashed line) [see Clausen et al., 2011, results [Nakamura et al., 2000; O’Brien Figure 4]. et al., 2003]. MLT is calculated using the Tsyganenko “T89c” model, which includes an external field component depending on Kp [Tsyganenko, 1989], to trace the field line to the magnetic equator, where MLT is calculated, but the external field component changes MLT only very slightly relative to what is obtained using the International Geomagnetic Reference Field (IGRF) [Langel, 1992]. Field line tracing under both models is performed using the International Radiation Belt Environment Modeling library (http://virbo.org/IRBEM, accessed 3 December 2015). As E0 marches upward from 400 keV to over 700 keV, the local time peak moves away from midnight and toward dusk; however, even for the hardest class of SAMPEX events, the peak is centered 4 h later than the peak of EMIC activity. The simplest explanation of this is that the population of hard REP described in recent surveys as extending from dusk to midnight [Carson et al., 2013; Comess et al., 2013; Wang et al., 2014] is made up of two components: an EMIC-driven component near dusk with harder spectra and a CSS-driven component near midnight with softer spectra. We will test this idea in the sections below. We expect CSS-driven spectra to be softer than EMIC-driven spectra for two reasons. First, there is no limit on how low the energy of electrons engaging in CSS can be as long as there are magnetic inhomogeneities on the scale of the gyroradius, while there is generally a lower energy resonance threshold for EMIC wave scattering, and second, the peak at midnight remains distinct and appears without any dusk-like peak for E0 < 400 keV. But the peak near midnight in the E0 < 400 keV band and the peak near 1600 MLT in the EMIC waves are too narrow and too well separated to simply fit each of the high E0 bands in Figure 1 to a weighted sum of those two distributions. Some other effect is needed to fill in the middle (around 2100 MLT), and at this stage we only speculate. One possibility is that the EMIC waves responsible for some hard REP are a subset that for some reason are biased to later MLT than the bulk of either H or He band waves. Another possibility is that in the CSS mechanism, the highest-energy electrons, as they drift toward midnight from the afternoon sector, encounter sufficient magnetic curvature to precipitate them earlier than the lower energy electrons, which have to wait until true midnight to encounter sufficient curvature.

3. Solar Wind and Geomagnetic Indices We next examine whether the geomagnetic and solar wind environment can help us distinguish between hard REP (E0 > 500 keV) events that may originate through EMIC waves and CSS. Usanova et al. [2012] found that EMIC waves beyond geosynchronous orbit seen by the THEMIS correlated strongly with solar wind dynamic pressure (Pdyn ) and with the AE and SYMH geomagnetic indices. Clausen et al. [2011] found similar strong correlations for EMIC waves at geosynchronous orbit for Pdyn and solar wind proton density Np , with a less dramatic effect for AE and no apparent connection with Dst, which is comparable to SYMH but sampled at a slower (1 h) cadence. We start with the hypothesis that among our hard REP events (E0 > 500 keV), those with both the hardest e-folding energy and local times closest to dusk are most likely to be related to EMIC waves, while those with softer e-folding (closer to 500 keV) and closest to midnight are most likely to be related to CSS. Figure 2 shows a scatterplot of the events in our database with MLT and E0 . The data points in red, our preferred candidates to be EMIC driven, are those with E0 > 750 keV and 1400 < MLT < 2000. The data points in blue, considered most likely to be from CSS, have E0 < 600 keV and MLT between 2200 and 0200. The local time ranges for each SMITH ET AL.


8604


10.1002/2016JA022346

category were chosen based on the arguments in the previous section; the minimum E0 for the red sample and maximum E0 for the blue sample were designed to give a sample size of about 100 events in each class. The precise numbers are 113 in the red sample, 117 in the blue sample, and 818 in the remaining (black) data sample, which we hypothesize as being a mix of EMIC-driven and CSS-driven events. In this section and the next, we test our hypothesis by seeing if events in the red sample will show characterisFigure 2. Scatterplot of MLT and E0 for the 1048 hard REP events tics that correlate with an EMIC origin, (E0 >500 keV) in the SAMPEX data sample. The regions we hypothesize as while those in the blue sample will most likely to be EMIC-driven and most likely to be CSS-driven are populated by red and blue points, respectively. not. Figure 3 shows the distributions of Pdyn , Dst, AE , and Np for all times covered by our SAMPEX data set (1992–2004) and for the three classes of event shown in Figure 2, shown as solid lines in the matching colors. These parameters were retrieved from the OMNIWeb Plus service of the Space Physics Data Facility at NASA’s Goddard Space Flight Center. The events in the red sample (hypothesized to be EMIC wave driven) skew to higher Pdyn , higher AE , more negative Dst, and higher Np than either the random times or the other two samples of hard REP. The blue

Figure 3. Distributions of the frequency of occurrence of four context parameters: solar wind dynamic pressure (Pdyn ) and proton density (Np ) and geomagnetic indices Dst and AE . Red: events in the sample hypothesized to be due to EMIC waves. Blue: those hypothesized to be due to CSS. Black: the remaining hard REP events. Dashed line: all times during our period of SAMPEX coverage from 1992 to 2004.

SMITH ET AL.


8605


10.1002/2016JA022346

Figure 4. Maximum likelihood fits of the black distributions (unclassified hard REP events) from Figure 3 to weighted sums of the distributions of the red (EMIC) sample and blue (CSS) sample. Green: individual fits using each parameter distribution alone. Orange: joint fit using all four parameter distributions at once.

sample (hypothesized to be CSS driven) peaks at lower Pdyn even than the randomly selected times. With all four parameters, their distribution for the black sample (intermediate values of E0 or MLT or both) can be matched very well by a weighted sum of the distributions for the blue and red samples. Using a maximum likelihood analysis, with the possible models being all possible weighted sums of the red and blue distributions and the data being the black distribution, we found that the highest-likelihood percentage contribution of the red distribution (EMIC driven) to the black distribution is similar for all four parameters: 46%, 54%, 38%, and 44% for Pdyn , Dst, AE , and Np , respectively. Figure 4 shows these optimum weighted curves in green in comparison to the black sample. The similarity of the contributions in each individual fit lends support to the underlying assumption that there are two distinct populations making up the overall hard REP sample. When doing a joint likelihood maximization over all the histograms of all four parameters, the best value for the fraction of the black sample contributed by the red sample is 44%. Under the assumption that all 113 events in the red sample are EMIC driven, all 117 events in the blue sample are CSS driven, and 44% of the events in the black sample are EMIC driven, the total fraction of the SAMPEX hard REP set that is EMIC driven would be 45%. While this is no doubt an indirect approach, it is still intriguing that it is completely independent of the results of Figure 1, yet suggests a qualitatively similar conclusion, in that the MLT range of hard REP in Figure 1 is roughly halfway between the peak of EMIC waves and midnight, where soft REP have a peak.

4. IMAGE/EUV Observations 4.1. Data Set EMIC-driven precipitation is expected to occur only when significant plasma density (the plasmasphere itself, a plume, or smaller structures) extends to L shells in the outer radiation belt. Lorentzen et al. [2000] showed that a cold plasma density of only 10 el cm−3 is sufficient to bring 1.7 MeV electrons into resonance with typical EMIC waves and that the reduction of the resonant energy with density is very small beyond that, its value depending more on other parameters [see also Denton et al., 2015]. Usanova et al. [2013] used data from the Cluster spacecraft to show that the likelihood of EMIC waves is ∼20 times higher inside than just outside a SMITH ET AL.


8606


10.1002/2016JA022346

plume, defined as a local density enhancement of > 10 el cm−3 , demonstrating that wave generation, as well as electron resonance, typically prefers this level of plasma density. IMAGE/EUV provided images of the entire plasmasphere in the 30.4 nm line of sunlight resonantly scattered by cold He+ [Sandel et al., 2000]. There are four limitations in the direct use of the IMAGE/EUV data as a test of sufficient plasma density to drive EMIC waves causing precipitation seen at SAMPEX. First, the EMIC density threshold of ∼10 el cm−3 in total (not He+ ) plasma density is about a factor of 4 lower than the lowest level of total plasma density that IMAGE/EUV can reliably detect, (40 ± 10) el cm−3 , as estimated by Goldstein et al. [2003] using comparisons of IMAGE/EUV maps with in situ plasma data. Second, it is possible that smaller density enhancements exist than can be resolved with IMAGE/EUV. Third, the relation between He+ and total plasma density is not a constant but can be a function of both radius and how recently the plasmasphere has been disturbed, although Goldstein et al. [2003] found that when a sharp plasmapause is present it is collocated for the two species. Finally, the mapping of a single magnetic field line from SAMPEX in low-Earth orbit to the equator is uncertain (mostly in L rather than MLT), particularly near midnight, where many of our events take place, and in disturbed geomagnetic conditions. Nonetheless, IMAGE/EUV can give us an overview of the state of the plasmasphere at the time of any given event, and it is primarily for this qualitative purpose that we use the data here. We expect that the plasmasphere will be more extended overall at times of EMIC-driven precipitation than at other times and that no such extension is necessarily expected at times of CSS-driven precipitation. We followed the analysis procedures of the EUV team to generate images of the plasmasphere at times of hard REP observed by SAMPEX [see Gallagher et al., 2005, Appendix A]. The raw EUV images were corrected for projection considering the angle of the spacecraft position vector relative to the equatorial plane, using the program xform.pro (http://euv.lpl.arizona.edu/euv/software/xform.html, accessed 23 June 2016). Image quality was determined by screening out times of very high background automatically, followed by a visual search for severe discontinuities in the image resulting from the stitching together of images from the instrument’s three telescopes, which together cover the whole plasmasphere. We required a difference of 500 keV) seen by SAMPEX/PET seem to be caused by a mechanism that does not correlate with factors expected to favor EMIC waves. First, only those events at the most dusk-like local times approach the local time peak EMIC waves, and these are correlated with the hardest values of E0 (section 2.2). Second, the solar wind and geomagnetic conditions conducive to EMIC waves are seen primarily in those events that are both nearest to dusk and highest in E0 (section 3). Finally, in the few cases where data are available, the extended plasmasphere expected to produce EMIC waves and scatter electrons is seen only in those events that are, again, near dusk and highest in E0 (section 4.2). If CSS is considered as the alternate mechanism near midnight, then simultaneous proton precipitation does not necessarily help in distinguishing the two causes (see section 1.2 and Yahnin et al. [2014]). Thus, we suggest that a significant fraction of the events attributed to EMIC wave scattering in recent surveys due only to their hardness [Comess et al., 2013] (our data set) or to the additional detection of ion precipitation [Carson et al., 2013; Wang et al., 2014] may in fact be CSS-driven events instead. While the rest of our discussion will feature this hypothesis, we note one other idea here. Hendry et al. [2016], in their statistical correlation of POES precipitation events and EMIC wave measurements from the ground, noted the difference between two classes of EMIC waves: those with a distinct rising tone, named “intervals of pulsations of diminishing periods” (IPDP) waves by Troitskaya [1961], and those without (narrowband Pc1 waves). IPDP EMIC waves were shown by Hendry et al. [2016] to favor the earlier (dusk-like) MLTs almost exclusively and were more likely to be found in association with POES precipitation; non-IPDP EMIC waves were more extended in MLT, more likely to appear near midnight and somewhat less likely to be associated with precipitation. Yet the number of non-IPDP Pc1 wave events driving relativistic electron precipitation was also statistically significant. If it happens that the non-IPDP class of waves is not strongly associated with parameters like Pdyn and plasmaspheric extension, and only the IPDP class is, then the dichotomy we are demonstrating in this paper could actually be a dichotomy related to the two classes of EMIC (or at least Pc1) waves, rather than between EMIC waves and CSS. In fact, two distinct modes of 30 keV proton precipitation coming from IPDP and non-IPDP Pc1 waves were identified by Yahnina et al. [2003]. Precipitation caused by the IPDP waves had all the usual hallmarks of EMIC wave-driven precipitation: concentration at dusk, extended plasmasphere, disturbed conditions, and recent proton injections and generally was accompanied by the precipitation of > 30 keV electrons and < 20 keV protons. The non-IPDP events were concentrated in the morning (possibly extending to the dayside, where data were not available), happened during geomagnetically quiet times, with very low plasma density at geosynchronous orbit, and were much less likely to be accompanied by electrons and lower energy protons. This dichotomy somewhat parallels ours and is intriguing in the context of our results, but the non-IPDP events, which might be hoped to correspond to our near-midnight class of hard REP events, were never detected near midnight by Yahnina et al. [2003]. Returning to the CSS hypothesis, Yahnin et al. [2014] give cases of relativistic precipitation that show distinct CSS signatures. The POES data to directly examine 92 events with precipitation of electrons >1 MeV and ions of 30–80 keV. They classed 65 of these events as showing predicted characteristics of CSS: isotropy occurred at the outer trapping boundary, this boundary migrated to lower L with higher energy and was at even lower L for the ions (since the 30–80 keV band lies well above the energy of ions with gyroradius equal to that of >1 MeV electrons; see section 1.2). The remaining 27 events showed no such evolution in L with particle gyroradius and occured well inward of the outer trapping boundary; they were considered by Yahnin et al. [2014] to be EMIC driven. These events were therefore only 29% of their hard REP events that included ion precipitation, and they were 13% of their hard REP events overall. This low percentage is not necessarily inconsistent with our estimate of 44%, since we excluded many events that contain significant >1 MeV precipitation but have E0 < 500 keV, which probably increased the percentage of EMIC-driven events in our sample relative to that of Yahnin et al. [2014]. We note that with our current sample of SAMPEX data, we cannot measure proximity of the precipitation spike to the outer trapping boundary because we are observing only electrons in the bounce loss cone and do not see the trapped population. We are still left with one significant puzzle: even the population of precipitation events that we are attributing to EMIC wave scattering peaks at later MLT than EMIC waves themselves (compare the duskward edge of the E0 > 700 keV peak in Figure 1 with the EMIC wave data). This problem is only exacerbated in studies that do not attribute most hard precipitation events near midnight to another cause like CSS. Based on the MLT distributions of waves shown in Hendry et al. [2016], a preferred connection between precipitation and non-IPDP SMITH ET AL.


8609


10.1002/2016JA022346

waves over IPDP waves could explain a later-peaked MLT; however, in that work the authors found that it was the earlier-peaking IPDP waves that were most strongly correlated with REP. Even for a single wave population, however, it is not necessarily to be expected that hard REP caused by EMIC waves must have exactly the same distribution of occurrence (in L, MLT, and geomagnetic conditions) as the waves themselves. Precipitation will only happen when two additional conditions are met: that there are many relativistic trapped electrons in the given L shell available to precipitate and that they are high enough in energy to meet the resonant condition. These requirements, under different conditions, could introduce biases in the behavior of precipitation events relative to the behavior of EMIC waves, such as toward low L where the trapped electrons more reliably exist, toward high L where the low magnetic field pushes the resonance condition down to lower energies, and toward earlier MLT where plasmaspheric plumes are more likely to exist (with higher densities that also drive the resonance energy downward). None of these potential biases would explain a hard REP distribution biased toward midnight relative to the EMIC distribution. There is one other tendency that could, however: on a given drift shell, in a stretched field, the lowest magnetic field strength seen by a particle is at midnight at the magnetic equator (L. Woodger, private communication, 2016). This suggests that electrons drifting through EMIC waves that cover a broad range in MLT might be most likely to resonate and precipitate close to midnight. It may also be that the pitch angle distribution of electrons drifting through the afternoon sector is preconditioned by EMIC wave interactions in a way that makes precipitation by another mechanism (or by more EMIC waves) further toward midnight more likely. This might be expected if EMIC wave scattering is generally gradual and if it is more effective for electrons at large equatorial pitch angles than small ones near the loss cone. The opposite is, however, predicted to be the case: pitch angle diffusion rates due to EMIC waves have been calculated to be strong only at low equatorial pitch angles for 1–2 MeV electrons and negligible at large pitch angles [Albert, 2003; Ni et al., 2015]. This makes using EMIC waves to “prime the pump” for another precipitation mechanism without causing precipitation themselves at too early an MLT a finely tuned proposition. For them to scatter electrons toward the loss cone but not into it, there would have to be a pitch angle distribution heading into the afternoon sector that was heavily populated at intermediate pitch angles (say 20–30∘ ) but empty within several degrees of the edge of the loss cone. A possibly relevant case study of hard precipitation observed from a balloon in the context of CSS is given by Millan et al. [2007]. This event, on 19–20 January 2000, was by far the longest-duration hard REP event ever seen from a balloon (lasting over 3 h, when ∼15 min is more typical), and it was solidly in the premidnight sector that is problematic for the EMIC mechanism (ranging from 1920 to 2240 MLT). It was on the softer side for a hard REP seen from a balloon; at E0 = 290 keV, it would not even have been classed as hard REP by our SAMPEX definition. But e-foldings even this hard are quite rare in balloon data, possibly because the balloon gamma ray spectrometers are sensitive in a somewhat lower energy band than SAMPEX/PET and the electron spectrum may not be exactly exponential over the whole range from a few hundred keV to 5 MeV. While this event was discussed in Millan et al. [2007] as potentially related to the EMIC mechanism, it was noted in that work that it occurred during a period of magnetic field stretching in the tail (which would be favorable to CSS) and that it was capable of accounting for an accompanying dropout in >2 MeV electrons seen in GOES. Of the BARREL hard REP events associated with EMIC wave observations, two occur right at dusk, earlier in MLT than most of the SAMPEX hard REP [L. W. Blum et al., 2015] and one closer to midnight [Li et al., 2014]. The latter event reminds us that in no individual case can a near-midnight MLT contradict the EMIC wave model for hard REP, since EMIC waves can occasionally occur at any MLT; the contradiction we point out is statistical in nature and refers to our expectations for the majority of the SAMPEX hard REP cases. In future work, position at the outer trapping boundary can be used as a diagnostic of current sheet scattering for any given event, as can the expected softening of the spectrum with increasing L [West et al., 1978; Imhof et al., 1979; Yahnin et al., 2014]. Any population of precipitation events expected to consist mostly of current sheet scattering should also correlate to times where the tail is particularly stretched. More speculatively, enhanced ULF waves may also play a supporting role, either by transiently decreasing RC in the tail or by providing enhanced radial transport to replenish the particles capable of being lost, as is thought to be the case with magnetopause shadowing on the day side [Hudson et al., 2014]. SMITH ET AL.


8610


10.1002/2016JA022346

5.2. Overall Contributions to Outer Belt Losses Aside from illuminating the underlying physics, the other major goal for this sort of work is to determine what mechanism(s) are primarily responsible for the loss of MeV electrons from the outer belt. Magnetopause shadowing [Turner et al., 2012; Hudson et al., 2014; Ozeke et al., 2014], hard REP possibly due to EMIC wave scattering [Millan et al., 2002, 2007], Figure 7. As Figure 1 but without the EMIC wave data and with the and microburst precipitation from choy axis scaled to show the actual number of events identified in each rus [Thorne et al., 2005] have all been category in each hour of MLT. A trace showing the sum of the three hardest precipitation bands has been added. shown to be capable of emptying the belt in days or less, although microburst precipitation is probably usually accompanied by a greater amount of acceleration of the trapped population at the same time, so that it may not be well described as a pure loss mechanism. Figure 7 shows the same curves as in Figure 1 (except for the EMIC wave distribution), but now they have not been normalized to unity; the area under the curves here represents the total number of times that each class of precipitation was detected during a pass (but still corrected for differing exposure with MLT). There may well have been multiple episodes of precipitation at different values of L in some passes; in each pass, the event that caused the highest peak in HILT count rates (>1 MeV) was selected, and others would have been missed. But it is apparent that the hardest values of E0 , and those near dusk—i.e., the population we would attribute to EMIC wave scattering—are a small fraction of the overall population of events showing >1 MeV electrons in HILT: ∼45% of those with E0 >500 keV as estimated in section 3 and 5% of the total population of REP events broken down in Figures 1 and 7. Under the assumption that hardest REP events are a mix of EMIC-driven and current sheet scattering events, with perhaps even one or more other mechanisms contributing, approaches not used in this work are needed to sort them out on an individual basis and sum up their contributions to overall losses in different geomagnetic conditions. The number of separate events is not sufficient to estimate the contribution to losses; estimates are needed for the intensity, duration, and extent in L and MLT of each event, and there is a chance that this could reveal a larger role for EMIC waves than their small fraction of overall REP events would suggest. But we note that the longest hard REP event ever seen from a balloon [Millan et al., 2007], which may have caused a dropout in >2 MeV electrons at geosynchronous orbit, had context that suggests it may have been driven by CSS (see above). If CSS-driven events are much longer in duration than EMIC-driven ones, then low-Earth orbiting spacecraft like SAMPEX will see more of the longer CSS events, since each pass under the outer belt is like a snapshot, whereas balloons will observe the two types of event in their correct proportion by number. But the question of whether the hardest events are mostly EMIC driven or due to current sheet scattering misses an even more general question. Figure 7 demonstrates that even if we were to attribute all the hardest events, which peak at premidnight MLT, to EMIC wave scattering, both those events and the microburst events are overshadowed in number by the majority of the events we see, which are neither hard and dusk-like nor bursty and dawnlike but which occur at all local times and have a distinct peak at midnight. Although these events have somewhat softer spectra than the hard premidnight population, they have also been selected by looking at >1 MeV electrons detected in SAMPEX/HILT, so they are certainly cases of relativistic precipitation and may even dominate the contribution of the harder events depending on geomagnetic conditions and what threshold value of electron energy is considered. Finally, there may be some bias introduced by the bounce loss cone restriction. While it makes for much cleaner measurements—we know that everything we see is precipitating and was scattered at the MLT where we observe the precipitation—it also restricts us preferentially to cases of relatively strong pitch angle scattering. We note that Meredith et al. [2011], in specifically observing >1 MeV electrons freshly injected into the drift loss cone during all the phases of high-speed solar wind stream driven storms, found a broad peak in MLT centered in the noon-to-dusk quadrant, i.e., matching the MLT distribution of EMIC waves. This stands SMITH ET AL.


8611


10.1002/2016JA022346

in contrast to all the other surveys, including ours, that found a premidnight peak that is difficult to reconcile with the EMIC distribution [Vampola, 1971; Imhof et al., 1986; Carson et al., 2013; Comess et al., 2013; Wang et al., 2014]. It may be worthwhile to explore the possibility that EMIC wave scattering proceeds more often as gradual pitch angle diffusion that feeds primarily the drift loss cone while current sheet scattering more often fills the entire loss cone. Acknowledgments D.M.S. and E.P.C. thank Jerry Goldstein and Dennis Gallagher for assistance with analyzing and understanding the IMAGE/EUV data. DMS thanks Jacob Bortnik, Yuri Shprits, and Maria Usanova for useful conversations on EMIC waves and other phenomena and Paul O’Brien, Mark Looper, Mark Moldwin, and Joseph Mazur for help in understanding the magnetic coordinates used for the various spacecraft. We thank the referees for criticisms and insights that resulted in a major transformation of the paper from its earliest form. This work made use of geomagnetic and solar wind data available through NASA/GSFC’s Space Physics Data Facility’s OMNIWeb service (http://omniweb.gsfc.nasa.gov/), SAMPEX data available through the SAMPEX Data Center at the California Institute of Technology (http://www.srl.caltech.edu/sampex/ DataCenter/), and IMAGE/EUV data available from the University of Arizona (http://euv.lpl.arizona.edu/ euv/). Processed data products and code can be provided by the first author on request.

SMITH ET AL.

References Albert, J. M. (2003), Evaluation of quasi-linear diffusion coefficients for EMIC waves in a multispecies plasma, J. Geophys. Res., 108, 1249, doi:10.1029/2002JA009792. Allen, R. C., J.-C. Zhang, L. M. Kistler, H. E. Spence, R.-L. Lin, B. Klecker, M. W. Dunlop, M. André, and V. K. Jordanova (2015), A statistical study of EMIC waves observed by Cluster: 1. Wave properties, J. Geophys. Res. Space Physics, 120, 5574–5592, doi:10.1002/2015JA021333. Anderson, B. J., R. E. Erlandson, and L. J. Zanetti (1992), A statistical study of Pc 1–2 magnetic pulsations in the equatorial magnetosphere. I—Equatorial occurrence distributions. II—Wave properties, J. Geophys. Res., 97, 3075–3101, doi:10.1029/91JA02706. Blum, L., X. Li, and M. Denton (2015), Rapid MeV electron precipitation as observed by SAMPEX/HILT during high-speed stream-driven storms, J. Geophys. Res. Space Physics, 120, 3783–3794, doi:10.1002/2014JA020633. Blum, L. W., et al. (2015), Observations of coincident EMIC wave activity and duskside energetic electron precipitation on 18–19 January 2013, Geophys. Res. Lett., 42, 5727–5735, doi:10.1002/2015GL065245. Carson, B. R., C. J. Rodger, and M. A. Clilverd (2013), POES satellite observations of EMIC-wave driven relativistic electron precipitation during 1998–2010, J. Geophys. Res. Space Physics, 118, 232–243, doi:10.1029/2012JA017998. Clausen, L. B. N., J. B. H. Baker, J. M. Ruohoniemi, and H. J. Singer (2011), EMIC waves observed at geosynchronous orbit during solar minimum: Statistics and excitation, J. Geophys. Res., 116, A10205, doi:10.1029/2011JA016823. Clilverd, M. A., R. Duthie, R. Hardman, A. T. Hendry, C. J. Rodger, T. Raita, M. Engebretson, M. R. Lessard, D. Danskin, and D. K. Milling (2015), Electron precipitation from EMIC waves: A case study from 31 May 2013, J. Geophys. Res. Space Physics, 120, 3618–3631, doi:10.1002/2015JA021090. Comess, M. D. (2011), Relativistic electron losses from the outer Van Allen Belts, PhD thesis, Univ. of California, Santa Cruz, Calif. Comess, M. D., D. M. Smith, R. S. Selesnick, R. M. Millan, and J. G. Sample (2013), Duskside relativistic electron precipitation as measured by SAMPEX: A statistical survey, J. Geophys. Res. Space Physics, 118, 5050–5058, doi:10.1002/jgra.50481. Cook, W. R., A. C. Cummings, J. R. Cummings, T. L. Garrard, B. Kecman, R. A. Mewaldt, R. S. Selesnick, E. C. Stone, D. N. Baker, and T. T. von Rosenvinge (1993), PET: A proton/electron telescope for studies of magnetospheric, solar, and galactic particles, IEEE Trans. Geosci. Remote Sens., 31, 565–571, doi:10.1109/36.225523. Denton, R. E., V. K. Jordanova, and J. Bortnik (2015), Resonance of relativistic electrons with electromagnetic ion cyclotron waves, Geophys. Res. Lett., 42, 8263–8270, doi:10.1002/2015GL064379. Foat, J. E., R. P. Lin, D. M. Smith, F. Fenrich, R. Millan, I. Roth, K. R. Lorentzen, M. P. McCarthy, G. K. Parks, and J. P. Treilhou (1998), First detection of a terrestrial MeV X-ray burst, Geophys. Res. Lett., 25, 4109–4112, doi:10.1029/1998GL900134. Gallagher, D. L., M. L. Adrian, and M. W. Liemohn (2005), Origin and evolution of deep plasmaspheric notches, J. Geophys. Res., 110, A09201, doi:10.1029/2004JA010906. Goldstein, J., M. Spasojević, P. H. Reiff, B. R. Sandel, W. T. Forrester, D. L. Gallagher, and B. W. Reinisch (2003), Identifying the plasmapause in IMAGE EUV data using IMAGE RPI in situ steep density gradients, J. Geophys. Res., 108, 1147, doi:10.1029/2002JA009475. Gvozdevskij, B. B., and V. A. Sergeev (1995), Scattering on a current sheet—A possible mechanism of auroral proton precipitation, Geomagn. Aeron., 35, 151–155. Halford, A. J., B. J. Fraser, and S. K. Morley (2010), EMIC wave activity during geomagnetic storm and nonstorm periods: CRRES results, J. Geophys. Res., 115, A12248, doi:10.1029/2010JA015716. Hendry, A. T., C. J. Rodger, M. A. Clilverd, M. J. Engebretson, I. R. Mann, M. R. Lessard, T. Raita, and D. K. Milling (2016), Confirmation of EMIC wave driven relativistic electron precipitation, J. Geophys. Res. Space Physics, 121, 5366–5383, doi:10.1002/2015JA022224. Hudson, M. K., D. N. Baker, J. Goldstein, B. T. Kress, J. Paral, F. R. Toffoletto, and M. Wiltberger (2014), Simulated magnetopause losses and Van Allen Probe flux dropouts, Geophys. Res. Lett., 41, 1113–1118, doi:10.1002/2014GL059222. Imhof, W. L. (1988), Fine resolution measurements of the L-dependent energy threshold for isotropy at the trapping boundary, J. Geophys. Res., 93, 9743–9752, doi:10.1029/JA093iA09p09743. Imhof, W. L., J. B. Reagan, and E. E. Gaines (1977), Fine-scale spatial structure in the pitch angle distributions of energetic particles near the midnight trapping boundary, J. Geophys. Res., 82, 5215–5221, doi:10.1029/JA082i032p05215. Imhof, W. L., J. B. Reagan, and E. E. Gaines (1979), Studies of the sharply defined L dependent energy threshold for isotropy at the midnight trapping boundary, J. Geophys. Res., 84, 6371–6384, doi:10.1029/JA084iA11p06371. Imhof, W. L., H. D. Voss, J. B. Reagan, D. W. Datlowe, and E. E. Gaines (1986), Relativistic electron and energetic ion precipitation spikes near the plasmapause, J. Geophys. Res., 91, 3077–3088, doi:10.1029/JA091iA03p03077. Klecker, B., D. Hovestadt, M. Scholer, H. Arbinger, M. Ertl, H. Kaestle, E. Kuenneth, P. Laeverenz, E. Seidenschwang, and J. B. Blake (1993), HILT: A heavy ion large area proportional counter telescope for solar and anomalous cosmic rays, IEEE Trans. Geosci. Remote Sens., 31, 542–548, doi:10.1109/36.225520. Langel, R. A. (1992), IGRF, 1991 revision, EOS Trans. AGU, 73, 182–182, doi:10.1029/91EO00151. Lee, J. J., G. K. Parks, K. W. Min, M. P. McCarthy, E. S. Lee, H. J. Kim, J. H. Park, and J. A. Hwang (2006), Relativistic electron dropouts by pitch angle scattering in the geomagnetic tail, Ann. Geophys., 24, 3151–3159, doi:10.5194/angeo-24-3151-2006. Li, Z., R. M. Millan, and M. K. Hudson (2013), Simulation of the energy distribution of relativistic electron precipitation caused by quasi-linear interactions with EMIC waves, J. Geophys. Res. Space Physics, 118, 7576–7583, doi:10.1002/2013JA019163. Li, Z., et al. (2014), Investigation of EMIC wave scattering as the cause for the BARREL 17 January 2013 relativistic electron precipitation event: A quantitative comparison of simulation with observations, Geophys. Res. Lett., 41, 8722–8729, doi:10.1002/2014GL062273. Lorentzen, K. R., M. P. McCarthy, G. K. Parks, J. E. Foat, R. M. Millan, D. M. Smith, R. P. Lin, and J. P. Treilhou (2000), Precipitation of relativistic electrons by interaction with electromagnetic ion cyclotron waves, J. Geophys. Res., 105, 5381–5390, doi:10.1029/1999JA000283. Meredith, N. P., R. M. Thorne, R. B. Horne, D. Summers, B. J. Fraser, and R. R. Anderson (2003), Statistical analysis of relativistic electron energies for cyclotron resonance with EMIC waves observed on CRRES, J. Geophys. Res., 108, 1250, doi:10.1029/2002JA009700. Meredith, N. P., R. B. Horne, M. M. Lam, M. H. Denton, J. E. Borovsky, and J. C. Green (2011), Energetic electron precipitation during high-speed solar wind stream driven storms, J. Geophys. Res., 116, A05223, doi:10.1029/2010JA016293.


8612


10.1002/2016JA022346

Meredith, N. P., R. B. Horne, T. Kersten, B. J. Fraser, and R. S. Grew (2014), Global morphology and spectral properties of EMIC waves derived from CRRES observations, J. Geophys. Res. Space Physics, 119, 5328–5342, doi:10.1002/2014JA020064. Mewaldt, R. A., C. M. S. Cohen, A. W. Labrador, R. A. Leske, G. M. Mason, M. I. Desai, M. D. Looper, J. E. Mazur, R. S. Selesnick, and D. K. Haggerty (2005), Proton, helium, and electron spectra during the large solar particle events of October–November 2003, J. Geophys. Res., 110, A09S18, doi:10.1029/2005JA011038. Millan, R. M., and R. M. Thorne (2007), Review of radiation belt relativistic electron losses, J. Atmos. Sol. Terr. Phys., 69, 362–377, doi:10.1016/j.jastp.2006.06.019. Millan, R. M., R. P. Lin, D. M. Smith, K. R. Lorentzen, and M. P. McCarthy (2002), X-ray observations of MeV electron precipitation with a balloon-borne germanium spectrometer, Geophys. Res. Lett., 29, 2194, doi:10.1029/2002GL015922. Millan, R. M., R. P. Lin, D. M. Smith, and M. P. McCarthy (2007), Observation of relativistic electron precipitation during a rapid decrease of trapped relativistic electron flux, Geophys. Res. Lett., 34, L10101, doi:10.1029/2006GL028653. Min, K., J. Lee, K. Keika, and W. Li (2012), Global distribution of EMIC waves derived from THEMIS observations, J. Geophys. Res., 117, A05219, doi:10.1029/2012JA017515. Miyoshi, Y., K. Sakaguchi, K. Shiokawa, D. Evans, J. Albert, M. Connors, and V. Jordanova (2008), Precipitation of radiation belt electrons by EMIC waves, observed from ground and space, Geophys. Res. Lett., 35, L23101, doi:10.1029/2008GL035727. Nakamura, R., M. Isowa, Y. Kamide, D. N. Baker, J. B. Blake, and M. Looper (2000), SAMPEX observations of precipitation bursts in the outer radiation belt, J. Geophys. Res., 105, 15,875–15,886, doi:10.1029/2000JA900018. Ni, B., et al. (2015), Resonant scattering of outer zone relativistic electrons by multiband EMIC waves and resultant electron loss time scales, J. Geophys. Res. Space Physics, 120, 7357–7373, doi:10.1002/2015JA021466. O’Brien, T. P., K. R. Lorentzen, I. R. Mann, N. P. Meredith, J. B. Blake, J. F. Fennell, M. D. Looper, D. K. Milling, and R. R. Anderson (2003), Energization of relativistic electrons in the presence of ULF power and MeV microbursts: Evidence for dual ULF and VLF acceleration, J. Geophys. Res., 108, 1329, doi:10.1029/2002JA009784. Ozeke, L. G., I. R. Mann, D. L. Turner, K. R. Murphy, A. W. Degeling, I. J. Rae, and D. K. Milling (2014), Modeling cross L shell impacts of magnetopause shadowing and ULF wave radial diffusion in the Van Allen belts, Geophys. Res. Lett., 41, 6556–6562, doi:10.1002/2014GL060787. Rodger, C. J., T. Raita, M. A. Clilverd, A. Seppälä, S. Dietrich, N. R. Thomson, and T. Ulich (2008), Observations of relativistic electron precipitation from the radiation belts driven by EMIC waves, Geophys. Res. Lett., 35, L16106, doi:10.1029/2008GL034804. Rodger, C. J., A. T. Hendry, M. A. Clilverd, C. A. Kletzing, J. B. Brundell, and G. D. Reeves (2015), High-resolution in situ observations of electron precipitation-causing EMIC waves, Geophys. Res. Lett., 42, 9633–9641, doi:10.1002/2015GL066581. Sandanger, M., F. Søraas, K. Aarsnes, K. Oksavik, and D. S. Evans (2007), Loss of relativistic electrons: Evidence for pitch angle scattering by electromagnetic ion cyclotron waves excited by unstable ring current protons, J. Geophys. Res., 112, A12213, doi:10.1029/2006JA012138. Sandel, B. R., et al. (2000), The extreme ultraviolet imager investigation for the IMAGE mission, Space Sci. Rev., 91, 197–242. Selesnick, R. (1993), PET Electron Calibrations, Caltech/SRL Tech. Rep. #93-1, California Institute of Technology, Pasadena, Calif. Sergeev, V. A., and N. A. Tsyganenko (1982), Energetic particle losses and trapping boundaries as deduced from calculations with a realistic magnetic field model, Planet. Space Sci., 30, 999–1006, doi:10.1016/0032-0633(82)90149-0. Sergeev, V. A., E. M. Sazhina, N. A. Tsyganenko, J. A. Lundblad, and F. Soraas (1983), Pitch-angle scattering of energetic protons in the magnetotail current sheet as the dominant source of their isotropic precipitation into the nightside ionosphere, Planet. Space Sci., 31, 1147–1155, doi:10.1016/0032-0633(83)90103-4. Smith, D. M., R. P. Lin, K. A. Anderson, K. Hurley, and C. M. Johns (1995), High-resolution spectra of 20–300 keV hard X-rays from electron precipitation over Antarctica, J. Geophys. Res., 100, 19,675–19,686, doi:10.1029/95JA01472. Thorne, R. M., and C. F. Kennel (1971), Relativistic electron precipitation during magnetic storm main phase, J. Geophys. Res., 76, 4446, doi:10.1029/JA076i019p04446. Thorne, R. M., T. P. O’Brien, Y. Y. Shprits, D. Summers, and R. B. Horne (2005), Timescale for MeV electron microburst loss during geomagnetic storms, J. Geophys. Res., 110, A09202, doi:10.1029/2004JA010882. Troitskaya, V. A. (1961), Pulsation of the Earth’s electromagnetic field with periods of 1 to 5 seconds and their connection with phenomena in the high atmosphere, J. Geophys. Res., 66, 5–18, doi:10.1029/JZ066i001p00005. Tsyganenko, N. A. (1989), A magnetospheric magnetic field model with a warped tail current sheet, Planet. Space Sci., 37, 5–20, doi:10.1016/0032-0633(89)90066-4. Turner, D. L., Y. Shprits, M. Hartinger, and V. Angelopoulos (2012), Explaining sudden losses of outer radiation belt electrons during geomagnetic storms, Nat. Phys., 8, 208–212, doi:10.1038/nphys2185. Usanova, M. E., I. R. Mann, J. Bortnik, L. Shao, and V. Angelopoulos (2012), THEMIS observations of electromagnetic ion cyclotron wave occurrence: Dependence on AE, SYMH, and solar wind dynamic pressure, J. Geophys. Res., 117, A10218, doi:10.1029/2012JA018049. Usanova, M. E., F. Darrouzet, I. R. Mann, and J. Bortnik (2013), Statistical analysis of EMIC waves in plasmaspheric plumes from Cluster observations, J. Geophys. Res. Space Physics, 118, 4946–4951, doi:10.1002/jgra.50464. Vampola, A. L. (1971), Electron pitch angle scattering in the outer zone during magnetically disturbed times, J. Geophys. Res. Space Physics, 76, 4685–4688, doi:10.1029/JA076i019p04685. Wang, Z., Z. Yuan, M. Li, H. Li, D. Wang, H. Li, S. Huang, and Z. Qiao (2014), Statistical characteristics of EMIC wave-driven relativistic electron precipitation with observations of POES satellites: Revisit, J. Geophys. Res. Space Physics, 119, 5509–5519, doi:10.1002/2014JA020082. West, H. I., R. M. Buck, and M. G. Kivelson (1978), On the configuration of the magnetotail near midnight during quiet and weakly disturbed periods: Magnetic field modeling, J. Geophys. Res., 83, 3819–3831, doi:10.1029/JA083iA08p03819. Yahnin, A. G., T. A. Yahnina, N. V. Semenova, and B. B. Gvozdevsky (2014), Relativistic electron precipitation as seen By NOAA POES, in Physics of Auroral Phenomena, Proc. XXXVII Annual Seminar (Apatity, Russia, 2014), pp. 46–50, Kola Science Center, Russian Academy of Sciences, Apatity, Murmansk, Russia. Yahnina, T. A., A. G. Yahnin, J. Kangas, J. Manninen, D. S. Evans, A. G. Demekhov, V. Y. Trakhtengerts, M. F. Thomsen, G. D. Reeves, and B. B. Gvozdevsky (2003), Energetic particle counterparts for geomagnetic pulsations of Pc1 and IPDP types, Ann. Geophys., 21, 2281–2292, doi:10.5194/angeo-21-2281-2003. Yuan, Z., M. Li, Y. Xiong, H. Li, M. Zhou, D. Wang, S. Huang, X. Deng, and J. Wang (2013), Simultaneous observations of precipitating radiation belt electrons and ring current ions associated with the plasmaspheric plume, J. Geophys. Res. Space Physics, 118, 4391–4399, doi:10.1002/jgra.50432.

SMITH ET AL.


8613