The Global Statistical Response of the Outer Radiation Belt During Geomagnetic Storms

Using the total radiation belt electron content calculated from Van Allen Probe phase space density, the time‐dependent and global response of the outer radiation belt during storms is statistically studied. Using phase space density reduces the impacts of adiabatic changes in the main phase, allowing a separation of adiabatic and nonadiabatic effects and revealing a clear modality and repeatable sequence of events in storm time radiation belt electron dynamics. This sequence exhibits an important first adiabatic invariant (μ)‐dependent behavior in the seed (150 MeV/G), relativistic (1,000 MeV/G), and ultrarelativistic (4,000 MeV/G) populations. The outer radiation belt statistically shows an initial phase dominated by loss followed by a second phase of rapid acceleration, while the seed population shows little loss and immediate enhancement. The time sequence of the transition to the acceleration is also strongly μ dependent and occurs at low μ first, appearing to be repeatable from storm to storm.


Introduction
The outer radiation belt is a toroidal region of the magnetosphere between~2 and 7 Earth radii from the center of the Earth populated by trapped electrons with energies from hundreds of 100s of keV to multiple MeV (Mauk et al., 2013). Typically, the outer radiation belt exists in a quiescent state. However, during geomagnetic storms the outer radiation belt becomes increasingly dynamic, at times leading to the filling of the slot region which generally separates the outer from the inner radiation belt Loto'aniu et al., 2006), the creation of a third radiation belt (Baker et al., 2013;Mann et al., 2016;Turner et al., 2013;Yuan & Zong, 2013a), and the generation of highly energetic electrons capable of disrupting satellite operations (Baker et al., 1994;Wrenn, 1995). These dynamics are controlled by various loss, acceleration, and transport mechanisms including magnetopause shadowing (West et al., 1972) and wave-particle interactions (Schulz & Lanzerotti, 1974), driven by enhanced storm time solar wind and geomagnetic activity (Murphy et al., 2015;. The overall response of the radiation belt during storms results from a superposition of these processes, culminating in an extremely complex and coupled system. For example, comparisons of prestorm and poststorm fluxes of relativistic electrons at geosynchronous orbit suggest that the dynamics of the outer radiation belt can be difficult to predict based on geomagnetic activity alone (Anderson et al., 2015;Reeves et al., 2003). However, when the dynamics of storm time outer radiation belt electrons are characterized as a function of L shell and electron energy  or solar wind driver (e.g., Hietala et al.,

RESEARCH LETTER
2014; Kilpua et al., 2015;Miyoshi & Kataoka, 2005;Yuan & Zong, 2013b), the storm time response is generally more predictable. Turner et al. (2015) demonstrated that enhancements in the flux of lower-energy electrons (100s of keV) are common during storms, especially at low L shells (L <~4.6 R E ). Miyoshi and Kataoka (2005) showed that the geosynchronous electron flux tended to be higher during storms driven by corotating interaction regions (CIRs) than those driven by coronal mass ejections (CMEs; c.f., Borovsky & Denton, 2006). Kilpua et al. (2015) demonstrated that storms driven by CME sheath or ejecta were dominated by electron loss. Yuan and Zong (2013b) used low-altitude observations of relativistic electrons to demonstrate that CME-driven storms produce larger radiation belt enhancements at lower L shells than the CIR-driven storms (c.f., Shen et al., 2017). Interestingly, the sum of individual results from these studies does not form a coherent description of the response of the outer radiation belt to geomagnetic storms; rather the radiation belt exhibits a complex response dependent on electron energy, L shell, and storm driver. Here we suggest that a global measure of the energetic electrons in the outer radiation belt, analyzed in a way that normalizes the duration of each important stage in a storm, can provide new information and well-defined patterns in outer radiation belt behavior.
Using Van Allen Probes observations, we characterize the response of the outer radiation belt during 73 storms. Electron phase space density (PSD) and total radiation belt electron content (TRBEC) are used to characterize the radiation belt dynamics as a function of the first and second adiabatic invariants μ and K such that reversible adiabatic effects are removed. Using TRBEC, we demonstrate that storm time electron radiation belt dynamics are statistically characterized sequentially by an initial period dominated by loss followed by a subsequent period dominated by rapid acceleration. Significantly, our results reveal a level of simplicity and repeatability in storm time radiation belt electron dynamics not previously appreciated: the μ dependence of these results being important for understanding the dominant processes which act to shape energetic particle dynamics during storms.

Data and Analysis
This study uses data from the Van Allen Probes and OMNIweb (King & Papitashvili, 2005) to statistically characterize the solar wind and response of the radiation belt during 73 geomagnetic storms. We use 52 storms as detailed by Turner et al. (2015) from September 2012 to February 2015, which were identified using a Sym-H threshold of À50 nT. These storms are supplemented with 21 storms from February 2015 to March 2016 using the same criteria. These storms are composed of mostly CME-driven storms (~57%). This database (supporting information, Table S1) is used to perform a superposed epoch analysis of storm time radiation belt electron dynamics.
The storm time response of the outer radiation belt is characterized using observations from the Magnetic Electron Ion Spectrometer (MagEIS) (Blake et al., 2013;Claudepierre et al., 2015) and Relativistic Electron Proton Telescope (REPT) (Baker et al., 2012) instruments to calculate the TRBEC (or radiation belt content; . The TRBEC is a proxy for the total number of electrons within the radiation belt and allows the dimensionality of the complex multidimensional MagEIS and REPT data sets to be reduced. TRBEC is ideal for characterizing macroscale changes in the radiation belt for statistical studies , including the response of the radiation belt to substorms (Forsyth et al., 2016). The TRBEC used here is derived by integrating electron PSD over the three adiabatic invariants, μ, K, and L * . The PSD f is calculated using the methodology detailed in Boyd et al. (2014), supporting information Text S2. TRBEC N can then be calculated from PSD according to where R E is the radius of the Earth, m 0 is the mass of an electron, and μ 0 is the permeability of free space. By integrating equation (1) TRBEC can be calculated for different electron energy populations (integral over fixed μ range) and different regions of the outer radiation belt (integral over fixed ranges in K and L*) while removing reversible adiabatic effects, such as the Dst effect (Kim & Chan, 1997).

Geophysical Research Letters
In this study PSD is integrated over each half orbit for discrete values of μ and K giving TRBEC N(μ, K); the lower limit of the integral is L* = 3, and the upper limit is L* at apogee. Three values of μ are used, μ 1 = 150 MeV/G, μ 2 = 1,000 MeV/G, and μ 3 = 4,000 MeV/G, corresponding to the radiation belt seed, relativistic, and ultrarelativistic electron populations. At L* = 5, μ 1 corresponds to~300-keV electrons, μ 2 to~1-MeV electrons, and μ 3 to~2.5 MeV. Note that the Van Allen Probes is a near-equatorial mission which limits the pitch angle coverage when the probes are off the equator. Hence, we restrict our analysis to K ∈ [0.02,0.9] R E G 1/2 ; outside this range, errors in PSD, and thus TRBEC, can be large. The calculation of TRBEC N(μ, K) allows the investigation of radiation belt seed, relativistic, and ultrarelativistic electron populations as a function of K (low K corresponding to near-equatorially trapped electrons) across the outer radiation belt (integration of each half orbit in L*) with adiabatic effects removed.
It is important to note that variation in the strength of the Earth's magnetic field during storms changes the L* at the apogee of the Van Allen Probes and hence the upper limit of integration over L*. For instance, during the storm main phase L* at apogee typically decreases due to compressions of the dayside magnetic field resulting in an increase in the local magnetic field strength (c.f., Figure 1). Thus, while adiabatic effects are removed using PSD, there does exist an unavoidable variation in the integral limits of L* when calculating TRBEC. Low Earth orbiting satellites, such as Solar Anomalous and Magnetospheric Particle Explorer (SAMPEX) (e.g., , can overcome this limitation. Future work will utilize low Earth orbiting observations to investigate variation in TRBEC over fixed L ranges. For each storm, three times are independently identified from solar wind and Dst to define the start of the storm t s , the epoch t E , and the end of the storm t f . The epoch is taken as the time of minimum Dst during each storm. The start of each geomagnetic storm is identified by enhanced solar wind driving, and the end of each storm is determined by recovery of Dst following the end of enhanced solar wind driving (see supporting information Text S3 for specific details of the algorithm). These phases correspond most clearly to the definition of the storm main and recovery phases. It is important to note that these two phases are not the same duration for each storm. To perform a superposed epoch analysis, the initial phase of each storm is normalized to 30 hr and the subsequent phase is normalized to 120 hr (e.g., Yokoyama & Kamide, 1997). Hence, our study differs from previous studies which define the start and end of the storm as a fixed number of days pre-epoch and postepoch (e.g., Kataoka & Miyoshi, 2006) or use the time of the maximum electron flux in a fixed number of days pre-epoch and postepoch (e.g., Reeves et al., 2003). Figure 1 illustrates the identification of the three storm times and evolution of TRBEC during a storm occurring from 28 February to 6 March 2013. Figures 1a-1d show the solar wind dynamic pressure P dyn , north-south component of the interplanetary magnetic field (IMF) B Z , solar wind velocity V SW , and Dst. Figures 1e-1g show the electron PSD at three μ values and the TRBEC for each μ at K = 0.1 R E G 1/2 ; the smallest value of continuously observed by the Van Allen probes and characteristic of electrons mirroring close to the magnetic equator (Boyd et al., 2016).
In the solar wind, the initial phase of the storm is characterized by negative B Z and a rapid increase in solar wind dynamic pressure and velocity. P dyn peaks and B Z minimize during the initial phase of the storm and rapidly approach quiet values (P dyn < 2 nPa, B Z~0 nT) during the second phase of the storm. During the second phase V SW continues to grow to a peak of 638 km/s and then decays toward quiet time values over the subsequent 4 days. During the initial phase, Dst rapidly decreases until the epoch, reaching a minimum of (e-g) The bottom three panels show the PSD at three μ values, μ = 150, 1,000, 4,000 MeV/G, corresponding to the radiation belt seed, relativistic, and ultrarelativistic electron populations, respectively. The dashed and solid black lines in the bottom three panels are the TRBEC N for each μ at fixed K = 0.1 R E G 1/2 from Van Allen A and B, respectively.

Geophysical Research Letters
À55 nT, and then slowly recovers during the second phase of the storm. During this storm the radiation belt shows a strong and very clear response to solar wind driving. A sharp decrease in both PSD and TRBEC is observed in the initial phase followed by a rapid increase exceeding prestorm values, representing a storm time radiation belt enhancement. Note that despite the limitation in calculating TRBEC over L*, TRBEC still captures the key dynamics of the geomagnetic storm, in this case an initial period of loss followed by a rapid enhancement of the outer radiation belt. The following sections present results of a superposed epoch analysis of TRBEC and solar wind and discuss the implications of these results for radiation belt dynamics. 1 R E G 1/2 . In each panel the median is shown in black, mean in red, and upper and lower quartiles in gray. During the initial phase of geomagnetic storms (t s -t E ) the solar wind shows a monotonic increase in V SW and decrease in B Z . Solar wind P dyn and V ρ also show a clear increase though are more dynamic, exhibiting multiple peaks during the initial phase. During this initial phase P dyn , V ρ , and IMF B Z reach their most extreme values. Following the initial phase of the geomagnetic storms (t E -t f ), P dyn , V ρ , and IMF B Z rapidly approach quiet values, while V SW continues to increase followed by a slow decay over an extended period. Geomagnetic activity increases as characterized by a rapid decrease in Sym-H during the initial phase and slowly approaches quiet values during the second phase. Despite the continued increase in V SW during the second phase, a rapid drop in V ρ leads to a significant decrease in P dyn . These patterns are consistent in the median, mean, and upper and lower quartiles.

Results
In terms of the dynamics of TRBEC, there is a clear μ-dependent response at fixed K = 0.1 R E G 1/2 . During the initial phase, at μ 1 , the response of TRBEC is strongly dependent on the quartiles. The lower quartile shows evidence of a small amount of loss, while the median, mean, and upper quartiles remain relatively constant. However, during the second phase the response is very well organized, the quartiles and mean and median values all show a rapid increase in TRBEC. This can be contrasted with the responses at higher μ where each quartile, and indeed almost all of the storms, shows the same behavior. At μ 2 the initial phase is characterized by a systematic net decrease reaching a minimum at the epoch time where it subsequently rapidly enhances during the second phase of the storm. At higher μ, μ 3 shows a similar pattern of net decrease followed by net increase; however, the minimum TRBEC observed is after the epoch time. Hence, the time which acceleration is observed is strongly μ dependent, but at μ 2 and μ 3 almost all storms show the same time sequence of loss and then acceleration across the quartiles. The statistical significance in the variation of TRBEC is tested using the Student's t test and the Mann-Whitney-Wilcoxon test which are used to determine if the means and medians of two distributions are statically different, respectively. Comparing the TRBEC distributions at t À30 and t 0 , and t 0 and t 30 , using the Student's t test and Mann-Whitney-Wilcoxon test demonstrates that the decrease in the

Geophysical Research Letters
TRBEC in μ 2 and μ 3 during the initial phase and increase in TRBEC across all μ during the second phase are statistically significant to the 99% confidence level.  Figure 3 shows the median value of TRBEC N from the superposed epoch analysis, the middle column shows the fractional change of TRBEC N relative to N 0 observed at t s defined as ΔN 0 = (N i À N 0 )/N 0 , and the right column shows the fractional change of TRBEC N as a function of time ΔN i = (N i À N i À 1 )/N i À 1 , where i denotes the time step. Across K the response of the radiation belt during storms is remarkably coherent, each μ showing a similar pattern as observed in Figure 2.
At the higher μ and across K, the left column of Figure 3 shows that the initial phase of storms is dominated by a net decrease or loss of electrons and the second phase is dominated by a net increase of electrons mediated either by acceleration or transport of electrons. This sequence is most discernible at the higher μ values corresponding to the relativistic (μ 2 ) and ultrarelativistic (μ 3 ) electron populations (middle and bottom rows of Figure 3). During the initial phase the TRBEC at these μ values decreases by an order of magnitude (yellow to green) and then rapidly increases by a similar factor (green to red) during the second phase of the storm. At lower μ (top row) the overall behavior of the seed population is different; a decrease during the initial phase is less apparent although a clear enhancement of TRBEC is observed during the second phase. Note the at μ 3 the TRBEC above K = 0.58 R E G 1/2 is nominally zero, due to low count rates at high energies.
The fractional change in TRBEC relative to the start of the storm, ΔN 0 (middle column), depicts periods of a storm during which TRBEC is either smaller, ΔN 0 < 0 (blue), or larger, ΔN 0 > 0 (red), than the initial value of TRBEC. Contours are drawn at ΔN 0 = À 0.5, 0, 1 (blue, black, and red) corresponding to a decrease by a factor of 2, no change, and an increase by a factor of 2 in TRBEC relative to the start of the storm. The plots of ΔN 0 as a function of μ and K very clearly illustrate the sequence of events discussed above, a net decrease followed by net increase in TRBEC during storms. During the initial phase of storms μ 1 exhibits a relative loss at K > 0.05 R E G 1/2 (ΔN 0 < 0). Below this value of K there is a relative increase (ΔN 0 < 0). Note that the Middle column, fractional change in TRBEC relative to the TRBEC relative to N 0 observed at t s defined as ΔN 0 = (N i À N 0 )/N 0 ; the contours in each panel are drawn at ΔN 0 = À 0.5, 0, 1 corresponding to a decrease by a factor of 2, no change, and an increase by a factor of 2 in TRBEC relative to the start of the storm (blue, black, and red, respectively). Right column, fractional change in TRBEC as a function of time defined as ΔN i = (N i À N i À 1 )/N i À 1 where the subscript i denotes time.

Geophysical Research Letters
Van Allen Probe orbit does not continuously observe PSD at these values of K, and so we cannot necessarily interpret these changes physically. In contrast, during the second phase μ 1 shows a relative increase in TRBEC across the spectrum of K (ΔN 0 > 0). At the higher μs, μ 2 and μ 3 , the initial phase and early second phase are characterized by a relative decrease in TRBEC followed by a relative increase from the start of storm. This is characterized by negative values of ΔN 0 (blue) during the initial phase and positive values of ΔN 0 (red) during the second phase of storms. The transition from decreases to increases occurs at t = 10 hr for μ 2 and slightly later at t = 15 hr for μ 3 .
The right column of Figure 3 shows the fractional change in TRBEC as a function of time ΔN i , identifying intervals when electron loss (negative values) or acceleration and transport (positive values) dominate. Three distinct features stand out in ΔN i . First, each μ is initially characterized by negative ΔN i (blue) and a net loss of electrons is also illustrated in ΔN 0 (middle column of Figure 3). Second, following an initial period of negative ΔN i , each μ shows a short period of large and positive ΔN i (red) characterizing a period of rapid increase in TRBEC across μ and K. The onset of this rapid increase in ΔN i is also a function of μ, observed initially at μ 1 at t = À5 hr, subsequently at μ 2 at t = 0 hr, and finally at μ 3 at t = 5 hr. There is also evidence that this increase is dependent on K, starting initially at lower K and moving to higher values. Finally, following the period of positive ΔN i , there is a transition to a quasi steady state in TRBEC at t > 20 hr for each of μ. This is characterized by small values of ΔN i that fluctuate between positive and negative.
Overall, Figures 2 and 3 demonstrate three novel aspects of storm time radiation belt electrons. First, the radiation belt has a clear sequence of events responding in a statistically repeatable manner during storms. This response is characterized by an initial period dominated by net decreases followed by a short period of rapid increases in TRBEC. Second, this response is observed across nearly all K at fixed μ demonstrating a coherence in K or pitch angle in this response. Finally, this response is μ dependent; the transition from decreases in TRBEC to increases occurs first for low μ 1 and subsequently for μ 2 and μ 3 or initially in the radiation belt seed population followed by the relativistic and ultrarelativistic populations.

Discussion
In this paper we have performed a superposed epoch analysis of the dynamics of the Earth's outer radiation belt during 73 geomagnetic storms observed by the Van Allen Probes predominantly driven by CMEs (57%). The superposed epoch analysis uses electron PSD (Boyd et al., 2016) to calculate the TRBEC Forsyth et al., 2016) as a function of the first and second adiabatic invariants. The use of PSD and TRBEC in the superposed epoch analysis has two key advantages for studying storm time electron dynamics. First, PSD removes adiabatic or reversible changes, such as the Dst effect, in the derivation of TRBEC. By removing adiabatic effects, changes in TRBEC are the result of real changes in the number of outer radiation belt electrons, at least to the extent that this can be monitored across the L* range observed along the Van Allen Probes orbit. Second, with TRBEC global electron dynamics throughout the outer radiation belt can be investigated as opposed to single point measurements at a fixed L shell.
In the superposed epoch analysis the start of each storm is defined by enhanced solar wind driving, which leads to enhanced geomagnetic activity such as increased ultralow frequency (ULF; e.g., Mathie & Mann, 2001;Murphy et al., 2015) very low frequency (VLF; Aryan et al., 2016), and electromagnetic ion cyclotron (EMIC) wave activity (Halford et al., 2016;Usanova et al., 2012). The epoch time t E is defined as the minimum in Dst, and the end of each storm is defined as the end of enhanced solar wind driving and recovery of Dst to nominally quiet values. By characterizing epoch times based on enhanced solar wind and geomagnetic activity, the dynamics of storm time radiation belt electrons can be studied accounting for the varying length of the storm. Further, this analysis provides three independently defined and physics-based epoch times determined from observed quantities associated with enhanced geomagnetic activity and thus enhanced dynamics in radiation belt electrons and TRBEC. These epoch times allow us to cross compare and statistically characterize storm time electron dynamics based on enhanced solar wind driving and geomagnetic activity avoiding the pitfalls when using fixed epochs that can mask patterns.
A clear sequence of events is statistically observed in both the solar wind driving and the response of the three radiation belt electron populations to this driving. During the initial phase of storms V SW , V ρ , and P dyn rapidly increase and B Z becomes negative. B Z , V ρ , and P dyn reach extreme values during the initial 10.1002/2017GL076674

Geophysical Research Letters
phase and rapidly approach quiet values during the second phase. V SW peaks during the second phase of storms and subsequently slowly decays. In TRBEC the initial phase is characterized by a net decrease in TRBEC and the second phase is characterized by a net increase in TRBEC across nearly all K, especially evident in the relativistic populations. Figures 2 and 3 also show a clear dependence of TRBEC on the first adiabatic invariant μ. Enhancements in TRBEC are first observed at lower μ and later at higher values of μ.
Our observations demonstrate that statistically the global dynamics of the seed, relativistic, and ultrarelativistic populations in the outer radiation belt are not necessarily controlled by a delicate balance of loss and acceleration at any given point during a storm. Rather, these electron populations show a very well defined initial phase dominated by loss (characterized by a net decrease in TRBEC) followed by a short period of rapid acceleration (characterized by a net increase in TRBEC) followed by a transition to either a balance between loss and acceleration or a quiescent period with limited loss and acceleration. This is emphasized in the middle and right columns of Figure 3. Figure 3 also reveals a very clear coherency in the dynamics of the three radiation belt populations. There is evidence of a time-dependent response in K, smaller K responding first followed by large K; this suggests a pitch angle-dependent response in the recovery which may be mediated by pitch angle-dependent waveparticle interactions (e.g., Li et al., 2014) and will be investigated in detail in future work. Overall, our findings agree with and expand upon previous studies that have separately demonstrated a global coherence in different populations in the outer radiation belt. Kanekal et al. (2001) compared relativistic electron flux observed by Polar at high altitudes with SAMPEX at low altitudes near the footprint of magnetic field lines and found a remarkable global coherence in the dynamics of relativistic electron in the outer radiation belt (see also Chen et al., 2016). The work here demonstrates that the global coherence of radiation belt electrons exists across the entire outer radiation belt from lower-energy seed electrons up to ultrarelativistic electrons.
The statistical comparison between solar wind and outer radiation belt observations presented here points to a statistically likely and systematic sequence of events controlling storm time radiation belt dynamics. During the initial phase of storms negative B Z , enhanced dynamic pressure, and a rapid increase in solar wind velocity push the magnetopause inward (Shue et al., 1998;Sibeck et al., 1991) and drive enhanced magnetospheric activity in the form of ULF (Murphy et al., 2015) and VLF (Aryan et al., 2016) waves. The inward motion of the magnetopause leads to the rapid loss of radiation belt electrons via magnetopause shadowing (Ozeke et al., 2014; further enhanced by outward radial diffusion via ULF waves (Mann et al., 2016;Murphy et al., 2015;Ozeke et al., 2014) and precipitation driven by enhanced EMIC (Usanova et al., 2012) and VLF wave activity (Orlova et al., 2014).
During the second phase of the storm B Z , V ρ , and P dyn decay toward quiet values. As a result, the magnetopause begins to withdraw retreating away from the Earth and the outer radiation belt and loss via magnetopause shadowing is reduced. At this point, the outer radiation belt transitions from a period dominated by loss to a short period dominated by a rapid enhancement in the radiation belt seed, relativistic, and ultrarelativistic populations. During the second phase, V SW peaks and slowly decays, providing a mechanism for enhanced ULF wave power during the second phase of geomagnetic storms (Mathie & Mann, 2001;Pahud et al., 2009;Rae et al., 2012). In terms of radiation belt dynamics, the electron seed population is the first to recover during the second phase of storms. This recovery is likely via the injection of low-energy electrons via substorms driven by the release of nightside energy stored via reconnection and negative B Z during the initial phase of storms (e.g., Baker et al., 1998;Jaynes et al., 2015). These substorms also drive enhanced VLF (Meredith et al., 2004) and ULF Rae et al., 2011) activities. The recovery of the seed population of electrons and enhanced ULF and VLF wave activities subsequently leads to recovery of the relativistic and ultrarelativistic electron populations via radial diffusion (e.g., Li et al., 2017;Mann et al., 2016;Ozeke et al., 2012Ozeke et al., , 2017Shprits et al., 2005;Su et al., 2015) and local acceleration (e.g., Horne et al., 2005;Li et al., 2014;Reeves et al., 2013;Thorne et al., 2013). A key feature of the second phase of storms is the rapid enhancement of the three electron populations as opposed to a slow or gradual recovery. Any theory or modeling of radiation belt dynamics must be able to reproduce this rapid enhancement during storms. Future work will concentrate on studying the physical processes occurring during this phase of storms and attempts to distinguish between the causality of the possible mechanisms leading to this rapid enhancement in radiation belt electron fluxes.

Conclusions
Our superposed epoch analysis represents a refocusing of radiation belt research, demonstrating that statistically, storm time radiation belt dynamics throughout the entire outer radiation belt are repeatable. This repeatability is characterized by an enhanced solar wind driving which leads to rapid loss at the start of a storm followed by a rapid enhancement in the outer radiation belt seed, relativistic, and ultrarelativistic electron populations during storms. Future work will exploit this response and separation of electron dynamics into an initial phase dominated by loss followed by a second phase dominated by acceleration to attempt to quantify the causality and the importance of the role of various loss (e.g., Millan & Thorne, 2007; and acceleration processes (e.g., Elkington, 2006;Horne et al., 2006;Mann et al., 2012;Thorne, 2010) in the dynamics of electrons during geomagnetic storms as well as extend the analysis into solar minimum, a period dominated by CIR-driven storms as opposed to CME-driven storms as studied here.