Methods
The ESA Swarm mission, launched in late 2013, comprises three identical spacecraft equipped for making simultaneous, high accuracy and high cadence magnetic and electric field measurements. This study uses data from the Swarm A satellite, in a ~450 km polar orbit, equipped with the fluxgate magnetometer instrument [14] measuring magnetic field vectors at 50 samples/sec and the Electric Field Instrument [30] measuring ion velocity vectors at 16 samples/sec based on observed ion distributions from two sensors, which are converted into a 2 Hz E-field data product. Under the assumption of frozen flux of the observed ions based on ideal Ohm’s Law where E = -v x B the inferred velocity vectors can be converted into electric field vectors in the plane perpendicular to the magnetic field.
The automatic event identification algorithm used here has been designed to extract useful scientific data from synchronous magnetic and electric field measurements while addressing known caveats in the electric field data (e.g. uncertainties regarding offsets and magnitudes of the electric field instrument data [31], [32]). It is based on the methodology employed by [33]. Only magnetic latitudes of between +- 60 and +- 80 degrees magnetic latitude (MLAT) are considered for the analysis. This avoids low-latitude phenomena such as plasma bubbles [34] and localised extreme high latitude phenomena such as polar cap traversals [35]. The magnetic and electric field data is rotated into the mean-field aligned (MFA) frame, where z points along the direction of the mean magnetic field, x points towards the geomagnetic North Pole, and y completes the triad facing eastwards. A sliding 3-minute window is used for mean field calculation.
For event detection and selection, the Poynting flux is calculated by crossing the E- and B-field time series after applying a 2nd order Savitzky-Golay filter with window size 60.5 seconds, to remove any residual mean field influences or large-scale electric fields, as well as any uncertainties with the E-field baselines which are a known artefact in the electric field data. The modulus of this Poynting flux is then low-pass filtered with a 120.5 sec moving average filter to obtain its magnitude envelope. Where the magnitude of this Poynting flux envelope exceeds an empirically determined threshold, the event is flagged and event duration defined both backwards and forwards in time until the magnitude of the Poynting flux envelope drops below a second, lower, empirically determined threshold. For the analysis presented here, the thresholds are 25 and 8.75, respectively. This time window then defines a single event. Only E-field data sets flagged with the quality flags 1 (“use in consultation with EFI TII team at University of Calgary”) [36] is used in the analysis.
The 2 Hz E-field estimates are provided for both the horizontal and the vertical sensors. Based on the caveats described in [36] the outputs from both TII sensors are averaged. Since the analysis in this study used high-pass filtering to focus on relatively small scales, it is deemed acceptable to use the full 3-D vector for Poynting flux calculations. A separate test performed using only the along-track component of the E-field, which is identical in both sensors, to calculate the Poynting flux, also reproduced the observed northern preference.
The selected events must all occur at locations between 60 and 80 degrees magnetic latitude (MLAT) and event length must be at least 150 seconds long. The time series are extended with zeros for 30 seconds at the beginning and the end, which serves as padding to allow all filter sizes to fully capture the energy content in each event without edge-effect distortion. For the selected events, the electric and magnetic field data are then passed through a Savitzky-Golay filter of 2nd order and of various window sizes, from 1 (no effective high-pass filter) to 60.5 seconds, at 0.5-second intervals. The cross product of the two band-passed signals gives the Poynting flux in that frequency band. The Poynting flux is integrated over time along the spacecraft trajectory to obtain the integrated apparent energy flow through the satellite world-line as it crosses the event region. This is repeated for the entire range of low pass filter window sizes for each event. Average Poynting fluxes for each event are obtained by dividing the integrated Poynting flux values by the event duration. The median and quartiles are obtained for these three quantities from all of the events flagged by the algorithm.
A representative example of the analysis is shown in Figure 4 for an auroral zone Swarm A crossing event from 05:11:30 to 05:15:00 UT on 17 Nov 2016, flagged by the algorithm. It can be seen that there is good correspondence between B-field (panel (a)) and E-field (panel (b)) data, suggesting mostly positive (downwards) Poynting flux throughout. This is evidenced in panels (c) and (d) where the (parallel) Poynting flux remains largely positive. The Poynting flux shown in panel (c) is plotted for three Savitsky-Golay filter windows – 9 sec (blue) for small-scale phenomena, 27 sec (red) for mesoscales, and 47 sec (black) for perturbations associated with larger-scales. It can be seen that the Poynting flux reduces as progressively more low-pass filtering is applied to the constituent E-field and B-field time series. Panel (d) shows the time integrals of the fluxes in panel (c) demonstrating the accumulation of Poynting flux on the satellite’s world line as it crosses the perturbation region. It can be seen that the cumulative Poynting flux passing through the satellite during the event is approximately half the value for large scales as for meso- and small-scales, suggesting that large-scale perturbations associated with global field aligned current systems carry only half of the electromagnetic energy into the ionosphere during this particular event.