Observation of quiet-time mid-latitude Joule heating 1 and comparisons with the TIEGCM simulation

16 Joule heating is a major energy sink in the solar wind-magnetosphere-ionosphere sys-17 tem and modelling it is key to understanding the impact of space weather on the neu-18 tral atmosphere. Ion drifts and neutral wind velocities are key parameters when mod-19 elling Joule heating, however there is limited validation of the modelled ion and neutral 20 velocities at mid-latitudes. We use the Blackstone Super Dual Auroral Radar Network 21 (SuperDARN) radar and the Michigan North American Thermosphere Ionosphere Ob-22 serving Network (NATION) Fabry-Perot interferometer (FPI) to obtain the local night-23 side ion and neutral velocities at ∼ 40 ◦ geographic latitude during the nighttime of 16 24 July 2014. Despite being a geomagnetically quiet period, we observe significant sub-auroral 25 ion flows in excess of 200ms − 1 . We calculate an enhancement to the local Joule heat-26 ing rate due to these ion flows and find that the neutrals impart a significant increase 27 or decrease to the total Joule heating rate of > 75% depending on their direction. We 28 compare our observations to outputs from the Thermosphere Ionosphere Electrodynamic 29 General Circulation Model (TIEGCM). At such a low geomagnetic activity however, TIEGCM 30 was not able to model significant sub-auroral ion flows and any resulting Joule heating 31 enhancements equivalent to our observations. We found that the neutral winds were the 32 primary contributor to the Joule heating rates modelled by TIEGCM rather than the 33 ions as suggested by our observations. 34


Introduction
A significant fraction of the energy flowing through the magnetosphere-ionosphere system is lost to the atmosphere via Joule heating, which in the ionosphere-thermosphere system can be equated to frictional heating between charge carriers and neutral constituents within Earth's upper atmosphere (Vasyliunas & Song, 2005).Joule heating is the dominant magnetosphere-ionosphere energy input source, typically responsible for twice as much energy input compared to auroral power (Lu et al., 1996(Lu et al., , 1998;;Knipp et al., 2004;Lu et al., 2016) and up to 70% of the total ionospheric power input during geomagnetic storms (Knipp et al., 2004).This heating can cause ionospheric and thermospheric expansion (Rishbeth et al., 1969;Fuller-Rowell et al., 1997;Knipp et al., 1998;Lu et al., 2016; S. R. Zhang et al., 2017) which can result in enhanced ion outflow (Wahlund et al., 1992) and increased satellite drag that can reduce operational lifetime (Dang et al., 2022;Fang et al., 2022;Lin et al., 2022).It is therefore important that we understand the causes of Joule heating across all regions of the ionosphere.
Joule heating has been extensively studied at the high-latitudes (Kiene et al., 2019;Wang et al., 2020).Ion motion is controlled by magnetic reconnection between the interplanetary magnetic field (IMF) and Earth's magnetosphere, circulating due to ⃗ E× ⃗ B drift antisunwards at polar latitudes then returning sunwards at lower latitudes (Dungey, 1961;Cowley & Lockwood, 1992) under southwards IMF conditions.Motion of the neutrals in the thermosphere is driven by a combination of solar pressure gradients, coriolis forces and drag from ion motion (Rishbeth, 1977).Typically at high-latitudes, neutral velocities are small relative to the ion velocities such that Joule heating is primarily due to motion of the ions.High-latitude Joule heating calculations have therefore often discounted contributions from the neutrals.However during non-storm times and at lower latitudes, the velocities of the neutrals relative to the ions can be significant.
Using model simulations, Lu et al. (1995) calculated the neutrals to have an approximate 28% negative effect on Joule heating.Often the neutral velocities at mid-latitudes can exceed the ion velocities.Both Zou and Nishitani (2014) and Joshi et al. (2015) used Super Dual Auroral Radar Network (SuperDARN) data to show that neutral motion driven by expanded ⃗ E × ⃗ B ion drift due to intense geomagnetic storms can persist up to 20 hours after the recovery phase, resulting in neutral wind driven mid-latitude ion motion known as the disturbance dynamo effect.A study by Billett et al. (2018) focusing on the high-latitudes used a combination of SuperDARN and neutral wind model data to find that global Joule heating patterns have a significant dependence on UT due to neutral wind enhancements.Studies including the mid-lower latitudes and during periods of weaker geomagnetic activity must therefore include neutral wind contributions when calculating the Joule heating.
Joule heating is calculated as the dissipation rate of currents perpendicular to the magnetic field, ⃗ J ⊥ • ⃗ E (Lu et al., 1995) and the total Joule heating rate can be calculated with equation (1) (Baker et al., 2004), where ⃗ E = ⃗ −V i × ⃗ B is the electric field in the Earth's reference frame due to the ion motion, ⃗ V i , assuming a stationary neutral background, σ p is the conductivity in the direction of the electric field (Pedersen conductivity), V n is the velocity of the background neutrals and ⃗ B is the magnetic field strength.⃗ V n × ⃗ B accounts for the electric field generated by the neutral wind dynamo due to the drag imposed on charged particles in the ionosphere.
Equation 1 can be expanded into equation 2 (Billett et al., 2018) which conveniently breaks it down into the three terms that individually describe the main contributors to the total Joule heating.
Q i is the ion heating and is the heating that would be generated by ions moving against a stationary neutral background.Q w1 is the 1st wind correction term and accounts for the direction of the ions relative to the neutrals.If the neutrals and ions move in the same direction then the difference between their velocities is smaller and the frictional heating due to collisions between the neutrals and ions will be lower.Conversely, if they move in opposing directions the difference in their velocities will be greater and the heating will be larger, therefore Q w1 can act to either increase or decrease the total Joule heating.Q w2 is the heating that would be generated by the neutrals moving against a stationary ion background.Together, Q w1 and Q w2 create the wind correction term Q w , which describes the total heating accounted for by motion of the neutral wind relative to the ions.
Except for geomagnetic storms, where the twin cell ⃗ E× ⃗ B convection pattern expands to 40 • magnetic latitude (Walach & Grocott, 2019;Walach et al., 2021) from its high-latitude (> 60 • magnetic latitude) boundary, most ion flows at the mid-latitudes (40 • − 60 • magnetic latitude) are sub-auroral.At sub-auroral latitudes, ion motion is often associated with sub-auroral polarization streams (SAPS) (Clausen et al., 2012;Billett et al., 2022), penetrating electric fields (Maimaiti et al., 2018(Maimaiti et al., , 2019) ) and pressure gradient drifts (Hudson & Kelley, 1976;Greenwald et al., 2006;Liu et al., 2021)  Global circulation models are often used to study high-latitude ionosphere-thermosphere and Joule heating processes (Lu et al., 2016;Wang et al., 2020), however a lack of studies using global models focused on the mid-latitudes leaves some uncertainty in their reliability to provide accurate mid-latitude modelling.We therefore compare our ion and neutral observations and Joule heating estimations with equivalent outputs from the Thermosphere Ionosphere General Circulation Model (TIEGCM, see section 2).This paper is split into the following sections: Section 2 details the observed and modelled data used in this study, Section 3 provides an overview of the geomagnetic conditions and observations made during the night of 16 July 2014.Section 4 details the methods used to estimate the Joule heating rate while presenting the results of those estimations.Finally, section 5 discusses the results in context of the wider literature and scientific community.
2 Parameters and Models

Ion Motion
The Super Dual Auroral Radar Network (SuperDARN) (Greenwald et al., 1995;Chisham et al., 2007;Nishitani et al., 2019) is a series of high frequency radars in the northern and southern hemispheres that provide observations of ionospheric dynamics across high and mid-latitudes.In the northern hemisphere, SuperDARN comprises radars which have near total hemispheric coverage of the polar, high-latitude and mid-latitude regions.For this investigation we consider data from the Blackstone (BKS) radar due to its field of view (FOV) overlapping the FPI used in this study (see section 2.2).Each radar can electronically steer its look direction, centred on which it forms a beam typically 3 • wide and consisting of 75-100 range gates with a 45km range resolution.Each radar can sweep through 16-24 beams with the FOV being roughly 50 • where a full azimuthal scan across all beams takes 1-2 min.
SuperDARN radars detect field-aligned plasma irregularities in the E and F regions of the ionosphere by recording the backscattered signal from decameter scale electron density structures.Plasma irregularities in the F region drift with ⃗ E× ⃗ B velocities and their Doppler shift can be used to infer properties of the ionosphere.Due to refraction, the radar beam can reflect off of the ground, known as groundscatter.Groundscatter is typically characterised by a velocity of only a few ms −1 and produces a low spectral width, which is often sufficient to distinguish between ionospheric and ground scatter at the highlatitudes.At the mid-latitudes however, and particularly during periods of low geomagnetic activity, ionospheric scatter can often be much slower while exhibiting low spectral widths and these techniques can often eliminate observations of relevant ion motion.
Instead we use the algorithm developed by Ribeiro et al. (2011Ribeiro et al. ( , 2012) ) which has been specifically designed for identifying mid-latitude ionospheric scatter.The algorithm uses a 3 × 3 × 3 beam by range gate by scan boxcar filter that identifies individual clusters of scatter connected by range gate and scan and determines the ratio of fast to slow moving scatter within each cluster.Errors associated with the filtered velocities are derived using the method described by Ruohoniemi and Baker (1998), however we have modified the method so that velocities are removed if they are two median absolute deviations (Howell, 2005) from the median instead of two standard deviations from the mean.
This reduces the impact of unphysical outliers that result in excessive standard deviations due to the lower velocities associated with the mid-latitudes.Since an individual cluster of returned scatter can usually be attributed to either ionospheric backscatter or groundscatter, the algorithm identifies and marks which clusters contain ionospheric scatter.The Ribeiro et al. (2011) algorithm also automatically excludes backscatter from ranges within 315km of the radar to eliminate scatter originating from the E-region and meteor echoes at near ranges.Furthermore, in this study we modify the algorithm such that it can consider clusters spanning multiple beams similar to A. G. Burrell et al. (2018).
This whole approach enables a quantitative ionospheric/groundscatter classification of mid-latitude backscatter.

Neutral Wind Motion
The North American Thermosphere Ionosphere Observing Network (NATION) (Makela et al., 2012) was a network of five Fabry-Perot Interferometers that observed the neutral wind velocity and temperature in Earth's thermosphere across the mid and eastern parts of the United States of America.Each FPI observes the Doppler shift of the 630nm OI emission line that is assumed to peak at an altitude of 250km.The FPIs scan at an elevation angle of 45 • and take measurements in the geographic cardinal directions (north, east, south, west) and the Zenith through the nighttime period.Data are analyzed using the techniques described in Harding et al. (2014) to produce estimates of the horizontal neutral winds at ∼ 250km altitude.
Of the five FPIs in NATION, and assuming that the cardinal measurement locations are located at the peak emission altitudes, we only use data from the Michigan (ANN) instrument due to it being the only FPI that has all of its measurement locations intersecting with the BKS radar's FOV.

Geomagnetic Field
The 13th generation International Geomagnetic Reference Field (IGRF) model (Alken et al., 2021) profiles the Earth's tilted dipole as a function of time, geographic position and altitude.At high-latitudes, the magnetic field is mostly vertical, however at midlatitudes and lower, there can be a significant tilt to the angle of the field that needs to be accounted for when comparing the electric field generated by the neutral wind dynamo ( ⃗ V n × ⃗ B) with the electric field that is calculated from ⃗ E × ⃗ B drift, due to the difference in angle between the ion and neutral velocity vectors.The IGRF13 model provides the declination and inclination of the magnetic field at the ANN FPI's location, which at 250km altitude are −6.49• and 69.26 • respectively.This inclination and declination do not change significantly between assumed peak emission locations so we do not consider it in our calculations.

Auroral Boundary
Due to the mid-latitude location this study focuses on, it is important to identify if any observed ion flows are sub-auroral.It has been shown that the boundary between the region 1 and region 2 currents serve as a good approximation for the extent of equatorward boundary of the auroral oval, particularly on the duskside in the northern hemisphere (Kilcommons et al., 2017).The Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE) uses magnetometers on the Iridium constellation of telecommunication satellites to provide field aligned current measurements across both hemispheres.A spherical harmonic fit to the measured radial current densities alongside Ampere's law (B.J. Anderson et al., 2000;Coxon et al., 2018) produces 10-minute cadence global current density maps which can be used to determine the location of the Region 1 and 2 currents, where we use the boundary between them as a proxy for the boundary of the auroral oval.

Total Electron Content
The mid-latitude ionospheric trough plays a significant role in ionospheric processes in the mid-latitudes (P.C. Anderson et al., 1993;Kunduri et al., 2021;Liu et al., 2021) and can be identified using Total Electron Content (TEC) measurements from Global Positioning System (GPS) data.We use the TEC data to further investigate the midlatitude dynamics through identification of the trough.The TEC data are processed using the algorithms from Rideout and Coster (2006) and Vierinen et al. (2016).The TEC data are placed into 1 • ×1 • geographic latitude by geographic longitude cells integrated over 5 minutes.Furthermore we median filter the TEC data as described by Thomas et al. (2013) to reduce the geospatial noise among the dataset.

TIEGCM
The Thermosphere Ionosphere Electrodynamic Circulation Model (TIEGCM) (Richmond et al., 1992;Qian et al., 2014) is one of the most widely used thermosphere/ionosphere models within the upper atmospheric scientific community and is a fully three-dimensional time dependent model of Earth's ionosphere and thermosphere that solves the equations of continuity, energy and momentum for the three major ion and neutral species.
TIEGCM uses either the Weimer (Weimer, 2005) or Heelis (Heelis et al., 1982) electric field models as a driver for the ⃗ E× ⃗ B driven high-latitude ion convection and can operate using a 5 • × 5 • or 2.5 • × 2.5 • resolution in latitude and longitude.Wu et al.
(2017) compared TIEGCM's high-latitude thermospheric winds and ion drifts using the two electric field models to observational data and found that using the Weimer model produces more accurate simulations.In this study we run TIEGCM with the Weimer electric field model at a resolution of 2.5 • ×2.5 • .When using the Weimer electric field model TIEGCM takes the f10.7 solar radio flux, IMF By, IMF Bz, solar wind velocity and solar wind density as input drivers for the model.
TIEGCM produces estimates of the geographic meridional, zonal and vertical ion and neutral velocities at the specified run resolution.Outputs from TIEGCM in this study are taken at an altitude of 250km, which corresponds to the altitudes of the ion and neutral observations.As the ion and neutral line of sight velocities show little difference between assumed peak emission locations, we simply use the TIEGCM ion and neutral velocities located at the FPI location.
TIEGCM also models the Pedersen conductivity, which we use for calculating both the observational and modelled Joule heating values.Keeping the Pedersen conductivity consistent between observed/modelled methods allows us to better isolate the effect from differences between the observed and modelled ion and neutral velocities, which is the aim of this study.

Event Overview
In this section we present observations from 16 July 2014.First we present an overview of the geomagnetic conditions, before looking in detail at the ion and neutral velocity observations.

Geomagnetic Conditions
Geomagnetic conditions during this interval were quiet.Figure 1 shows the interplanetary and geomagnetic conditions, IMF By, IMF Bz, SYM-H, ASYM-H and Kp during the hours 0000-1000 UT.Panel 1a shows the solar wind speed while 1b presents the y and z components of the IMF.Of particular note is the slight negative IMF Bz between 0400 and 0700 UT indicating a southwards directed IMF, allowing magnetic reconnection to occur between the IMF and Earth's magnetic field.Panel 1c shows the auroral indices with significant enhancements to the AL, AU and derived AE index coincident with the period of southwards IMF, indicating an increased intensity of the auroral electrojet.Similarly, the SYM-H index (1d) shows an increase in ring current intensity from 0500 UT while the ASYM-H index (1e) shows an increase in asymmetries in the ring current.Figure 1f shows the Kp index of between 1.7 and 2.3, which indicates minor geomagnetic activity during this period.UT (for the east observation point) and 0700 UT (for the south observation point).after 0500 UT.The north facing observations especially, show a large increase in magnitude up to a peak value of −150 ms −1 whereas the south facing observations peak at roughly −80ms −1 .The zonal velocities show similar magnitudes with the East look direction peaking at −30ms −1 and the west look direction at −40ms −1 .Although in our ANN data, the south observations stop after 0630 UT and a 3 hour data gap occurs in the east observations between 0400 and 0700 UT due to the presence of the moon in that direction, the trend in the data between the opposing observation points are similar enough that we can assume that there are no significant changes in the spatial distribution of the neutral wind flows over the ANN FPI.The vertical velocities are shown to fluctuate highly relative to their greatest magnitude, ranging from values of 5ms −1 to −20ms −1 .

FPI Observations
When comparing TIEGCM's output, we find that the modelled meridional velocities are similar in magnitude to the observations.At the north location, TIEGCM's neutral velocities are closely aligned with these observations.TIEGCM's velocities at the southern location follow a similar pattern and although the observations stop after 0600 UT, the trend of increasing magnitude from 0200 UT is apparent in both the observed and modelled data.The meridional velocities then match well with our observations.Both TIEGCM's velocities at the east and west locations are shown to have a large difference to the observations.Although the trend of a somewhat sinunsoidal variation in both the observed and modelled neutral velocities between 0000 UT and 0900 UT, peaking at 0500 UT, are somewhat similar, TIEGCM's velocities are significantly faster, especially at the peak.TIEGCM is more accurate when estimating the mid-latitude meridional neutral flows than in the zonal direction in this case.

Methodology
Estimation of the Joule heating over the region requires at least the two-dimensional ion and neutral velocities for use in Equation 1.We use a technique similar to L-shell fitting (Villain et al., 1987;Ruohoniemi et al., 1989) which has been used for other midlatitude ion studies (Clausen et al., 2012;Kunduri et al., 2018;Maimaiti et al., 2018Maimaiti et al., , 2019)), whereby the ion motion is assumed to be constant across some area.If such a flow is observed by a SuperDARN radar then the line of sight velocities vary azimuthally across the beams such that if a beam crosses the flow perpendicular to the flow direction it will return with a zero velocity.Conversely, if the beam is sounding in the direction parallel/antiparallel to the full flow then it will return its full velocity.We can fit a cosine curve to the line of sight velocities against beam azimuth, where the magnitude of the fit provides the full 2D ion flow perpendicular to the magnetic field.
Initially a pre-defined area around the FPI was used where velocities in that area were selected and fit to a cosine curve, however this resulted in poor fits as it became apparent that there were multiple flow patches within the FPI region.Therefore, in order to accurately capture the dynamic ion motion over the FPI, we manually identified the individual ionospheric scatter patches, first by time-integrating the scans over periods of 10 minutes to reduce temporal variability and then by marking the boundaries of each patch spatially and temporally.We then selected the highest magnitude velocity from each beam within each defined patch and fit to those points.A minimum of five unique beams were used to constrain the fits, which although is less than used in other studies, e.g.Thomas and Shepherd (2018); Kunduri et al. (2018), manual (rather than automated) selection and review of the points ensures that they are still constrained to the fit.This further allowed deselecting beams at the sides of patches if by inspecting the fits it became clear that part the flow does not belong to the patch, ensuring that motion only belonging to that patch was captured.an azimuth range of only 10-15 • .If both of these were included in the same fit, as would occur with a static fitting area, we would be unable to accurately fit it to a sinusoid.This analysis shows that points contained within C and D are therefore part of two separate patches of scatter.Therefore, by including the velocity-azimuth points and their fit results in the patch determination process, we ensure that individual patches are accurately tracked.
By taking the fit magnitudes throughout the interval we are able to estimate the two dimensional ion flow during this period.Then by applying equation 2 with the IGRF magnetic field strength we can calculate the total local Joule heating across this interval.At high-latitudes, the quasi-vertical magnetic field results in the ion drift travelling approximately parallel to the Earth's surface in the same plane as the neutrals.However, since the magnetic field is inclined (69.26 • to the horizontal) at this latitude, the ions and neutrals reference planes are instead inclined roughly 20 • relative to each other.
The BKS radar only measures the ion velocity in the component perpendicular to ⃗ B, it is therefore necessary to mention that the derived fitted velocities may not be the representative of the full three dimensional ion velocities, but only the two dimensional velocities perpendicular to the magnetic field inclination.Due to the magnetic field inclination, it is also important that calculations of the coupling between the ions and neutrals are made in the same plane relative to each other.As there is no estimate of the ion velocity in the direction parallel to the magnetic field, it is not possible to calculate the ion velocity in the plane horizontal to the Earth's surface.Since the FPI observes the geographic horizontal and vertical directions, the neutral wind velocity is obtainable fully in three-dimensions.By applying the three-dimensional rotation matrix transformation given by equation 3, where N Bz , N Bx and N By are the neutral wind components in the z, x, and y directions in magnetic field aligned coordinates respectively, and assuming the magnetic field is entirely in the z direction.N mer , N zon & N ver are the geographic meridional, zonal and vertical neutral wind components, θ is the angle subtended by the great circle lines connecting the FPI location to the geographic and magnetic north pole, while ϕ is the angle subtended between the magnetic field and the plane horizontal to Earth's surface at the location of the FPI.
As the observed neutral velocities are in the line of sight direction (45 • inclination) of the FPI, we calculate the horizontal components of each cardinal observation, W h , using equation 4 (Makela et al., 2012), where W LOS is the line-of-sight Doppler velocity, W v is the vertical neutral velocity and α is the elevation angle of the line of sight measurements.We assume that the zenith velocity measured above the FPI is consistent across the cardinal locations and so use that as the vertical velocity.The signs of the velocities are then changed such that positive velocities are directed northwards (meridionally).Because the FPI observations are made at different times with irregular cadences for each cardinal location, we linearly interpolate the observations so that the cadence of opposing east/west observations match.
If only one opposing cardinal measurement is available at a given time, such as after 0600 UT where no southwards observations were taken, we use the measurement we do have as the full meridional/zonal flow.

Results
Figure 7 shows IMF Bz (a) followed by the estimated observed (blue) and modelled (orange) magnitude of the full neutral wind vector (b). Figure 7c shows the fitted two dimensional ion velocities for each identified patch compared with those modelled by TIEGCM, while (d) shows the Pedersen conductivity, σ p at the FPI location as modelled by TIEGCM and used for calculating the Joule heating.The solid lines represent the magnitude of the fitted velocity while the shaded region is the RMSE error of the fits.The largest RMSE is less than 20ms −1 , which given that the two major patches (B and C) are always at least 100ms −1 indicates that the fits to determine the two dimensional ion velocities are excellent.The time boundaries for the plot have been restricted to between 0400 and 0800 UT since no significant ion patches were identified either side of these times.The observed neutral's speed is seen to steadily increase from ∼ 40 to ∼ 200ms −1 over the course of the night.TIEGCM overestimates the neutral velocities prior to 0600 UT, however afterwards, the total velocity magnitude is in line with the observations.Of the identified ion patches, two take precedence, patches B and C. Patch B appears at 0400 UT with velocities of 100ms −1 , increasing to in excess of 250ms −1 at 0500 UT before dissipating.Patch C starts at 0500 UT, hovering at between 100 and 200ms −1 until it also dissipates at 0700 UT.It is worth noting that the patches were only marked if they were at least covering part of the region within the FPI measurement locations.It is likely that the patches originated or dissipated outside of this area and merely traversed through the region over the FPI, we have only noted the times where the patch is contained within the FPI region.Furthermore, patch A was identified to occur between 0350 and 0420 UT, however the azimuthal span of the patch was not enough to satisfy the conditions we set in section 3 to fit a two-dimensional velocity, hence it is missing in this and further presentations of the patch ion velocities and Joule heating.TIEGCM's ion velocities remain a fairly steady 20−40ms −1 throughout the interval.As TIEGCM is a global large-scale model, it lacks the micro/mesoscale physics to capture the ion irregularities that produce the ion drift patches observed by the BKS radar, as evidenced here.The Pedersen conductivity is modelled to be relatively constant, although decreasing throughout the nighttime period.
Using the estimated two dimensional ion and neutral velocities, we have calculated the Joule heating rate and its components for each patch, assuming an altitude of 250km, plotted in Figure 8. Panels a, b, c, d and e show the Joule heating rate and components for patches A, B, C, D and E respectively.The blue line represents the ion heating rate, Q i , the orange and green lines the two wind correction terms, Q w1 and Q w2 respectively and the red line the total Joule heating rate of the patch, calculated as the sum of the three components.A negative Q w1 indicates that the direction of the ion and neutrals were aligned with each other, resulting in fewer collisional interactions and thus dampening the overall heating rate, a positive value indicates the ions and neutrals were opposed.Panel f shows the same components, but as modelled by TIEGCM.The final panel, g, compares the average total Joule heating rate from all the patches over the FPI region, with the total heating rate modelled by TIEGCM.The TIEGCM Joule heating has been calculated by using its ion and neutral velocities for use in Equation 2, while keeping the other parameters the same as the observations, therefore the only difference between the observed and modelled Joule heating rates, are the observed and modelled ion/neutral velocities.Based on the measurements, the most significant heating rate occurred between 0430 and 0500 UT which resulted from ion motion from patch B, with heating due to ion motion, Q i , peaking at roughly 236pWm −3 out of its total heating, Q j , of 237pWm −3 .
Patch C also exhibited some enhanced ion heating at 49.1pWm −3 , however this is somewhat lower than the severe excitations in patch B.
The positive magnitude of Q w1 in patch C and D indicates that the ions and neutrals were opposed in direction for most of each patch and so collisional interactions were increased.This increase is greatest in patch D at ∼ 0520 UT, where the total heating is increased by 47pWm −3 to 109pWm −3 , resulting in a 78% increase due to the ion-neutral directions.The impact of this term is further shown in patch D, where at ∼0550 UT, Q w1 is its most negative value and results in decreasing the total Joule heating by ∼ 76%.
The heating directly due to the neutrals Q w2 is low throughout the entire interval, hovering at around ∼ 5pWm −3 , therefore despite the large influence of the neutral wind direction on the total heating rate, the overall Joule heating magnitude is only significantly enhanced by ion motion.
The Joule heating enhancement observed in panel g at 0500 UT that peaked with a magnitude of 235pWm −3 is nearly 8 times higher than the TIEGCM modelled Joule heating of 30.8pWm −3 .Aside from ± 20 minutes of the 0500 UT peak, the TIEGCM Joule heating rate is significantly higher than the observational estimate.When we investigate the reason behind the heating we find that the larger magnitudes in the observed ion heating, Q i , especially in patch C, indicates that the observed heating is due to ion motion.Panel f shows that the magnitude of the total heating in TIEGCM is due to faster motion of the neutral winds, Q w2 , whose significance increases throughout the interval compared to the ion contribution.During this event, we find that our observations disagree with not just the magnitude of the modelled Joule heating rate at the midlatitudes, but the modelled Joule heating being from greater neutral wind motion disagrees with that calculated from our observations.

Discussion
The strong westwards driven ion flows that are observed in Figure 3 Kunduri et al., 2018).Kunduri et al. (2021) also found SAPS latitudinal distribution to correlate strongly with the ionospheric trough, which during this interval lies poleward of the FPI, suggesting that it would be unlikely for the flows to be a SAPS.
Confirmation of the ion flows (not) being part of SAPS would require ion flux measure-ments from satellite observations (Grocott et al., 2011;Clausen et al., 2012;Kunduri et al., 2017Kunduri et al., , 2018)), however coincidental measurements were not available for the interval of this study.Instead we compare with findings from Kunduri et al. (2017), which studied the latitudinal distribution of SAPS with correlation to the DST index.At 0600 UT (0030 MLT at the FPI location) the DST index was -1, which according to Kunduri et al. (2017), would place the mean SAPS position at 61 • magnetic latitude (∼ 51 • geographic) with a minimum of 59 • magnetic latitude (∼ 49 • geographic), which would still be at least 7 • poleward of the FPI location.Furthermore Nagano et al. (2015) calculated a quantitative estimation of the lower latitudinal boundary for SAPS keyed by SYM-H, which during this interval reached a minimum of -20nT.According to Nagano et al. (2015) this would result in a lower latitude boundary for SAPS of ∼ 58 • magnetic latitude (∼ 48 • geographic), still poleward of the FPI.We therefore suggest that the observations during this interval are not likely due to SAPS.If the ion enhancements are not due to high-latitude convection or to a SAPS, they may instead be part of a persistent quiettime mid-latitude nighttime feature (Greenwald et al., 2006;Clausen et al., 2012) that appears due to pressure gradient instabilities often found at the equatorward boundary of the ionospheric trough (Hudson & Kelley, 1976;Greenwald et al., 2006;Liu et al., 2021), which would align spatially with our observations.While we suggest the ions to be primarily responsible for the increased Joule heating rates, we might expect currents to be present in the region of those Joule heating enhancements.However, there is no evidence of field-aligned currents at the FPI location, which is concerning and could be investigated further in future works.We do provide one possible explanation for the lack of FAC's in the region, which could be due to ions and electrons flowing in the same direction at 250km altitude, potentially preventing any current from being produced.
Previous studies investigating mid-latitude nighttime ionospheric scatter have found ion velocities typically less than 100ms −1 (Greenwald et al., 2006;Maimaiti et al., 2018Maimaiti et al., , 2019)).They are often attributed to penetrating electric fields, driven by the neutral wind dynamo or due to pressure gradient forces.Given the magnitude of Q w2 is small compared to the total Joule heating rate, Q j , in this study, we infer that the ions are responsible for driving the increased Joule heating rate.Maimaiti et al. (2018Maimaiti et al. ( , 2019) ) carried out statistical studies of the nightside mid-latitude and sub-auroral ionospheric convection and found persistent westward flows between 20−90ms −1 depending on season and MLT, which is somewhat slower than our results, particularly as they found that the fastest flows occured in winter.Although Maimaiti et al. (2018) used the same groundscatter algorithm (Ribeiro et al., 2011) as in this study to remove low velocity non-ionospheric scatter, they also deployed the additional technique as described in (Ribeiro et al., 2012), where events were only considered if the 3rd and 97th percentile of their ion velocity distributions were greater than −120ms −1 and less than 120ms −1 respectively.This ensured that they only studied the quiet time mid-latitude nighttime scatter, however rare fast events may have been lost.By selecting active patches in this study, we have not considered low-velocity ionospheric scatter during this event, this will have skewed our velocities to a higher range than theirs.We believe that the higher ion velocities estimated in this study are therefore reasonable.Furthermore, despite our ion velocities being greater than other quiet time studies, they are significantly slower than other mid-latitude studies that occur under geomagnetically active periods.When enhanced ion velocities have been observed due to the equatorward expansion of auroral convection (Joshi et al., 2015) or SAPS (Clausen et al., 2012;Billett et al., 2022) velocities are observed in excess of 500ms −1 and up to 1000ms −1 .Our observed ion velocities therefore fall within a reasonable expectation when considering the geomagnetic activity and methods used in this study.
Strong ion motion has been shown to drive the neutral atmosphere into a similar direction as momentum is exchanged through frictional collisions.During both patches C and D the directions of the ions and neutrals are initially opposed, resulting in an increased Joule heating rate, however as both patches persist, the neutrals are slowly driven into the same direction as the ions, given by Q w1 decreasing.When the ion driving to the neutrals is at its greatest Q w1 would reach its peak negative value, and start to increase once the ion driving recedes and the neutrals retain momentum and start to drive the ions.In our observations Q w1 continues to decrease and never reach a negative peak over the tracked lifespan of both patches, with patch B lasting ∼ 1.5 hours and patch C ∼ 2 hours, suggesting the ions continue to drive the neutral motion throughout the period where we track them.Joshi et al. (2015) calculated the mid-latitude ion neutral coupling timescale during a geomagnetic storm and found a time-lag of ∼84 minutes for the neutrals to respond to the ion driving.Billett et al. (2022) found a response time of 2h for mid-latitude neutral wind to respond to pressure gradient forces.In the case of Joshi et al. (2015), ions were driven by expanded auroral convection during a geomagnetic storm, and for Billett et al. (2022) a SAPS event, with ion velocities several 100ms −1 faster than this study's quiet time events.Kosch et al. (2001) (Billett et al., 2018) as shown by Lu et al. (1995), or by assuming that F-region altitude measurements map down to a range of altitudes (Cai et al., 2014) ing rates up to the order of nWm −3 for geomagnetically active intervals, an order of magnitude higher than our observations.The majority of their observations however were in the tens, or hundreds of pWm −3 , which matches our observations, suggesting small patches of ion scatter at mid-latitudes are able to produce local Joule heating enhancements similar to those observed at high-latitudes.Their most dominant Joule heating values were coincident with the auroral region, where ion velocities are typically much higher, often in excess of 1000ms −2 , particularly during geomagnetically intense periods (such as in their studies).While we could compare our values with studies calculating height-integrated Joule heating rates by assuming that the electric field maps to lower altitudes, the neutral wind measurements however do not, and doing so would introduce significant uncertainty into our calculations that we have attempted to avoid by keeping the ion and neutral measurements as co-located as possible.Nethertheless, they can be used as an insight into the difference between auroral and sub-auroral Joule heating rates, which typically indicate higher magnitudes in the auroral region, with the difference of at least an order of magnitude being fairly common (X.X. Zhang et al., 2005;Lu et al., 2016;Billett et al., 2018).Our values being an order of magnitude smaller than those in the high-latitude studies is reasonable.If we consider the fact that the high-latitude studies occured during geomagnetically intense periods, while our mid-latitude study is during a quiet time period, our Joule heating values may be closer than expected, indicating that even small transient events can result in a significant Joule heating deposition in the mid-latitudes.
Baloukidis et al. ( 2023) compared statistical high-latitude Joule heating distributions estimated by using the European incoherent scatter scientific association (EISCAT) radars with TIEGCM.Their EISCAT Joule heating estimations ranged from altitudes of 80 − 150km altitude and did not include the neutral wind contributions, so are not directly comparable to our estimations in this study, but their comparisons to TIEGCM are still useful.They found that during low Kp, TIEGCM's modelled Joule heating was higher than their observed estimates.If we can assume that fast moving ion patches were averaged out in their low Kp statistical analysis, then our results of TIEGCM modelling higher Joule heating during low velocity ion events agrees with the findings from their study.At higher Kp, Baloukidis et al. (2023) also found TIEGCM's observed Joule heating was lower than their observed estimates.Although our study is a low Kp event, our periods of significant ion enhancements are more often associated with high levels of geomagnetic activity, so we can compare our fast moving ion patches to their high Kp analysis, whereby we also agree that TIEGCM's modelled Joule heating is lower than observed estimations.Similar to our findings, Baloukidis et al. (2023) remark that the difference in their discrepancies between TIEGCM and their observations are due to smallscale effects that amount to sub-grid variability within TIEGCM that it cannot resolve.
Due to this sub-grid variability, TIEGCM includes an empirically-derived multiplication factor named JOULEFAC to increase its internal Joule heating by a fixed factor of 1.5 (NCAR, 2016) in order for its neutral temperatures to better agree with statistical observations.One solution Baloukidis et al. (2023) propose is to adjust JOULEFAC with Kp so that different values are used for different levels of geomagnetic activity.Previous studies have manually adjusted the value of JOULEFAC to better reproduce realistic Joule heating values (Emery et al., 1999).Although there may be differences between optimised JOULEFAC values for high and mid-latitudes, optimised JOULEFAC values may work on a statistical level, however it could not account for small scale spatial or temporal events such as in this study.A better JOULEFAC for low Kp may bring TIEGCM's modelled Joule heating in line with our observed estimations for low veloc-ity patches, however there would still be a large and potentially greater difference for excited ion motion, such as patch B between 0430 and 0500 UT in this study.Furthermore, adjusting JOULEFAC may "correct" the numerical Joule heating value, however it might not solve discrepancies between whether greater ion or neutral motion produces Joule heating as occurs in this study.Rather, if focusing on localised studies, improvements should be made for TIEGCM to better model the microscale electrodynamics of the midlatitude ionosphere.

Conclusion
During the night of 16 July 2014 over mid-latidude North America the BKS Su-perDARN radar observed highly localised ion velocity enhancements of over 200ms −1 while ANN FPI observed neutral velocities over 150ms −1 despite the lack of strong geomagnetic drivers.The use of combined AMPERE and TEC datasets shows the ion enhancements are sub-auroral, and likely driven by plasma gradient instabilities, a common quiet-time nighttime mid-latitude occurrence observed at the equatorward edge of the mid-latitude trough.The ion velocity increases drove significant Joule heating enhancements to the region, of a similar magnitude to results from high-latitude studies, with the maximum increases only a single order of magnitude less than under high-latitude geomagnetically active periods.The neutral wind was shown to have a significant impact on the overall heating rate, accounting for on average between 24% and 43% of the total heating, while at the extremes increasing or decreasing the total heating rate by in excess of 75%.
Comparisons with modelled ion and neutral velocities from TIEGCM indicate that TIEGCM does not model equivalent enhancements to the ion velocities due to being a large-scale model that does not include microscale electrodynamical processes, resulting in an approximate 8 times smaller modelled Joule heating rate than during the peak observed estimates.Although TIEGCM does a good job of modelling the meridional neutral velocities, the zonal velocities were an order of magnitude higher than our observations, enough to amplify the total neutral wind velocity such that the mid-latitude Joule heating reported by the model was due to greater motion of the neutrals rather than the ions as our observations suggest.The strong neutral wind in the model also resulted in a greater modelled Joule heating rate than our observational estimates during quieter periods of the interval.
Opportunities for studying mid-latitude ion-neutral coupling and the Joule heating response are rare and limited intervals exist with measurements from coincident instruments, particularly during quiet times to study such events, nevertheless further work is needed to better understand the dynamics of the mid-latitude ionosphere-thermosphere, especially during non-geomagnetically intense periods.Further understanding and better representation of the mid-latitude dynamics could help produce more accurate models for Joule heating predictions.

Open Research Section
All data used for this study are available from open-source from nonprofit organisations.The authors acknowledge the use of SuperDARN data.SuperDARN is a collection of radars funded by national scientific funding agencies of Australia, Canada, China, France, Italy, Japan, Norway, South Africa, United Kingdom, and United States of America, and we thank the international PI team for providing the data.The authors acknowledge access to the SuperDARN database via the British Antarctic Survey (https:// www.bas.ac.uk/project/superdarn/data).Other data mirrors are hosted by the Virginia Tech SuperDARN group (http://vt.superdarn.org/) and the University of Saskatchewan (https://superdarn.ca/data-download).The radar data products used are the FI-TACF3.0library and version 5.0 of the Radar Software Toolkit (RST) (Thomas et al., that are mostly responsible for driving subauroral ion flows.Billett et al. (2022) observed significant equatorward and westward neutral wind disturbances during a SAPS event.They found that the response of the neutrals close to the SAPS was almost immediate and was likely driven by ion neutral coupling.They did however also find neutral disturbances further from the SAPS after a 2 hour time lag that they propose were due to pressure gradient and Coriolis forces from the SAPS heating.The difference in neutral response due to location indicates the importance of considering the mesoscale structure of ionospheric events when accounting for neutral particle motion.Furthermore a study byKiene et al. (2019) found that the difference in high-latitude Joule heating rates varied by as much as a factor of 10 due to local variations in the observed ion-neutral structure.If we are to accurately estimate the Joule heating rate during ionosphere-thermosphere disturbances it is then necessary to ensure that the ion and neutral measurements are as colocated as possible.The Blackstone (BKS) Super Dual Auroral Radar Network (SuperDARN) radar and the Ann Arbor (ANN) North American Thermosphere Ionosphere Observation Network (NATION) interferometer are two mid-latitude ground-based instruments, used for observing ion and neutral flows in the F-region of the ionosphere respectively, (see section 2) whose fields of view (FOV's) overlap each other, allowing for colocated observations of mid-latitude ion and neutral flows.As the majority of ion motion at the midlatitudes occur during quiet periods we search for quiet time coincident ion and neutral observations.Identifying times of high-quality colocated observations during quiet times is extremely difficult.Both instruments need to be operational, the BKS radar needs to observe ionospheric scatter during the nighttime period in the region over the FPI and the ANN FPI needs to have suitable (uncloudy) observation conditions.Intervals where all these requirements are satisfied are unfortunately rare.Nevertheless we have identified the nighttime of 16 July 2014 as a period where all the necessary conditions are met, allowing us to study the local quiet-time mid-latitude Joule heating during this interval.

FigureFigure 1 .Figure 2 .
Figure2ashows the IMF Bz followed by panels presenting measurements of ionospheric ion velocities for selected beams of the BKS SuperDARN radar between 0000-1000 UT, specifically beams 15 (b), 17 (c), 7 (d), 9 (e) and 17 (f) which are the beams that intersect through the FPI north, east, south, west and zenith assumed peak emission locations respectively.Negative velocities indicate line of sight ion motion away from the radar and positive velocities towards the radar.The velocity magnitude is given by the colorbar on the right.Portions of the observations that have been determined to be groundscatter according theRibeiro et al. (2011Ribeiro et al. ( , 2012) ) algorithm have been marked in grey.The horizontal dashed lines across each beam range gate panel show the range gate where the assumed peak FPI emission point is located, calculated using the standard Su-perDARN virtual height model.Across all beam range gate panels we observe enhancements of the ion velocities during the southwards IMF Bz interval between 0400 and 0700

Figure 3 Figure 3 .
Figure 3 corresponds to 0600 UT (0100 local time at the ANN FPI), chosen due to a strong westwards flow in the south-west region of the FPI area.By tracing a westwards line that starts at 40 • N ∼ 85 • W and finishes at 45 • N 105 • W, we can see that the strong ion velocities close to the FPI persist through multiple ionospheric scatter ranges and into the FOV of more westwards located radars.The AMPERE dataset indicates that the R1/R2 boundary at midnight is between 60 • and 50 • geographic latitude, approximately 20 • poleward of the FPI, we thus conclude that the observed flows are subauroral.From the TEC data we can see the formation of the ionospheric trough equatorward of the region 2 currents and poleward of the FPI, starting at ∼ 50 • geographic latitude at local midnight and wrapping around to ∼ 70 • geographic latitude at the duskside.A more detailed presentation of the ion flow data from the BKS radar overlooking the ANN FPI is provided by figure4, which shows the line of sight velocities for the BKS radar beam range gates that are assumed to contain the cardinal peak emission location of the FPI.Panels, a, b, c, and d are for the beams that slice through the north, south, east and west locations respectively at the range gate that contains the assumed peak emission location.Note, that the panels do not indicate that motion is north/south/east/westwards, instead positive values indicate motion towards the radar and negative away along the azimuth of the radar beam.The recorded line of sight ion velocities are indicated in blue.Shaded regions indicate the errors calculated using the method described in section 2.Compared in orange are the line of sight velocities from TIEGCM taken at an altitude of 250km at the same geographic latitude and longitude as the beam range gates and projected into the same direction as the radar beams.The line of sight velocities show high activity across all beams between 0400 UT and 0700 UT.The northern and eastward observations show an early spike at 0400 UT with line of sight velocities of approximately 160ms −1 .The southern observations show several spikes of high velocities from 0400 to 0630 UT peaking at −180ms −1 slightly after 0530 UT.The westward observations show high velocity spikes occurring between 0500 and 0600 UT, peaking at slightly less than −150ms −1 .An interesting observation is that the IMF Bz was directed northwards until after 0400 UT, however the north and southward BKS radar line of sight measurements show strong flows from as early as 0330 UT, and the eastwards observation starts to spike just before 0400 UT, indicating some driver other than the IMF Bz contributed to the fast ion motion.Furthermore, we can see that the westwards spikes begin (∼ 0510 UT) shortly after the strong eastwards observations end (∼ 0445 UT), which could indicate that it is the same patch of scatter that traverses across the FOV of the radar.The TIEGCM line of sight resolved ion velocities follow the same general trend over each observation point.Differences between each region can be identified most notably at 0500 UT, where the east location is around −10ms −1 while the southern point has model velocities of ∼ −25ms −1 .There are also slight variations in the magnitudes of the velocities due to the difference in the beam azimuth relative to TIEGCM's modelled three-dimensional ion flows.The TIEGCM line of sight ion velocities hover around their peak value of between 40−60ms −1 in all cells from roughly 0100 to 0300 UT, well before the first observed ion velocity spikes and southwards directed IMF Bz.They then decrease in velocity to close to 0ms −1 between 0500

Figure 5 Figure 4 .Figure 5 .
Figure 5 compares the FPI line of sight velocities (blue) with the neutral velocities modelled by TIEGCM at the assumed peak emission locations (orange).Panels a, b, c and d show the north, south, east and west observation directions respectively.Since TIEGCM's output velocities are given as geographic meridional and zonal magnitudes, we take TIEGCM's meridional flow for the north and south observations and the zonal flow for the east and west observations at each assumed peak emission location.We then project them into the same elevation angle as observed by the FPI.Positive velocities indicate motion to the north (meridionally) and east (zonally).The Zonal directions (east, west) show generally low velocities throughout the nighttime period, the meridional velocities however, show a gradual increase, particularly after the IMF Bz turns southward

Figure 6
Figure 6 shows an example of this fitting technique for the BKS scan from 0551 to 0601 UT.The top panel shows the scan of line of sight ion velocities plotted onto a geographic grid with the ion velocities corresponding to the colorbar to the right.Positive velocities indicate motion towards the radar, negative velocities away.Non-F-region ionospheric scatter or groundscatter identified by the Ribeiro algorithm has been colored grey.The location of the FPI is plotted at approximately 42 • north, 84 • west by the orange triangle.The assumed location of the peak neutral wind measurements are shown by the orange dots.The orange boxes mark what has been determined to be a patch of fast moving ionospheric scatter.Since we take the maximum velocity of each beam within a patch, it is only necessary to ensure that the highest velocity within a beam is included within the patch boundaries rather than needing to determine the exact spatial structure of the patch across all radar range gates.At the top right corner of each patch outline, a letter identifier (A, B, C, D and E) has been used to track each patch.At the time of the plot only patches C and D are present.Patches A and B occurred before 0551 UT, while E after 0601 UT and so are not shown here.The two panels below show the highest line of sight ion velocities in each beam for both the patches outlined (C) and (D), plotted against their beam azimuths and the resulting cosine fits for each of those cells.A beam azimuth of 0 • would point directly to magnetic north, negative azimuths indicate a westwards direction and positive eastwards.If we investigate the points used for fitting, patch C shows velocities that trend to positive at +90 • azimuth, while patch D shows velocities that trend to a negative at +90 • azimuth.Furthermore Patch C's eastmost beam and patch D's westmost beam show a difference of ∼ 150ms −1 within which persist through a longitude range of approximately 30 • from 80 • W to 110 • W, could indicate that the flows captured in the FPI region are part of a SAPS.SAPS do not typically occur during low geomagnetic activities, however Kunduri et al. (2017) found SAPS to occur 15% of the time in the nightside during relatively quiet conditions with velocities ∼ 100ms −1 ( Anderson et al. (2013)ghlatitude response times during geomagnetically quiet periods to be 3.3 hours.While we cannot calculate the full neutral coupling timescale because the neutrals never reach a steady state with the ions, the timescales in our observations can be viewed as the minimum value for the coupling timescale.Our values are close to the full values fromJoshi et al. (2015)andBillett et al. (2022), but are still smaller than those from the high-latitude timescales fromKosch et al. (2001), indicating that our values are reasonable.Studies byAruliah et al. (2005)and C.Anderson et al. (2013)investigated the impact that neutral winds have on Joule heating rate estimations.They calculated the highlatitude neutral wind dynamo to account for 29%(Aruliah et al., 2005)and 36% (C.Anderson et al., 2013)of the total Joule heating rates.Across patches B, C and D, the average neutral contribution (Q w ) to the total heating rate was 24.7%, 40.4% and 43.1% respectively, which is consistent with the previous studies, albeit at different latitudes.|75%|increaseorreduction in the total Joule heating rate depending on the neutral flow direction relative to the ions.When considering the multiplicative reduction, and the percentage decreases, our results show that the neutral winds have a significant reducing action on the overall Joule heating rate in line with the results obtained by the high-latitude studies ofBillett et al. (2018);Kiene et al. (2019).Although these studies did not show cases of the neutrals increasing the heating,Aruliah et al. (2005)and C.Anderson et al. (2013)did find that high-latitude neutrals were able to enhance or reduce the total Joule heating rates as similarly shown in this study.The increased heating rate magnitude of ∼ 75% in this study is symmetrical to the heating magnitude when the neutrals were decreasing the heating rate, im- Kiene et al. (2019)8)al contribution can be accounted for by the significantly stronger ion enhancements than in the other two patches, while their contributions although higher, still signify the majority of mid-latitude Joule heating response being due to the ions.Billett et al. (2018)indicated that the high-latitude Joule heating rate was nearly entirely eliminated when the neutral wind was pulled into the orientation of the ion flow.Kiene et al. (2019)used a scanning doppler imager with a SuperDARN radar to estimate high-latitude local Joule heating rates.They found that inclusion of the neutral winds in their Joule heating rate calculations dropped the total heating rate by a factor of ≃ 3 at high-latitudes.At the minimum value of Q w1 , which occurred in patch D, the Joule heating rate was decreased from 24.1pWm −3 to 5.61pWm −3 , representing a 4.2 times decrease, similar to the observations found inKiene et al. (2019).However, our observations vary substantially with the winds either contributing positively or negatively to the total heating rate, amounting to either a > Kiene et al. (2019)13)f our values to other studies are somewhat limited, however C.Anderson et al. (2013)andKiene et al. (2019)calculated high-resolution high-latitude local Joule-heating rates using instruments observing the ions and neutrals at 250km, which provides an excellent comparison to our midlatitude study.The Joule heating rate in this study peaks at ∼ 235pWm −3 .Both C.Anderson et al. (2013)andKiene et al. (2019)estimated the local high-latitude Joule heat-