Satellite Observations of the Influence of Energetic Electron Precipitation on the Mesosphere and Stratosphere in the Northern Hemisphere

The Earth’s atmosphere is influenced by energetic electrons coming from the magnetosphere. This energetic electron precipitation (EEP) is energized by the solar wind and directly affects in the high‐latitude mesosphere and lower thermosphere (MLT). EEP forms odd nitrogen (NOx) and hydrogen oxides (HOx) which destroy ozone. During winter EEP‐NOx descends to the stratosphere, establishing the indirect EEP effect. Several studies have found that EEP is related to changes in temperature and winds in the northern winter stratosphere. One of the most prominent effects of EEP is the influence on the northern polar vortex, a westerly wind system surrounding the winter pole in the middle atmosphere. Most studies of the EEP effect on dynamical features of the middle atmosphere have relied on either model simulations or reanalysis datasets which are mainly limited to stratospheric heights. We study here EEP effects on chemical and dynamical properties of the stratosphere and mesosphere in the northern hemisphere by using EOS Aura satellite’s measurements of atmospheric properties and POES satellites' measurements of precipitating electrons. We confirm earlier results showing that EEP decreases ozone and affects the temperature in the polar middle atmosphere and strengthens the stratospheric polar vortex. We show that EEP weakens the mesospheric polar vortex in late winter. This effect on polar vortex is partly due to changes in propagation and convergence of planetary waves. Accordingly, the EEP effect on the northern polar vortex depends on planetary waves not only in the stratosphere, as found in earlier studies, but also in the mesosphere.


Introduction
The Earth's atmosphere is subject to continuous but varying precipitation of energetic charged particles from space.The most common precipitating particles are electrons that are accelerated in the Earth's magnetosphere and precipitate to the atmosphere due to the interaction of the solar wind, a plasma flow from the Sun, with the magnetosphere.Energetic electron precipitation (EEP) is concentrated to the high-latitude mesosphere and lower thermosphere (MLT) where it ionizes atmospheric atoms and molecules.As a result, EEP forms odd nitrogen (NO x ) and odd hydrogen (HO x ) oxides (Solomon et al., 1981;Swider & Keneshea, 1973;Verronen et al., 2011), which act as catalysts in the destruction of ozone.Ozone is a radiatively active molecule species since ozone molecules absorb solar ultraviolet (UV) radiation and terrestrial infrared (IR) radiation and emit IR radiation (e.g., Kuhn & London, 1969).Thus, EEP-NO x and EEP-HO x (NO x and HO x produced by EEP) can affect atmospheric temperatures and dynamics via ozone destruction.While HO x compounds are short-living, lasting only hours or days, NO x compounds can survive even months if they are not photolyzed (e.g., Brasseur & Solomon, 2006).Therefore, EEP-NO y compounds (including NO x and their reservoir species, e.g., HNO 3 , N 2 O 5 ) can be sustained in winter during the descent to the polar upper stratosphere (Funke et al., 2005;Randall et al., 2007).
During the winter, the middle atmosphere (i.e., stratosphere and mesosphere) is characterized by a strong westerly wind surrounding the cold and dark polar region.This wind system, the polar vortex, is fundamentally determined by the meridional temperature difference balanced by thermal winds (e.g., Andrews et al., 1987), but its form and strength vary during the winter.One of the main drivers of polar vortex variability are planetary waves.These waves typically form in the troposphere and propagate to the middle atmosphere during the winter when the background flow is westerly and weaker than the so-called Rossby critical velocity which depends, for example, on the horizontal wave number of the waves (Charney & Drazin, 1961).As planetary waves travel, they dissipate and deposit easterly momentum, which the decelerates background westerly wind (Rossby, 1939).Additionally, planetary waves move air in meridional direction and drive the residual, also called Brewer-Dobson circulation, which transports air poleward in the middle atmosphere during the winter.Residual circulation is completed by the ascent of air in the equator and low latitudes and descent in the polar region.
Several modeling (e.g., Arsenovic et al., 2016;Baumgaertner et al., 2011;Rozanov et al., 2005) and observational studies (e.g., Andersson et al., 2014;Salminen et al., 2019;Seppälä et al., 2013;Zawedde et al., 2019) have shown that EEP decreases ozone and affects the temperature in the winter polar stratosphere and mesosphere.EEP-HO x and EEP-NO x destroy ozone in the polar mesosphere and upper stratosphere, which causes warming in the midwinter due to decreased radiative cooling, while in the late winter (i.e., after polar darkness) ozone destruction decreases radiative warming and, thus, temperature (Sinnhuber et al., 2018).In the northern hemisphere, the EEP effect is not limited to the polar mesosphere and upper stratosphere as EEP-caused effects decrease ozone and temperature in the polar lower stratosphere and strengthen the stratospheric polar vortex (Arsenovic et al., 2016;Baumgaertner et al., 2011;Salminen et al., 2019;Seppälä et al., 2013).Earlier studies have indicated that EEP affects the propagation of planetary waves in the northern winter hemisphere so that planetary waves divert more away from the polar vortex when EEP activity is high (Asikainen et al., 2020;Lu, Jarvis, & Hibbins, 2008;Seppälä et al., 2013).Decreased planetary wave convergence then leaves the polar vortex stronger and the residual circulation weaker, resulting in decreased adiabatic warming and ozone concentration in the polar lower stratosphere.A stronger westerly wind of the polar vortex diverts more planetary waves, which further strengthens the vortex and leads to a positive feedback loop of the wave-mean flow interaction (e.g., Andrews et al., 1987).In addition to the middle atmosphere, the EEP effect extends even to the surface level (Baumgaertner et al., 2011;Lu, Jarvis, & Hibbins, 2008;Maliniemi et al., 2013;Palamara & Bryant, 2004) where EEP activity correlates with the Northern Annular Mode (NAM) and North Atlantic Oscillation (NAO), climate modes explaining a major part of climate variability in the northern winter hemisphere (Hurrell et al., 2003).Variations in the polar vortex propagate to lower altitudes and are seen in the NAM and NAO with a lag of a week or two (Baldwin & Dunkerton, 2001), which explains the connection between EEP and these modes.
Most observational studies of the EEP effect on atmospheric chemical properties, for example, NO x (Funke et al., 2005;Seppälä et al., 2007) and ozone (Andersson et al., 2014;Damiani et al., 2016) are based on satellite observations from the beginning of 2000's to present.Most satellite observations before the 2000's do not cover dark regions (i.e., mid-winter high latitudes) since the measurement techniques of earlier satellite instruments relied on sunlight.On the other hand, most observational studies about the EEP effect on the dynamical properties, for example, on winds and planetary waves are based on reanalysis datasets (Asikainen et al., 2020;Lu, Clilverd, Seppälä, & Hood, 2008;Salminen et al., 2019;Salminen et al., 2020Salminen et al., , 2022;;Seppälä et al., 2009).While recent reanalysis datasets typically cover long time periods, extending back to the 1980's or even the 1950's, most of them cover only altitudes up to the upper stratosphere.Therefore, the EEP effects, for example, on winds and planetary waves in the mesosphere have remained elusive in the observational studies.Furthermore, earlier studies based on reanalysis data have indicated that the EEP effect on the northern stratospheric polar vortex depends on the equatorial stratospheric wind system called the Quasi-Biennial Oscillation (QBO) (Maliniemi et al., 2013(Maliniemi et al., , 2016;;Palamara & Bryant, 2004;Salminen et al., 2019Salminen et al., , 2020;;Seppälä et al., 2013) and on extreme events of polar vortex variability called the sudden stratospheric warmings (SSW) (Asikainen et al., 2020).QBO affects the propagation of planetary waves while SSWs are associated with enhanced planetary wave activity in the stratosphere.Earlier studies propose that the modulation of the EEP effect by QBO and SSWs is fundamentally driven by planetary waves (Asikainen et al., 2020;Salminen et al., 2019).Related to this, Salminen et al. (2022) showed, based on reanalysis data, that the EEP effect depends on the distribution of planetary waves so that the effect is stronger in those winter months in which planetary waves are focused on the equatorward side of the polar vortex.However, these findings on the role of planetary waves have, so far, not been confirmed with direct observations.Lu and Jarvis (2011) suggested that solar wind dynamic pressure may be linked to changes in the QBO so that the QBO is more easterly when both solar activity and solar wind dynamic pressure are high.While mechanisms for this effect are still unresolved, it underlines the global scale and complexity of solar windrelated effects on the atmosphere.EOS (Earth Observing System) Aura satellite measurements constitute one of the best-suited observational datasets to study the EEP related variability in the middle atmosphere.The Aura satellite carries the MLS (Microwave Limb Sounder) instrument which measures thermal microwave radiation of atmospheric gas at different wavelengths (Waters et al., 2006).These measurements can be used to derive air temperature and the concentration of several chemical species in the atmosphere.Aura MLS measurements cover pressure levels from the upper troposphere to the lower thermosphere at latitudes 82°S-82°N since August 2004.Compared to preceding satellite missions, Aura MLS measurements also cover dark regions and have only a few datagaps.Thus, they provide one of the longest and most complete global data series of middle atmosphere composition, which has been used in several recent studies of the effect of particle precipitation on atmospheric chemistry (e.g., Andersson et al., 2014;Damiani et al., 2016;Gordon et al., 2021;Jia et al., 2020;Lee et al., 2018;Verronen et al., 2011;Zawedde et al., 2019).However, the effect of particle precipitation on thermal and dynamical properties of the atmosphere has not been examined yet with Aura measurements.
In this study we examine interannual variability related to EEP in both dynamical and chemical properties of the stratosphere and mesosphere in the northern hemisphere by using Aura MLS measurements on atmospheric properties and POES (Polar Operational Environmental Satellites) measurements on electron precipitation.We also study the dependence of the EEP effect on planetary waves using these direct observations.In Section 2 we describe the datasets and methods used in this study.In Section 3 we study the EEP effect on the chemical properties of the atmosphere.In Section 4 we discuss the dynamical effects of EEP and in Section 5 the planetary wave modulation of the EEP effect.In Section 6 we discuss our results and present conclusions based on them.

Atmospheric Data
We use measurements of the MLS instrument of the EOS Aura satellite (Waters et al., 2006) to study variability in different chemical and dynamical properties of the atmosphere.Aura was launched in July 2004 and Aura MLS measurements are available from August 2004 onwards.Aura completes about 15 orbits per day on a Sunsynchronous and nearly polar orbit with a 98°inclination and 13:45 local time (LT) equator crossing, providing measurements with almost global coverage at latitudes between 82°S and 82°N.We use Aura MLS measurements of ozone (O 3 ) (Schwartz, Froidevaux, et al., 2020), nitric acid (HNO 3 ) (Manney et al., 2020), temperature (Schwartz et al., 2020b) and geopotential height (Schwartz et al., 2020a).Aura does not measure any NO x compounds, but HNO 3 is one of the NO y species and, thus, can be used to quantify descending EEP-NO y .EEP produces HNO 3 in ion-ion recombination reactions involving NO 3 and N 2 O 5 which are most efficient at altitudes below 80 km (Sinnhuber et al., 2012).We use the level 2 data (version 5) in which variables are given as pressure level profiles for each measurement made every 25 s.We removed unreliable and erroneous measurements from the data following the approach by Livesey et al. (2022).
Accuracy and reliability of Aura MLS data depend on the altitude and the used variable, since Aura MLS measurements are based on microwave radiation at a wavelength specific for each variable.Therefore, each variable has its own appropriate altitude range.Temperature and geopotential height are given between pressure levels 261-0.00046hPa, O 3 between 261 and 0.001 hPa and HNO 3 between 215 and 1 hPa.Estimated random uncertainty, precision, of individual measurements for O 3 is between ±0.03 and ±0.2 parts per million volume (ppmv) in the stratosphere and ±0.3 and ±3.4 ppmv in the mesosphere, for HNO 3 between ±0.6 and ±1.2 parts per billion volume (ppbv) in the stratosphere, for temperature between ±0.5 and ±1.2 K in the stratosphere and ±1.3 and ±4.0 K in the mesosphere, and for geopotential height between ±8 and ±20 m in the stratosphere and ±45 and ±190 m in the mesosphere.Estimated systematic uncertainty, accuracy, for O 3 varies between ±0.1 and ±0.4 ppmv in the stratosphere and ±0.1 and ± 0.9 ppmv in the mesosphere, for HNO 3 between ±0.1 and ±1.1 ppbv in the stratosphere, for temperature between 2 and 3 K in the stratosphere and between 9 and 3 K in the mesosphere, and for geopotential height between ±20 and ±120 m in the stratosphere and ±150 and ±700 m in the mesosphere.Vertical resolution of measurements for O 3 and HNO 3 is between 2 and 4 km in the stratosphere and 3-7 km for O 3 in the mesosphere.Vertical resolution of temperature is between 3.5 and 7 km in the stratosphere and 7-13 km in the mesosphere.Since geopotential height is based on vertically integrated temperature, its vertical resolution is not well-defined.
We compute daily and zonally averaged profiles of measurements in each 2.5°wide latitude band.The possible effect of uncertainties on the analysis becomes negligible after daily or monthly averaging.Figure 1, bottom panel shows the monthly ozone profiles averaged zonally over latitudes 70°N 82°N.We also compute daily geopotential height profiles in the latitude-longitude grid to calculate Eliassen-Palm flux (EP flux) components which can be used to study planetary wave propagation and convergence (Eliassen & Palm, 1961).Daily gridded values are calculated by interpolating measurements of a day to 2.5°× 2.5°latitude-longitude grid points with the Delaunay triangulation method.We interpolate gridded values separately using measurements of ascending and descending part of the satellite orbit and then average the two gridded sets.
Zonally averaged geopotential height is used to calculate the approximate zonal wind with the equation where u is the zonal wind, a = 6,371 km is the radius of the Earth, ϕ is the latitude, f = 2 Ω sin(ϕ) is the Coriolis parameter, Ω is the angular velocity of the Earth 2π/ 24 × 60 2 s), and Φ is the geopotential, that is, geopotential height multiplied with the gravitational acceleration 9.81 ms 2 .Overbar indicates a zonally averaged value.Equation 1 is an approximation of the zonal mean version of the meridional momentum equation for a fluid.The solved zonally averaged zonal wind, the so-called gradient wind, is accurate in monthly scales (Randel, 1987) et al., 2011;Smith et al., 2017).Even though wave motions are neglected in Equation 1, they affect the distribution of geopotential height and, thus, contribute to the solved zonal wind.EP flux is derived using latitudelongitude gridded geopotential height data by, first, calculating the longitudinal Fourier harmonics of geopotential in form of where A n and α n are the amplitude and phase of nth harmonic, λ is the longitude, z = H log (p/p 0 ) is the logpressure approximation of altitude and H = 6,700 m is the scale height.Prime (′) corresponds to difference relative to the zonal average value.Now A 0 equals Φ while Φ′ is the sum of the harmonic components.These harmonic components of geopotential can be used to calculate the corresponding harmonic components of geostrophic zonal wind (u′) and meridional wind (v′) as well as potential temperature (θ′) (Andrews et al., 1987).
Then quasi-geostrophic EP flux F in spherical coordinates can be calculated as where ρ 0 = ρ s (p/p s ) is the air density, ρ s = 1.204 kg/m 3 is the air density at the surface level, p is the air pressure, p s = 1,000 hPa is the air pressure at the surface level, and N 2 = gθ 1 (∂θ/∂z) is square of the Brunt-Väisälä frequency.We include the contribution to EP flux by planetary waves with zonal wavenumber from 1 to 3, that is, S = 3 in Equation 3. Derivation of F using gridded geopotential data is described in more detail by Andrews et al. (1987).The EP flux derived from satellite measurements of geopotential height was also used in the study by Leroy and Anderson (2007).While successive Aura measurements at the same latitude are longitudinally separated by about 24°in the ascending and descending part of the orbit, the gridded geopotential height data are accurate enough when deriving large scale wave structures, that is, waves with wavenumber 1-3.

Electron Precipitation Data
We use measurements of the POES satellites to quantify EEP.These satellites have measured electrons on a Sun-synchronous and nearly polar orbit since 1978.POES satellites carry the SEM (Space Environment Monitor) instrument package which includes MEPED (Medium Energy Proton and Electron Detector) instrument measuring electrons in three integrated energy channels (>30, >100 and >300 keV) with two orthogonally directed telescopes.MEPED electron flux measurements suffer from proton contamination and inhomogeneity between the two versions of SEM, SEM-1 in 1978-1998 and SEM-2 from 1998 onwards.In this study, we use MEPED data corrected and calibrated by Asikainen and Mursula (2013), Asikainen and Ruopsa (2019) and Asikainen (2019).We use the flux of electrons with energies over 30 keV averaged over the two telescopes and over corrected geomagnetic latitudes 40°-90°in the northern hemisphere.An average flux between the two telescopes is calculated using logarithmic values as was done for the UOulu POES dataset in the study by Nesse Tyssøy et al. (2022).Figure 1, top panel shows the time series of monthly averaged flux of precipitating electrons with energies over 30 keV.Precipitated electrons with energies over 30 keV penetrate to the altitudes lower than 80 km (e.g., Nesse Tyssøy et al., 2022) where they produce, for example, HNO 3 (Orsolini et al., 2018).

Regression Analysis
We use linear regression analysis to estimate the response of atmospheric variables to EEP.We use 2-month averages of all variables in winters 2004/2005-2021/2022 (18 winters) and calculate regressions separately for November-December, December-January, January-February and February-March.The regression model is where Y(t) is the response variable, that is, one of the atmospheric variables, EEP(t) is the explaining variable EEP, α is the constant term, β is the regression coefficient for EEP(t) and e(t) is the residual.After solving the regression coefficients with the least squares method, the response in Y is estimated as a ΔY = β × ΔEEP where ΔEEP is the standard deviation of EEP in the whole studied time period.Thus, ΔY is a change in Y which is related to a change of one standard deviation in EEP.We estimate significances of the responses by calculating the p-values for them with Student's t-test.To further test the reliability of the regression responses, we also used the Cochrane-Orcutt method (Cochrane & Orcutt, 1949) to calculate the regression responses and their significances since this method takes into account possible autocorrelation of residuals, which makes estimations of significances more accurate compared to the normal regression model.The responses and significances are almost the same with both methods.The responses also remain almost the same if we include the solar F10.7 radio flux index, indicating solar activity, and the QBO wind as explaining variables in addition to EEP.Since we use regression analysis to study interannual variations separately in 2-month periods, inaccuracies or possible biases of Aura measurements do not significantly affect our results.To test the modulation of the EEP effect on the atmosphere by planetary waves, we calculate regression responses in the different phases of PC1 and, separately, of PC2.We first classify months according to positive/negative PC1 (or PC2) phase so that those months, in which the examined PC is lower (higher) than the median value of that calendar month, are in the negative (positive) phase of that PC.We calculate separate regressions for PC1 and PC2 in which we include an indicator variable corresponding to the phase of the PC.The regression model in this case is

Principal Component
where I(t) is the indicator variable which is one in the positive PC phase and 0 in the negative PC phase, and a represents the difference in the constant term and b the difference in the regression coefficient for EEP(t) between the positive and negative PC phase.Now the response ΔY is ΔEEP × β in the negative PC phase and ΔEEP × (β + b) in the positive PC phase.

EEP Effects on Chemical Species
Figure 3 shows the EEP effect on zonally averaged ozone (top row, panels (a)-(d)) and nitric acid (bottom row, panels (e)-(h)) in November-December (Nov-Dec; column 1), December-January (Dec-Jan; column 2), January-February (Jan-Feb; column 3), and February-March (Feb-Mar; column 4).The responses were calculated using regression analysis as described in Section 2.3.In February-March period, we have used EEP values of January-February, that is, with 1-month lag, as we wanted to take into account the descending time of EEP-NO x (Funke et al., 2016).In the polar mesosphere, ozone is significantly decreased by over 0.1 ppmv at pressure levels 0.01-0.1 hPa in Nov-Dec, Dec-Jan and Jan-Feb.In Feb-Mar the negative ozone response to EEP in these levels is weaker (less than 0.1 ppmv) and insignificant.In the polar stratosphere, ozone is decreased 0.1-0.3ppmv in Dec-Jan, Jan-Feb and Feb-Mar.Especially in the polar upper stratosphere, decrease in ozone is clearly significant.HNO 3 is significantly increased by 0.1-0.3ppbv in the polar upper stratosphere as a response to EEP in all months.In the lower stratosphere, HNO 3 is significantly decreased by 0.1-0.2ppbv at latitudes 40° 60°N in Dec-Jan, Jan-Feb and Feb-Mar.
Ozone decrease by EEP in the northern polar mesosphere at pressures 0.1-0.01hPa has been detected in earlier studies (e.g., Andersson et al., 2014;Andersson et al., 2018;Arsenovic et al., 2016) and is most likely due to the direct EEP effect via HO x (Andersson et al., 2012;Verronen et al., 2011).In the polar upper stratosphere at 1-10 hPa, ozone is most likely decreased due to the descending NO x compounds, since few precipitating electrons can reach the stratosphere.This indirect EEP effect is indicated in Figure 3 as increased HNO 3 (which is a reservoir species for NO x ) in the polar upper stratosphere.While HNO 3 molecules do not themselves destroy ozone, HNO 3 can transform back to reactive NO x compounds in photolysis or a reaction with OH molecules (e.g., Brasseur & Solomon, 2006).Moreover, increase in HNO 3 most probably indicates increases in the amounts of other NO y compounds.Damiani et al. (2016) used Aura MLS measurements and found similar responses of HNO 3 and O 3 to geomagnetic activity, a proxy of EEP, in the polar mesosphere and upper stratosphere in the southern hemisphere.EEP related ozone decrease in the polar lower stratosphere and HNO 3 decrease in the midlatitude lower stratosphere, where EEP-NO y does not appear, are most probably related to changes in the residual circulation, which is considered in Section 4.

EEP Effects on Atmospheric Dynamics
Figure 4 shows the EEP effects on zonally averaged ozone (top row, panels (a)-(d); same as in Figure 3 but responses are given as percentages of climatology), temperature (middle row, panels (e)-(h)) and zonal wind (bottom row, panels (i)-(l)) in 2-month periods between November and March (columns 1-4).Relative to the climatology, EEP decreases ozone approximately by 10% in the polar mesosphere and by 1%-10% in the polar stratosphere depending on the period.EEP significantly increases temperature in the polar lower mesosphere and upper stratosphere by 1-3 K in Dec-Jan and by 3-6 K in Jan-Feb and Feb-Mar.In the polar lower stratosphere EEP is associated with a temperature decrease of 2-4 K in Dec-Jan and Jan-Feb.EEP also cools the polar middle and upper mesosphere (at pressures lower than 0.1 hPa) by 1-4 K in Jan-Feb and Feb-Mar.In addition to the effects in the polar region, EEP causes significant warming (1-2 K) in the mid-and low-latitude upper mesosphere in Dec-Jan, Jan-Feb and Feb-Mar.There are also limited but significant positive and negative temperature responses (1-2 K) at lower latitudes in the lower and upper stratosphere, respectively, in Jan-Feb and Feb-Mar.The significant temperature responses at lower latitudes are opposite to the responses at high-latitudes at the same altitude.
EEP significantly strengthens the westerly zonal wind by 3-8 m/s at latitudes higher than 60°in the stratosphere in Dec-Jan and Jan-Feb (Figures 4j and 4k).This response corresponds to a strengthening and minor northward shift of the polar vortex.In the upper mesosphere EEP causes a weak but significant increase of zonal wind (less than 3 m/s) at 60°N latitude in Nov-Dec (Figure 4i) and Dec-Jan.In Jan-Feb and Feb-Mar (Figure 4l), there is a significant negative zonal wind response (5-10 m/s) to EEP in the mesosphere poleward of 40°N.
During the winter a decrease in ozone in the polar darkness leads to a decrease in radiative cooling (e.g., Brasseur & Solomon, 2006;Sinnhuber et al., 2018).This explains, at least partly, the positive temperature response to EEP in the upper stratosphere and lower mesosphere from December to March.In the polar lower stratosphere the negative ozone response to EEP is associated with decreased temperature.These negative responses are most likely due to weakened residual circulation and downwelling inside the vortex which transport less ozone to the polar lower stratosphere and adiabatically warm it less effectively.Decreased horizontal mixing of air between mid-and high-latitudes may also contribute to a decrease in both temperature and ozone.Similar temperature and zonal wind responses to EEP in the stratosphere were found in earlier studies based on reanalysis data (e.g., Salminen et al., 2019;Seppälä et al., 2013) and modeling (e.g., Arsenovic et al., 2016;Baumgaertner et al., 2011).The negative temperature response at high-latitudes at 0.01-0.001hPa in Jan-Feb and Feb-Mar is also probably due to decreased downwelling of the residual circulation, since ozone is not decreased by EEP at these levels.The negative zonal wind response in the mesosphere is in accordance with the thermal wind shear balance since the poleward temperature gradient decreases (increases) in the lower (upper) high-latitude mesosphere.
Figure 5 shows EEP effect on the EP flux and its divergence in Nov-Dec (panel (a)), Dec-Jan (panel (b)), Jan-Feb (panel (c)) and Feb-Mar (panel (d)).EEP decreases upward EP flux and its convergence in Dec-Jan, and these responses are significant poleward of 50°N in the middle and upper mesosphere and in the lower stratosphere.However, in Jan-Feb, upward and equatorward EP flux is increased and converging more in the lower mesosphere and upper stratosphere at 30°-60°N latitudes.In Feb-Mar, EEP increases upward and equatorward EP flux and its convergence in an even wider region in the mesosphere and upper stratosphere than in Jan-Feb.Upward EP flux is increased by EEP also in the lower stratosphere at 60°-75°N in Feb-Mar.
The EEP effect on the EP flux is in agreement with the effects on temperature, zonal wind and ozone.In Dec-Jan, EP flux convergence is decreased in the high-latitude stratosphere and mesosphere (Figure 5b), which accelerates the westerly wind of polar vortex (Figure 4j) and weakens the residual circulation, leading to decreased ozone and temperature in the polar lower stratosphere (Figures 4b and 4f, respectively).Asikainen et al. (2020) and Salminen et al. (2022) found also similar EP flux divergence responses to EEP in the polar lower stratosphere.
In Jan-Feb (Figure 5c), EP flux converges more at mid-and low-latitude mesosphere, while in the lower stratosphere it still diverges more at the latitudes poleward of 60°N.Increased EP flux convergence in the mesosphere contributes to the decreased westerly wind in the mesosphere (Figure 4k).Increased EP flux convergence is related to increased temperature in the high-latitude lower mesosphere and upper stratosphere (Figure 4g) which is probably due to stronger residual circulation and downwelling (e.g., Andrews et al., 1987).In Feb-Mar (Figure 5d), EP flux converges even more in response to EEP at all latitudes in the mesosphere which is seen as a larger and stronger negative EEP effect on zonal wind in that region.Lu et al. (2013) found also that EP flux concentrates to higher altitudes in the stratosphere as a response to solar wind dynamic pressure (which correlates with EEP) and they proposed that this is due to modified wave guide related to the enhanced polar vortex.
In addition to EP flux, which corresponds to planetary wave activity, gravity waves modulate the residual circulation, temperature and winds in the mesosphere and even higher (Fritts & Alexander, 2003;Holton, 1983), and may also contribute to the EEP effect.As in the case of planetary waves, propagation of gravity waves is influenced by background wind and temperature distribution (Lindzen, 1981).When the stratospheric polar vortex is stronger than on average, for example, due to EEP, less eastward gravity waves propagate to the mesosphere, leading to a weaker mesospheric polar vortex and increased residual circulation and adiabatic warming in the polar lower mesosphere (Holton, 1983;Karlsson et al., 2009).Earlier studies (Limpasuvan et al., 2016;Liu & Roble, 2002;Pedatella & Harvey, 2022) have found that temperature in the polar upper mesosphere and lower thermosphere depends on the strength of stratospheric polar vortex similarly as it depends on EEP shown here.The roles of planetary and gravity waves in this temperature response are still unresolved.

EEP Effect Modulated by Planetary Waves
Figure 6 shows EEP effects on zonally averaged zonal wind in the positive and negative F z PC1 phase (first and second row, respectively) and in the positive and negative F z PC2 phase (third and fourth row, respectively) in Nov-Dec, Dec-Jan, Jan-Feb and Feb-Mar (columns 1-4, respectively).The responses were calculated as described in Section 2.4 and there are nine cases in each 2-month period of each phase.In the positive PC1 phase (panels (a)-(d)), EEP significantly increases zonal wind by 1-5 m/s at high latitudes in the lower stratosphere and in the middle and upper mesosphere at mid-latitudes in Dec-Jan.In Jan-Feb, the response in the lower stratosphere is marginally significant.In the negative PC1 phase (panels (e)-(h)), EEP causes a negative zonal wind response in the mesosphere in Jan-Feb, in which the response is only marginally significant, and in Feb-Mar, in which the response is clearly significant.While magnitudes of the responses in Jan-Feb and Feb-Mar in the negative PC1 phase are rather large in many regions, for example, in the high-latitude stratosphere (10-15 m/s), these responses are mainly statistically insignificant.
In the positive PC2 phase (Figure 6, panels (i)-(l)) EEP causes a strong and significant positive effect on zonal wind (5-15 m/s) in the high-latitude stratosphere and lower mesosphere in Dec-Jan and in the high-latitude lower and middle stratosphere in Jan-Feb.In positive PC2 Jan-Feb and Feb-Mar, there is a significant negative zonal wind response to EEP in the mid-and high-latitude mesosphere.In the negative PC2 phase (Figure 6, panels (m)-(p)) EEP significantly strengthens the zonal wind in the mesosphere in Nov-Dec and weakens it in Feb-Mar by 1-5 m/s.In the stratosphere, EEP effects on zonal wind are weaker (less than 5 m/s) than in other PC phases and significant only in limited regions.
Results shown in Figure 6 agree with the study by Salminen et al. (2022) who examined the EEP effect on stratospheric polar vortex and its modulation by planetary waves with re-analysis data.EEP both strengthens the stratospheric polar vortex and weakens the mesospheric polar vortex especially when there are more planetary waves propagating at the equatorward flank of the vortex (at latitudes lower 60°N) and less inside the vortex (at latitudes higher than 60°N).These responses in the positive PC2 phase are significant and more consistent than the responses in the two PC1 phases or in the negative PC2 phase.As seen in Figure 5b, EEP significantly increases EP flux divergence at 50°-60°N, that is, diverts planetary waves away from the outer flank of the vortex in the lower stratosphere in Dec-Jan.In Jan-Feb (Figure 5c), EEP increases EP flux or planetary wave convergence in the mesosphere at latitudes 30°-50°N.If planetary waves propagate more at latitudes southward of 60°N and less at latitudes poleward of 60°N, as in the positive PC2 phase, they are more vulnerable to EEP related changes in wave propagation and divergence than in the opposite case, that is, the negative PC2 phase.

Discussion and Conclusions
We studied here the effects of EEP on the northern winter middle atmosphere (stratosphere and mesosphere) by using Aura MLS satellite measurements of chemical and dynamical properties of the atmosphere and POES MEPED measurements of precipitating electrons.We found that EEP increases NO y compounds, indicated by HNO 3 , in the upper stratosphere already in the beginning of winter (November-December).This is in accordance with the study by Seppälä et al. (2007), who used the observations of GOMOS satellite, and the study by Funke  2014) who used the observations of MIPAS satellite.We showed that EEP is related to a decrease in ozone in the polar middle mesosphere in all the winter and in the polar stratosphere and lower mesosphere from December onwards.These findings agree with several model based studies (e.g., Andersson et al., 2018;Arsenovic et al., 2016;Baumgaertner et al., 2011;Rozanov et al., 2005), as well as studies based on reanalysis data (Salminen et al., 2019).Andersson et al. (2014) studied ozone observations of three satellites (GOMOS, SABER, Aura) and found that EEP can cause significant decreases of ozone in the northern polar mesosphere at 70-80 km altitudes, which agrees with our results.
We found that EEP is related to a significant increase (decrease) of temperature in the polar lower mesosphere and upper stratosphere (lower stratosphere) in December-March and to a strengthening of the stratospheric polar vortex in December-February.Several studies based on reanalysis data (e.g., Asikainen et al., 2020;Salminen et al., 2019;Seppälä et al., 2013) and models (Arsenovic et al., 2016;Baumgaertner et al., 2011) have found similar EEP effects on temperature and polar vortex as here.Moreover, we found that the westerly zonal wind in the high-latitude mesosphere is weakened by EEP in January-March.We showed that changes in the high-latitude zonal winds and temperatures are related to changes in EP flux, that is, planetary wave propagation.When EEP is strong, EP flux converges less in the high-latitude stratosphere and mesosphere in early winter (December-January).In late winter (January-February, February-March) increased EEP is related to increased EP flux convergence in the mesosphere.In the stratosphere, the EEP effect on EP flux convergence found here is similar to that found in earlier studies based on reanalysis data (Asikainen et al., 2020;Lu et al., 2013;Salminen et al., 2022).
Our results based on satellite data also confirm the findings by Salminen et al. (2022) who used reanalysis data to show that EEP affects the stratospheric polar vortex most significantly if planetary wave propagation is focused at lower latitudes than on average.Moreover, we showed that in these planetary wave conditions the EEP effect on the mesospheric polar vortex is more significant than in other conditions.In December-January increased EEP is associated with an increase in EP flux divergence in the lower stratosphere at latitudes 50-60°N, which suggests that planetary waves propagating outside the polar vortex enable, at least partly, the EEP effect on the polar vortex.The distribution of planetary waves is fundamentally determined by the source of waves and by the waveguide but we cannot separate contribution by these two factors here and, thus, they remain to be examined in future studies.
This study augments the observational evidence of the overall EEP effect in the middle atmosphere.It underlines the importance of the interdependency between the EEP effect and planetary waves.Measurements of other satellites, such as SABER, GOMOS and MIPAS, should be used to confirm findings presented here and to further study the EEP effect in the whole middle atmosphere and its modulation by planetary waves.Especially the EEP effect on mesospheric dynamics should be examined in future studies, since Aura data exhibits larger uncertainties and biases there, and earlier studies based on reanalysis data have mainly analyzed the EEP effect in the stratosphere.While several studies based on modeling the atmosphere and climate have examined the atmospheric effects of EEP, they have not considered the role of planetary wave distribution in modulating the EEP effect.Thus, planetary wave modulation of the EEP effect should be studied with atmospheric models in future.
The correctly modeled EEP effect on atmosphere and climate would improve weather and climate forecasts at least in the regions of the northern winter hemisphere in which the polar vortex has a significant impact.

Figure 1 .
Figure 1.Top panel: monthly averaged flux of precipitating electrons with energies over 30 keV in the northern hemisphere.Bottom panel: monthly values of ozone volume mixing ratio (VMR) at pressure levels 0.0002-200 hPa averaged zonally over latitudes 70°-82°N.Gray color in the bottom panel correspond to times of missing data.
Analysis of F zSalminen et al. (2022) showed that the distribution of upward propagating planetary waves (i.e., vertical EP flux) determines the strength of the EEP effect on the stratospheric polar vortex.Here we determine planetary wave/EP flux distribution similarly asSalminen et al. (2022) by calculating the two leading principal components (PCs) and the corresponding empirical orthogonal functions (EOFs) of vertical EP flux (F z in Equation3) at 30 hPa pressure level at latitudes 30°-82.5°N.F z in these latitudes and pressure level represents the flux of planetary waves which can potentially affect in the upper stratosphere in the northern wintertime stratosphere.The two EOFs, EOF 1 and EOF 2, are shown separately for December, January, February and March in Figure2.Note that both EOF 1 and EOF 2 are almost the same in each winter month.EOF 1 (left panel) represents the latitudinal variability in the overall vertical EP flux and is equal to the climatological distribution of vertical EP flux at that pressure level.EOF 1 peaks close to 60°-62.5°Nlatitude and positive (negative) values of the corresponding PC, PC1, imply stronger (weaker) overall upward EP flux than average.EOF 2 (right panel) represents the leading variability in the latitudinal distribution of vertical EP flux so that the corresponding PC, PC2, is positive when the flux is concentrated more equatorward and less poleward than on average, and vice versa if PC2 is negative.Both EOFs of Figure2and the corresponding PCs are similar to those derived in the study bySalminen et al. (2022) who used ERA-Interim and ERA-40 reanalysis data.

Figure 2 .
Figure 2. The two leading empirical orthogonal functions, EOF 1 (left panel) and EOF 2 (right panel), of vertical EP flux at 30 hPa pressure level.EOFs are calculated with principal component analysis separately for December, January, February and March.

Figure 3 .
Figure3.EEP effects on zonally averaged ozone (top row, panels (a)-(d)) and nitric acid (bottom row, panels (e)-(h)) volume mixing ratio in November-December, December-January, January-February and February-March (columns 1-4, respectively).Yellow and red colors correspond to positive responses and blue to negative responses.Gray area represents regions out of data coverage.Gray, black and magenta contours surround the regions which are significant at 90%, 95% and 99% level, respectively.

Figure 5 .
Figure 5. EEP effects on EP flux (arrows) and EP flux divergence (color) in November-December (panel (a)), in December-January (panel (b)), in January-February (panel (c)) and in February-March (panel (d)).Arrows have constant length and represent only the direction of response.Only significant ( p < 0.05) EP flux responses are shown.Color and contour markings are similar as in Figure 3.

Figure 6 .
Figure6.Zonally averaged zonal wind responses to EEP in positive and negative phase of F z PC1 (first and second row, respectively) and in positive and negative phase of F z PC2 (third and fourth row, respectively) in November-December, December-January, January-February and February-March (columns 1-4).The markings are similar to those in Figure3.