Simulated seasonal impact on middle atmospheric ozone from high-energy electron precipitation related to pulsating aurorae

. Recent simulation studies have provided evidence that pulsating aurorae (PsA) associated with high-energy electron precipitation is having a clear local impact on ozone chemistry in the polar middle mesosphere. However, it is not clear if PsA are frequent enough to cause longer-term effects of measurable magnitude. There is also an open question of the relative contribution of PsA-related energetic electron precipitation (PsA-EEP) to the total atmospheric forcing by solar energetic particle precipitation (EPP). Here we investigate the PsA-EEP impact on stratospheric and mesospheric odd hydrogen, odd nitrogen, 5 and ozone concentrations. We make use of the Whole Atmosphere Community Climate Model and recent understanding on PsA frequency, latitudinal and magnetic local time extent, and energy-ﬂux spectra. Analysing an 18-month time period cover-ing all seasons, we particularly look at PsA-EEP impacts at two polar observation stations located at the opposite hemispheres: Tromsø in the NH and Halley in the SH. We ﬁnd that PsA-EEP can have a measurable impact on ozone concentration above 30 km altitude, with ozone depletion by up to 8% seen in winter periods due to PsA-EEP-driven NO x enhancement. We also 10 ﬁnd that direct mesospheric NO x production by high-energy electrons ( E > 100 keV) accounts for about half of the PsA-EEP-driven upper stratospheric ozone depletion. A larger PsA-EEP impact is seen the SH where the background dynamical variability is weaker than in the NH. Clearly indicated from our results, consideration of polar vortex dynamics is required to understand PsA-EEP impacts seen at ground observation stations, especially in the NH. We conclude that PsA-EEP has potential to make an important contribution to the total EPP forcing, thus it should be considered in atmospheric and climate 15 simulations.

In this paper, we study the chemical impacts of PsA-EEP at polar latitudes. We use the Whole Atmosphere Community Climate Model with its lower ionospheric chemistry extension (WACCM-D), together with a plausible estimate for PsA-EEP forcing based on observations reported in the literature, to simulate the PsA-EEP impact. We analyse the atmospheric response for over an 18-month period, including all seasons of the year. Considering both the Northern and Southern hemispheres, observation stations from our global simulations in order to understand the expected impact that could be measured by groundbased instrumentation such as radiometers (Daae et al., 2014;Newnham et al., 2018), and the next generation EISCAT_3D ionospheric radar system (McCrea et al., 2015). Finally, we discuss our results in the context of overall EPP impact and the current challenges in representing EPP forcing in simulations.

Model and simulations 100
Here we use the Whole Atmosphere Community Climate Model (WACCM) version 4, described in detail by Marsh et al. (2007) and Marsh et al. (2013). Running simulations in a specified dynamics mode, the model temperature, horizontal winds, and surface pressure below 50 km were nudged to NASA GEOS5.1 re-analysis data (Rienecker et al., 2008). At the levels above 50 km up to the model upper boundary at 6 × 10 −6 hPa the model dynamics are free-running, although there is a degree of control coming from specified dynamics below. We make use of the variant WACCM-D which includes a representation 105 of the lower ionospheric chemistry of both positive and negative cluster ions and was particularly designed for EPP studies . WACCM-D captures a full range of observed EPP impacts in the middle atmosphere (Andersson et al., 2016;Kalakoski et al., 2020), in contrast to the standard WACCM which includes only a parameterization of HO x and NO x production. Recent D-region studies using WACCM-D include work on seasonal changes in ion composition and comparison of the latitudinal extent of solar proton events against ionospheric observations (Orsolini et al., 2018;Heino et al., 2019). To create a typical PsA-EEP forcing for our simulations, we utilize energy-flux spectra and ionization rates published by Turunen et al. (2016). These are based on ionospheric observations of the EISCAT radar and the KAIRA riometric observations during a PsA event on the 17th of November, 2012. In their Figure 2, Turunen et al. (2016) presented several different ionization 115 rate profiles which differ especially at altitudes below 80 km due to larger electron flux differences and uncertainties present at high electron energies >100 keV. Selecting the PsA-EEP spectrum that is in good agreement with the statistical median spectrum of Tesema et al. (2020), we make use of the resulting "MCMC median" ionization rate profile. The electron spectrum for that was inverted by Turunen et al. (2016)  where previous statistical studies have observed a high occurrence rate of pulsating aurora (Partamies et al., 2017;Tesema et al., 2020). In summary, with these simplifying assumptions we aim at simulating and analysing the full potential of PsA atmospheric impacts. Note that the same PsA-EEP forcing characteristics are applied throughout the year, which allows for direct comparisons between seasonal atmospheric responses.
To demonstrate the impact of the PsA-EEP forcing in WACCM-D simulations, Figure 1 shows two snapshots of global 135 electron concentration on 18th of January at ≈78 km altitude. Overall features include higher values on the dayside ionosphere from photoionization as well as higher values in the auroral regions due to particle precipitation. Very high electron concentration are shown with red color and occur at the time and place of PsA-EEP forcing. During every other day, these high-ionization PsA-EEP patters remain at the same magnetic local times and rotate once around the magnetic poles, following the magnetic latitudes of the forcing. The locations of Tromsø and Halley stations are marked on the map, both being within the latitude 140 bands that are directly affected by the PsA-EEP forcing.
We simulate the time period between January 2010 and June 2011 (18 months). Three simulations were made, with different PsA-EEP forcing scenarios: 1) no-PsA, i.e. zero ionization for a reference; 2) full-PsA, i.e. the MCMC median ionization from Turunen et al. (2016), 3) thermo-PsA, like full-PsA but the ionization below 85 km (≈ 4 × 10 −3 hPa) set to zero. Comparisons between the full-PsA and no-PsA simulations gives us an estimate of PsA atmospheric impacts, while the thermo-PsA simula-145 tion can be used to separate the impacts from thermospheric and mesospheric forcing. All simulations included the background EPP forcing used in WACCM, i.e. solar protons, auroral electrons, and galactic cosmic rays. Note that the simulation period is in the ascending phase of the solar cycle right after a record minimum in solar activity, thus the background EPP forcing was relatively low, making it easier to identify the PsA-EEP impact. For example, maximum daily Ap in the 18-month period was 54.6, as opposed to the maximum Ap of 203.9 for the full cycle of 2001 -2011. The temporal resolution of the electron, NO x , HO x , and ozone data is one hour, thus the diurnal variations are included. In 160 the panels a -d, the absolute concentrations from the full-PsA simulation display the overall seasonal variability as well as the vertical distributions at 10 -110 km. The electron concentration increases towards higher altitudes due to the increasing ionization from the solar short-wave radiation and EPP. NO x displays two characteristic maxima, one in the stratosphere (at ∼ 30 km) and another in the lower thermosphere (at ∼ 110 km), and more (less) NO x in the mesosphere (stratosphere) during wintertime. Denitrification due to reactions with chlorine leads to very low concentrations in the lower stratosphere around 165 20 km during winter periods. The HO x concentration is higher during summer due to its production being driven by solar ultraviolet and Lyman-α radiation, and maximizes around the stratopause (at ≈ 50 km). In wintertime, largest concentrations are seen in the mesosphere. Ozone has two maxima in the summer: the stratospheric ozone layer peaking at 20-30 km and the secondary maximum at mesopause at around 90 km. In wintertime, the mesospheric ozone concentration is higher than in the summer due to less loss from diminished solar radiation and photodissociation. The tertiary maximum development in the 170 middle mesosphere around 70 km at the vicinity of the polar winter terminator contributes to higher wintertime concentrations as well.
Highlighting the PsA-EEP impact, differences between the full-PsA and no-PsA simulations are shown in the panels e -h of Figure 2. The electron concentration clearly enhances at 60 -90 km during summer/daytime due to the added PsA-EEP ionization. In general, in the wintertime and at night there is much less impact than in the summer. However, because of the 175 absence of solar radiation, a larger portion of the negative charge below 90 km is held by ions, not electrons (e.g. Verronen et al., Orsolini et al., 2018). Above 90 km, the wintertime differences exhibit a variation between increases and decreases which are relatively small and are related to the internal variability of the model coming from its free-running dynamics at the upper altitudes. Also, the ionization from auroral electrons, which is the same in all simulations, becomes dominant over PsA-EEP at altitudes above ≈90 km (not shown). The NO x concentrations show a similar, relatively small variability around 180 100 km, but only during the winter periods when there is more dynamical variability. Focusing on the main NO x features, increases are seen at 80 -90 km throughout the year. Early in the winter season, the NO x increase due to PsA-EEP is observed in the mesosphere, from where it further descends into the stratosphere reaching down to about 20 km by the end of the winter season. The increase related to the descent disappears and then appears several times during the winter. A layer of PsA-EEP NO x persists at about 25 km altitude until the end of the simulation period (midsummer). The HO x response shows some 185 diurnal variability, i.e. cyclic increases and decreases. Overall, however, there is an increase of HO x at 70 -80 km altitudes from the direct PsA-EEP impact. Around 60 km, the NO x increase leads to chemical loss of wintertime HO x concentrations Figure 7 presents the relative NO x and O 3 response at ≈ 40 km altitude, i.e. in the upper stratosphere. At the beginning of January, Tromsø is located within the polar vortex. There, the PsA-EEP impact is clearly seen as increased NO x and decreased O 3 concentrations. There is considerable variability of the relative impact within the vortex with a range of responses up to about +180% and down to about −6%, respectively. In the middle of January, the polar vortex has moved away from Tromsø.

260
Although the PsA-EEP impact inside the vortex is quite similar compared to the situation in the beginning of the month, this time none of it can be seen at the Tromsø location. The situation changes back at the end of January when the polar vortex is over Tromsø again. Thus the variability in the NO x and O 3 response seen in Figures 2, 5, and 6 is due to the evolution of the NH polar vortex over the winter period. Clearly, a global model like WACCM is a powerful tool when interpreting results from a single polar station like Tromsø, as demonstrated here. Although not shown, the SH vortex is much more stationary 265 with respect to the Halley location and the PsA-EEP impacts there do not display similar large variability.

Discussion
The particle forcing used in this study was recently validated by a statistical analysis of in-situ particle spectra from lowaltitude spacecraft measurements (Tesema et al., 2020). It was concluded that the spectrum does indeed represent the observed median spectrum for PsA particle precipitation very well. The simulation results presented in this study thus provide insight 270 into the effects of the median PsA-EEP forcing. Tesema et al. (2020) also showed that while the low flux PsA forcing causes no atmospheric changes, the high flux PsA forcing could severely deplete mesospheric ozone. It is therefore desirable to investigate both the atmospheric sensitivity threshold towards the low flux scenario as well as the NO x production and descent during the high flux forcing. As reported by Partamies et al. (e.g. 2017), the solar wind driving during pulsating aurora does not need to be extreme although the wind speed is typically elevated. This can be particularly important for simulation runs 275 during the solar minimum and the declining phase of the solar activity, because these time periods are known to associate with a frequent high-speed streams in the solar wind (Asikainen and Ruopsa, 2016). The question is whether PsA forcing during consecutive nights would lead to a stronger cumulative effect in the atmosphere than what we have seen in this study. Some variations in the PsA MLT extent is expected due to changes in the solar wind driving. The latitude extent of PsA, which maps to the ionosphere from the outer radiation belt source region, is likely to undergo little variability from event to event (Sandhu