ULF Geomagnetic Activity Signatures in the Atmospheric Parameters in Antarctica

The study of the solar wind-Earth’s magnetosphere and atmosphere interaction is a topic of great interest. The solar wind energy is transferred to the Earth’s environment also through ultralow frequency (ULF, 1 mHz–5 Hz) waves of the geomagnetic field, with higher efficiency at high latitudes where magnetic reconnection processes occur, making the polar cap an important laboratory for these investigations. Several studies suggest that the atmosphere responds to the geomagnetic activity driven by the solar wind, although the interaction processes are not yet completely understood. In this context, the results of recent investigations, showing the coupling on timescales of 1–2 days between geomagnetic ULF activity and the middle-low (h < 50 km) atmosphere in the polar cap, are summarized, based on geomagnetic measurements at Terra Nova Bay, in Antarctica (λ 80°S) and atmospheric parameters from the reanalysis dataset.


Introduction
In the latest years, the Sun-Earth environment is studied to explain observed physical phenomena in the context of space weather/climate, such as climate changes. It is well known that the Sun continuously transfers its energy to the Earth's environment through radiation and solar wind (SW). Although the Sun's radiation represents the main source affecting the Earth's atmosphere, the SW energy plays an important role during high geomagnetic activity time intervals [1]. On this regard, the magnetosphere-ionosphere represents a complex system able to partially convert SW impacting energy through nonlinearly related physical processes. Such effects are more evident at high latitudes where reconnection processes between interplanetary magnetic field (IMF), carried out by the solar wind, and magnetospheric field occur, making the polar cap an important laboratory to study the SW-atmosphere interactions. Solar wind-driven electrodynamic processes and ultralow frequency (ULF, 1 mHz-5 Hz) waves seem to lead to both diffusion and precipitation processes of energetic electrons in the outer radiation belts, leading also to chemical [2] and microphysical [3] processes in the atmosphere, characterized by different timescales.
The longer-term response, characterized by timescales of several weeks, is usually attributed to the odd nitrogen (NOx) production, due to precipitating energetic electrons, in the mesosphere and lower thermosphere. During the polar winter, NOx can live long enough to be transported downward into the stratosphere where it chemically perturbs the ozone distribution, altering the radiative balance in that region of the atmosphere.
This can in turn affect the overall circulation in the stratosphere, and such changes can propagate to the surface level, eventually leading to detectable changes in surface air temperatures, through dynamical coupling processes occurring on timescales of several weeks (e.g., [2,[4][5][6]).
Conversely, the shorter-time response (<1 day) of the atmosphere to the SWmagnetosphere coupling processes is probably related to changes in the atmospheric electrodynamics. It can be attributed to electric conductivity variations in the lower atmosphere by ionization mechanisms and/or to changes of the polar cap electric potential induced by SW perturbations [7]. The consequent modulation of the current density which flows from the upper boundary (being as low as 60 km, [8]) through the troposphere to the ground in the global electric circuit (GEC) [3,[9][10][11] could influence cloud formation through the release of latent heat, which in turn can affect atmospheric dynamics [12].
This work represents a review of our investigations on the experimental observation of the possible short (within $1 day) timescale response of the atmosphere to the SW dynamics, observed during 2003-2010, which correspond to the solar cycle 23 and the beginning of solar cycle 24. We analyzed the geomagnetic field variations monitored at the Mario Zucchelli station, at Terra Nova Bay (TNB, AACGM latitude λ = 80.01°S and longitude φ = 306.94°E) in Antarctica, and atmospheric parameters at tropospheric and stratospheric heights, provided by ERA-Interim and Monitoring Atmospheric Composition and Climate (MACC) reanalysis archives (http://apps.ecmwf.int/datasets).
The ERA-Interim is a global atmospheric reanalysis dataset, continuously updated in real time (see [13] and references therein). Global atmospheric and surface parameters from 1 January 1979 are available from the surface up to 0.1 hPa as atmospheric fields on model levels and pressure levels, with a temporal resolution of 6 h, and as surface fields with a temporal resolution of 3 h. The data assimilation system used to produce ERA-Interim is based on a 2006 release of the ECMWF Integrated Forecast Model (IFS Cy31r2). The MACC dataset is a global reanalysis dataset of atmospheric composition data, produced by assimilating satellite data into a global model and data assimilation system (see [14] and references therein). The system includes a four-dimensional variational analysis (4D-Var) with a temporal resolution of 12 h analysis window. The ERA-Interim and MACC data, at 1 day resolution used for our studies, have been retrieved from the Meteorological Archival and Retrieval System at ECMWF. Solar wind parameters and interplanetary magnetic field are monitored by using OMNI data, time-shifted to the bow shock nose (i.e., the subsolar position of the supersonic-to-subsonic transition regions) and collected on CDAWeb (http:// cdaweb.gsfc.nasa.gov). Geomagnetic activity was monitored by using a triaxial search-coil magnetometer data, recorded at TNB, at a sampling rate of 1 s.
The atmospheric parameters must be whitened to filter the longer period components (essentially 1 year and 6 months), which would obscure the weak effects produced by the ULF geomagnetic activity [15].
This review is structured as follows: in Section 2, we shortly introduce the interactions occurring between ULF waves and the energetic electrons in the outer radiation belt, leading particle precipitations, as well as polar cap potential difference related with SW parameters; in Section 3, we present the experimental evidence of SW effects on the atmospheric parameters in Antarctica at stratospheric and tropospheric heights; finally, in Section 4, we discuss the estimated timescale response of atmospheric parameters and the possible physical processes involved in SW-atmosphere coupling processes. For greater clarity, we described data analysis and methods in each section.

The solar wind-magnetosphere coupling processes at high latitudes
The estimation of the Earth's surface temperature and lower atmosphere energy budget significantly changes due to small amount, distribution, or radiative properties of clouds [16]: therefore, they represent one of the largest sources of uncertainty in predictions of climate change [17]. Even small atmospheric electrical modulations can affect aerosol nucleation processes and cloud condensation nuclei production in troposphere and thus modify cloud properties. In this regard, the polar cap electrodynamics and the energetic particle precipitation seem to be important SW-atmosphere coupling mechanisms, responsible for atmospheric changes on timescales from several weeks to days. It is also known that a global electric current flows in the global electric circuit. It is generated mainly by charge separation in clouds at the tropics and maintains the global ionosphere at a potential of about 250 kV. Variations above and below this value occur in the high-latitude regions due to SW-magnetosphere-ionosphere coupling processes. In this section, we briefly discuss the solar wind-magnetosphere coupling processes, which could produce observable effects in the stratospheric and tropospheric dynamics due to energetic particle precipitation from the outer radiation belt, as well as to the polar cap electrodynamics at high latitudes.
Generated by a variety of instabilities, ULF waves transport energy throughout the magnetosphere and are observed on the ground as continuous pulsations (Pc ,  Table 1). They can play important roles in the energization and loss of radiation belt particles (see [34] for a review). In particular, ULF waves can interact with the relativistic electrons (>300 keV) magnetically trapped in the radiation belts (L $ 5-7 Re, λ $ 60-70°, where L is the McIlwain parameter). In that regions, the charged particle is subject to gyro, bounce, and drift periodic motions (see Table 1), each one characterized by different timescales [35][36].
In particular, wave-particle interactions are theoretically predicted [37,38] because drift and bounce motion frequencies of trapped electrons are in the Pc5 (1-7 mHz) and Pc1-2 (100 mHz-5 Hz) frequency range, respectively. Experimental evidence confirms diffusion/acceleration of energetic electrons by Pc5 magnetospheric waves [39][40][41][42] and their precipitations after pitch angle scattering, due to gyro-resonant interaction with electromagnetic ion cyclotron (EMIC) waves [38][39][40][41]. Such waves are in the Pc1-2 frequency range and are generated at the magnetic equator by unstable distributions of ring current ions during geomagnetic storms [43]. Moreover, recent investigations show that Pc5 waves have their origin also in the leading edge of the corotating interaction regions (CIR) [42], while the origin of the Pc1-2 waves, observed at $80°latitude, appears to be due to substorm/storm-related instabilities and, in the dayside, to solar wind compressions of the magnetopause [44]. After conversion into Alfvén (shear) waves, ULF waves propagate along the geomagnetic field lines and can be observed on the ground at high latitudes [45,46]. An example of energetic electron flux enhancements observed at geosynchronous orbit (L $ 6.6) associated to Pc5 power fluctuations measured on the ground and in the magnetosphere is shown in Figure 1, [42]; the electron flux seems to be delayed by $2 days with respect to pulsation power.

The roles of relativistic electron precipitation and the polar cap electrodynamics in the global electric circuit model at high latitudes
As discussed by Tinsley and Yu [48], the galactic cosmic rays (GCR) flux is responsible for almost all the production of ionization below 15 km of altitude, which determines the conductivity in that region and at high latitudes. However, the MeV electrons and their associated X-rays produce ionization in the stratosphere and higher troposphere, which can affect the local conductivity (see also [7]). Moreover, due to electric potential difference between the ionospheric layer and the ground, a vertical current density is present: It is directed along the stratosphere-troposphere-ground direction z, and it varies horizontally, due to variations in the local vertical column resistance represented by its resistivity σ and by variations in the local ionospheric potential φ(E). While σ is affected by the GCR and precipitating MeV electron fluxes from the outer radiation belt into the atmosphere, φ(E) is strongly dependent on SW-magnetosphere coupling process. In particular, the potential increases in the polar cap where geomagnetic field lines map to magnetospheric regions characterized by coupling processes with the interplanetary magnetic field (IMF); because J z , flowing through clouds in the troposphere, responds to conductivity and potential changes occurring up to 120 km altitude, it is very effective in linking stratosphere and ionosphere with clouds. Recent reviews by Lam and Tinsley [3] and Mironova et al. [7] well discuss the atmosphere response to current density changes near the poles caused by the interaction of the solar wind with the geomagnetic field; such Periodicity correspondence between relativistic electrons and ULF waves is marked in bold. Table 1.
ULF wave classification and the characteristic timescales for the three types of trapped particle motion (see also [37]).
changes arise in addition to the day-to-day variability in J z , caused by thunderstorm activity and present at all latitudes.
In particular, the B y (dawn-dusk) and B z (south-north) components of the IMF represent the main parameters controlling the electrodynamics in the upper atmosphere (i.e., the ionosphere): B z is important in the reconnection process in the dayside magnetopause, while B y affects the latitudinal dawn-dusk electrostatic potential asymmetry [49].
An example of polar cap electric potential in the southern hemisphere is show in Figure 2, obtained using the Weimer [50,51] ionospheric electrodynamic model. In this example, we assigned values of AE5 nT for B y and B z , assuming a solar wind speed of V SW = 450 km/s and a solar wind number density of n SW =3cm À3 .Itis clearly seen that two regions of polar cap potential, characterized by positive (dawnward) and negative (duskward) values of approximately AE40 kV, are emerging during periods characterized by SW-magnetosphere coupling (i.e., during B z < 0), indicating higher dawn-dusk polar cap potential drop. Conversely, during closed magnetosphere (B z > 0, left panels), lower polar cap potential difference is obtained. It implies that E z increases at high latitudes during IMF-magnetosphere reconnection (B z < 0) conditions. Furthermore, both σ and E Z in Eq. (1) can be significantly influenced by geomagnetic activity. In this regard, ULF activity could play an important role in the GEC model, leading to the clouds' physical changes and radiative properties. Indeed, the ULF activity in general is associated to storms and substorms, i.e., to the occurrence of a southward IMF and, at high latitudes, polar cap electric field; therefore, ULF activity can be regarded also as a proxy of the polar cap potential difference [52,53].  [42]. It clearly emerges a higher correlation with the >600 keV electron flux at a shorter-time delay (1.8-2 days) with respect to the >2 MeV electrons (2-2.4 days), with an approximately 9 h difference, probably due to the timescales of the acceleration processes, in agreements with [47].

Experimental evidence of signatures of solar wind effects in the atmospheric parameters
In their study, Francia et al. [15] examined surface air temperature measured at the automatic weather station ENEIDE, located at TNB, during 2007-2008, while signatures of ULF activity and polar cap potential difference were found by [54] in the stratosphere and troposphere, using the ERA-Interim reanalysis dataset. Figure 3 shows the global wavelet (GW), i.e., the time-averaged wavelet [55,56], separately for summer and winter months, of Pc5 and Pc1-2 powers at TNB and of the polar cap potential difference φ cap [54].
Common power peaks emerge at $27 days, the Sun synodic rotation period, and its first subharmonics (13.5 and 9 days), in the SW-related parameter φ cap and correspondingly in the ULF activity. In particular, in Figure 4 signatures of ULF activity and polar cap potential difference in temperature T and zonal wind U at tropospheric and stratospheric heights above TNB are observed. In their work [54], cross-wavelet W xy and wavelet coherence γ 2 [56] are computed between SW-driven and atmospheric parameters; the analysis was supported by the Monte Carlo test for the estimation of significance levels.
In can be seen from Figure 4 that the correspondence at 27 days is high and statistically significant in both the troposphere and the stratosphere during winter months, while at 13.5 days it is restricted to h < 10 km. During summer, the crosswavelet W xy for T shows very low values and an ambiguous correspondence at the Computed polar cap electric potential, obtained using Weimer [50,51]    27 day periodicity for all input parameters. Interestingly, a W xy and γ 2 attenuation can be seen, especially during winter months, near to the tropopause position ($7-10 km). Regarding U, high and significant W xy and γ 2 values are found mostly in the stratosphere, at $27 and $13.5 day periodicities, during winter months. During summer the pattern is similar but with lower values.
Analyzing the tropospheric temperature, specific humidity Q and cloud cover (CC) in Antarctica, Regi et al. [57] found further experimental evidence of SWatmosphere coupling. In particular, the authors analyzed 54 high geomagnetic activity time intervals, by using the superposed epoch analysis with the Monte Carlo test [58].
The results, shown in Figure 5, revealed a clear correspondence between Pc1-2 and variations in temperature, specific humidity, and cloud cover. The most significant correspondence is found at latitudes higher than 85°S in both T and Q parameters. Positive variations occurred at the epoch and 1 day after, suggesting that Pc1-2 power affects the specific humidity and the tropospheric temperature within 1 day; negative variations occurred at 2 days after the epoch.
Regarding the cloud cover (zonal mean) composite averages, computed at low (LCC, h < 3 km), medium (MCC, 3 < h < 6 km), and high (HCC, h > 6 km) altitudes, (see Figure 5, bottom panels), we found that the main effect of the ULF activity on these parameters consists in an increase in the HCC and MCC at the epoch. Moreover, we also found a significant variation at 1 day after and a decrease at 2 days after the epoch.
As discussed in the Introduction, GEC affects atmospheric parameters, such as cloud cover through several proposed microphysical processes (e.g., [3,10,48]) able to produce changes in the atmospheric parameters, such as temperature. In particular, the accumulation of charge on droplets and aerosol particles, most importantly the interstitial cloud condensation nuclei (CCN) and ice-forming nuclei (IFN), directly affects scavenging rates.
Scavenging is due to collisions between the nuclei and droplets, entailing sizedependent collection of nuclei and changes in their size distribution and overall concentration. It affects a number of microphysical processes, which cause changes in macroscopic cloud properties and in turn partitioning of energy flow in the system. In particular, the charge can increase or decrease the scavenging rates, depending on size, changing the concentrations and size distributions. Size distribution changes in CCN produce size distribution changes in droplets, affecting coagulation, precipitation, latent heat transfer, and cloud cover. Scavenging of iceforming nuclei by supercooled droplets promotes contact ice nucleation, releasing latent heat. The latent heat changes cause storm invigoration [59] and in winter storms can cause changes in the amplitude of Rossby waves and blocking. Since the typical lifetime of CCN in an air mass can be up to 10 days, the change in their properties can affect also later cycle of evaporation/condensation, flux of latent heat, and the amount of water vapor released into the troposphere. Figure 6 shows the results of superposed epoch analysis (SEA) conducted by [57] on Q and T at the epoch for different orientations of the IMF. It shows correspondence in T and Q, more evident during the IMF B z < 0 and B y >0 conditions, i.e., when the interplanetary electric field is efficiently transmitted into the high-latitude ionosphere and the dawn-dusk polar cap potential difference φ cap , due to E y = ÀV SW B z , as well as the vertical E z = V SW B y , increases (see also Figure 2). In particular, the vertical electric field during southward IMF conditions can be directly transferred in the ionospheric polar cap in the southern hemisphere, leading to a vertical current J z = σE z from the ionosphere to the ground through the stratospheric and tropospheric layers.

The observed timescales of the response of the atmosphere to SW variations
Timescales are important in order to discriminate the involved physical processes in the SW-atmosphere coupling. In their pioneering work, Francia et al. [15] found that, during 2007 and 2008, the ULF activity, in both Pc1-2 and Pc5 frequency ranges, is correlated with the surface air temperature with different delays. Their results, shown in Figure 7a, indicate that the temperature is significantly correlated with the Pc1-2 power at a time lag of 1 day. Although lower, the correlation with the Pc5 power is also significant, reaching the maximum value when the temperature is delayed by 3 days with respect to the Pc5 power.
In the meanwhile, Regi et al. [57] computed the cross-correlation between high cloud cover (HCC, h > 6 km) over Antarctica and logPc1-2 time series at TNB during 2003-2010, band-pass filtered at $27 days. They found several Antarctic regions characterized by significantly high correlation values. Examples of positive cross-correlations are shown in Figure 7b. Their experimental results suggest an average time delay of approximately 1 day of the response of the atmospheric cloud cover to the SW-driven Pc1-2 ULF power intensification. Such delay is consistent with the timescales of electrodynamicsÀ/microphysical-related processes here proposed. In [57] a possible relationship with large-scale atmospheric transport was investigated by means of a correlation analysis between Pc1-2 power and CH 4 ($27 day band-pass filtered) concentrations from MACC dataset. CH 4 is a long life chemical tracer, commonly used to characterize the middle atmosphere transport, also analyzing it in the framework of quasi-Lagrangian or conservative coordinate systems [60] and as a proxy to validate satellite measurements [61]. The results of the correlation analysis between Pc1-2 power and CH 4 , during 2003-2010 (not shown here), indicate that the correlation coefficients are generally lower and not significant and superimposable to the other tropospheric parameters. They suggest that the transport is not modulated by the 27 day periodicity of the ULF activity. However, the different time delays in different years could be due to variations in large-scale transport pattern (see [57] for details).

Summary and conclusions
In this work we presented a short review of our experimental results regarding the possible relationship between SW and atmospheric parameters at high latitude, in Antarctica, at Terra Nova Bay. Examining possible relationship between geomagnetic activity in the Pc1-2 and Pc5 frequency range and the polar cap potential difference with stratospheric and tropospheric parameters, the results provided in [54] can be summarized as follows: • Common power peaks emerge at $27 days, the synodic Sun rotation period, and its first subharmonics (13.5 and 9 days), in the polar cap potential difference φ cap and in the ULF activity ( Figure 3) and in the temperature and zonal wind at tropospheric and stratospheric altitudes ( Figure 4).
• The correspondence is more evident during winter, when solar radiationdriven processes are absent.
• Around the tropopause, approximately at 8 km, the correspondence is low, and in the stratosphere it appears mostly at the 27 day periodicity.
Further investigations [57] at tropospheric heights at high latitudes indicate that SW-driven electrodynamic processes and energetic particle precipitation related with enhancement of Pc1-2 activity can affect tropospheric temperature, specific humidity, and cloud cover. The response is quick (within $1 day) at ground and in the troposphere. These results suggest that the electrodynamics modulate the physical properties of clouds, probably through electron scavenging microphysical mechanism. It is a matter of fact that electron scavenging is strongly dependent on vertical tropospheric-stratospheric conductivity variations, due to energetic particle precipitation driven by ULF waves, and on the vertical electric current, modulated by polar cap potential, associated to SW-magnetosphere reconnection processes. More recently, evidence of an SW signature in the mesosphere [62] has been published, indicating that the SW can affect the atmosphere through the whole atmospheric column. As discussed by [63], the processes involved in each atmospheric layer are almost certainly different, and transport phenomena could be important.
Our conclusions are supported by the observed short (<1-2 days) delay response in the atmospheric parameters at troposphere altitudes and at ground with respect to the much longer delay expected for chemical mechanism [15,57]. However, this matter should be further investigated as underlined by [63] in particular through the examination of the time delays at stratospheric and mesospheric altitudes and at lower latitudes; the study of the dependence on different interplanetary conditions might be also useful for a more deep understanding of the atmosphere response to the SW.
Science Center for Space Science and Engineering Research (Virginia Tech) who provided the ionospheric electrodynamic model. The authors would also like to thank the MACC and ERA-Interim project data provider. The reanalysis data were provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) and can be downloaded from the ECMWF Data Server (http://apps.ecmwf.int/ datasets). Measurements of the geomagnetic field fluctuations at Terra Nova Bay are supported by the Italian PNRA (Programma Nazionale di Ricerche in Antartide, PdR2013/B2.09).