Diagnostic study of geomagnetic storm-induced ionospheric changes over VLF signal propagation paths in mid-latitude D-region

. We performed a diagnostic study of geomagnetic storm-induced disturbances that are coupled to the mid-latitude D-region by quantifying the propagation characteristics of very low frequency (VLF) radio signals from transmitters located in Cumbria, U.K. (call sign GQD) and Rhauderfehn, Germany (DHO) and received in southern France (A118). We characterised the diurnal VLF amplitudes from two propagation paths into ﬁve metrics, namely the mean amplitude before sunrise (MBSR), the midday amplitude peak (MDP), the mean amplitude after sunset (MASS), the sunrise terminator (SRT) and the sunset 5 terminator (SST). We analyzed and monitored trends in the variation of signal metrics for up to 20 storms, to attribute the deviations in the signal amplitudes that were attributable to the storms. Five storms and their effects on the signals were examined in further detail. Our results indicate that relative to pre-storm levels the storm-day MDP exhibited characteristic decreases in about 80 % (67 % ) of the events for the DHO-A118 (GQD-A118) propagation paths, respectively. The MBSR showed decreases of about 60 % (77 % ) whereas the MASS decreased by 67 % (58 %


Introduction
The terrestrial magnetosphere is formed by the interaction between the solar wind and the earth's magnetic field (McPherron et al., 2008).In contrast, the ionosphere is largely the result of solar photoionization of the neutral atmosphere balanced against chemical recombination and particle transport (Prolss, 2004;Kelley, 2009).The magnetosphere and ionosphere are coupled via the geomagnetic field effectively tying these seemingly disparate regions into a global magnetosphere-ionosphere (M-I) system (Blanc, 1988;Nwankwo et al., 2016).Within the concept of an open magnetosphere (Dungey, 1961) energy is transferred from the solar wind to the M-I system via magnetic reconnection (Cassak, 2016).Variations within the interplanetary environment, driven by solar disturbances, affect the M-I system particularly when the interplanetary magnetic field (IMF), embedded within the solar wind, is oriented southward relative to the outer northward-directed geomagnetic field (Gonzalez et al., 1999;Liu and Li, 2002).A geomagnetic storm is triggered when the solar-terrestrial interaction is sufficiently intense to energize the ring current (Jordanova et al., 2020) and to solicit a negative response in ground-based magnetometers in terms of the Disturbance storm time (Dst) index (Russel et al., 1974;Mayaud, 1980;Borovsky and Shprits, 2017).The strength of a geomagnetic storm is typically classified as small, moderate or intense for Dst values less than (<) -30 nT, < -50 nT, and < -100 nT, respectively (Gonzalez et al., 1994).The two leading drivers of geomagnetic storms are coronal mass ejections (CMEs) and corotating interactive regions (CIRs) (Baker, 2000).A CME is the impulsive release of solar material into interplanetary space from a solar active region that may, or may not, be associated with a solar flare (Youssef, 2012).Conversely, a CIR is the result of a high-speed stream (HSS) emitted from a solar coronal hole overtaking the background solar wind (Choi et al., 2009).In both cases the shock front at the leading edge of the interplanetary disturbance increases the ram pressure imposed on the dayside magnetopause causing a reconfiguration of the dayside Chapman-Ferraro current system (Chapman and Ferraro, 1930) and, in turn, the magnetosphere as a whole (Ganushkina et al., 2018).A typical geomagnetic storm has three phases consisting of 1) a sudden storm commencement (SSC) at the time of the increased ram pressure, 2) a main phase as the ring current is energized and 3) a recovery phase as the magnetosphere returns to a more quiescent ground state (Akasofu, 2018;Gonzalez et al., 1994).There are distinct differences between how and when CME-driven versus CIR-driven storms affect the earth (Borovsky and Denton, 2006).Typically, CME-driven storms are stronger, with regards to Dst (Tsurutani et al., 2006) and occur more frequently near the maximum in the sun's 11-year activity cycle (Gonzalez et al., 1999) whereas CIR-driven storms are weaker and occur predominantly on the decreasing phase of solar activity (Tsurutani et al., 1995).While the impacts of both CME and CIR-driven geomagnetic storms on the middle-to-upper atmosphere have been extensively studied and well known (Fuller-Rowell et al., 1994;Burch, 2016;Heelis and Maute, 2020), less certain are the geomagnetic storm effects in the lower ionosphere (Lastovicka, 1996;Kumar and Kumar, 2014).For the purposes of this report our attention is limited to CME-driven, moderate geomagnetic storms and the resulting impacts on D-region Very Low Frequency (VLF) radio-wave propagation within the earth-ionosphere wave guide (EIWG).The VLF frequency band spans the range from 3 kiloHertz (kHz) to 30 kHz. due to the loss of solar radiant photons whilst diffusive recombination continues unabated such that the D-region seamlessly coalesces into the lower E-region at about 90 km in the absence of other significant ionization sources (Thomas et al., 2007;Thomson and McRae, 2009).Prior to the space age, the detection of sudden ionospheric disturbances (SIDs) in the amplitude and phase of VLF radio-wave transmissions within the D-region was used as an established proxy technique for monitoring the occurrence of solar flares (Lincoln, 1964;Moral et al., 2013;Hegde et al., 2018) which were known to have a deleterious impact on radio-wave communications (Sauer and Wilkinson, 2008;Dellinger, 1937).The widely accepted standard for specifying the ionospheric electron density profile (EDP) is the empirically based International Reference Ionosphere (IRI) model which has evolved over time (Rawer et al., 1978;Rawer, 1981;Bilitza, 1990Bilitza, , 2001;;Bilitza and Reinisch, 2008;Bilitza, 2018) with special consideration given to the D-region of the lower ionosphere (Bilitza, 1981;Friedrich and Torkar, 1992;Danilov and Smirnova, 1995;Bilitza, 1998).An alternative approach for specifying the D-region is the use of specialized atmospheric models, such as the Whole Atmosphere Community Climate Model with D-region ion chemistry (WACCM-D), which are focused on the neutral atmosphere (Verronen et al., 2016;Andersson et al., 2016;Siskind et al., 2017) which, of course, is coupled to the ionized atmosphere via chemistry (Turunen et al., 1996;Schunk, 1996Schunk, , 1999;;Verronen et al., 2005;Turunen et al., 2009;Kovacs et al., 2016;Verronen et al., 2016;Turunen et al., 2016;Miyoshi et al., 2021) within the overall M-I system.
Techniques used to probe the ionosphere include both ground-based and space-based approaches.The earliest methods used by Appleton andBarnett (1925b, 1926) in short-range transmitter-to-receiver trials differentiated the reflected skywave from the direct groundwave to determine the height of the E-layer.About the same time, Merve Tuve and Gregory Breit (Tuve and Breit, 1925;Breit and Tuve, 1925) proposed a methodology of using pulsed radio-wave transmissions for measuring the heights of overhead reflecting layers.The Tuve-Breit methodology was the basis for ionosondes which for many years was the preeminent scientific technique for "sounding" the ionosphere (Bibl, 1998).Ionosondes typically operate in the high frequency (HF) domain from 3 to 30 MHz to derive the vertical ionization profiles of the E and F regions.Ionosondes are quite affordable leading to their widespread use in ionospheric characterization (Stamper et al., 2005).However, a key limitation is that ground-based ionosondes can effectively measure only the "bottom-side" ionosphere at and below the F-region peaks (Reinisch and Xueqin, 1983).A variant is the space-based "topside sounder" approach from which measurements of the topside ionosphere, above the F2 peak, can be obtained (Chapman and Warren, 1968;Benson, 2010).Ionospheric profiles of the E and F regions can also be obtained using incoherent scatter radars (ISR) (Robinson et al., 2009) which are typically operated at several hundred-megahertz (MHz) within the ultra-high frequency (UHF) range (Häggström, 2017) relying on the principle of Thomson electron scattering (Farley et al., 1961;Dougherty andFarley, 1961, 1963).Conversely, coherent scatter HF radars (Greenwald et al., 1995) rely on Bragg scattering (Takefu, 1989) from ionospheric structures to probe the ionosphere (Häggström, 2017).Examples of these advanced ionospheric measuring technologies include the Alouette topside sounder (Jackson, 1986), the Millstone Hill ISR (Evans, 1969a, b) and the Super Dual Auroral (HF) Radar Network (SuperDARN ) (Greenwald , 2021).More recently, researchers have leveraged the outstanding capabilities of Global Navigation Satellite Systems (GNSS), initially the Global Positioning System (GPS), for ionospheric characterization (Davies and Hartmann, 1997) using receivers on the ground (Mannucci et al., 1998;Prol et al., 2021) and in space (Mannucci et al., 2020).The Continuously Operating Reference Stations (CORS) (Snay and Soler, 2008) is a good example of a ground-based GNSS network whereas the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) (Yue et al, 2014) is an example of the related space-based approach.With regard to the lower E and D region, none of the aforementioned technologies are particularly effective at monitoring the bottommost ionospheric region (Cummer et al., 1998;Kumar et al., 2015) which is the focus of the present work.However, the perspective provided by considering the state of the local ionosphere allows us to assess our findings within the context of a coupled M-I system.
Reliance on VLF waves has proven to be an effective tool for monitoring and characterizing the lower ionosphere (Sechrist, 1974;Inan et al., 2010;Gross and Cohen, 2020).Early research within the VLF band was focused on lightning-induced "sferics" (Pierce, 1969), a technique that formed the basis of modern lightning detection networks (Betz et al., 2009).VLF sferics can travel significant distances within the EIWG (Wait ans Spies, 1964) and be detected as sound "tweeks" due to the dispersive nature of the ionosphere (Singh et al., 2016).D-region characteristics can be derived from these events although some care is required to account for their sporadic nature, both temporally and spatially (McCormick and Morris, 2018).Conversely, controlled experiments using known VLF frequencies and transmitter-receiver great circle paths (TRGCPs) can be used to characterize the lower ionosphere nominally to and from fixed locations (Kumar and Kumar, 2020).Well-known features in VLF TRGCP propagation are diurnal variations in amplitude and phase (Yokoyama and Tanimura, 1933;Pierce, 1955;Taylor , 1960;Chilton et al., 1964;Lynn, 1978;Barr et al., 2000;McRae and Thomson, 2000;Sharma and More, 2017) and characteristic signatures of sunrise and sunset (Walker, 1965;Crombie , 1964Crombie , , 1966;;Ries, 1967;Samanes et al., 2015;Sharma and More, 2017;Gu and Xu, 2020).These features can be explained within the context of the wave-mode theory of VLF propagation attributed to Budden (1951Budden ( , 1953Budden ( , 1957) ) and promulgated by Wait Wait (1960, 1961, 1963, 1964, 1968, 1970) and collaborators (Spies and Wait, 1961;Wait ans Spies, 1964).While the transient response of VLF TRGCP propagation to solar flares has been well documented (Mitra, 1974;Thomson and Clilverd, 2001;McRae and Thomson, 2004;AbdRashid et al., 2013;Palit et al., 2013;Rozhnoi et al., 2019), the associated characterization to geomagnetic storms has been less quantified due, perhaps, to the mixed ionospheric responses in the lower atmosphere from geoeffective CMEs and CIRs (Turner et al. , 2006;Kim et al., 2008;Laughlin et al., 2008;Verbanac et al., 2011;Soni et al., 2020).The focus of this effort is then to augment the limited body of research related to the impact of geomagnetic storms on VLF propagation along TRGCPs within the EIWG (Tatsuta et al., 2015).
The ionospheric response to geomagnetic storms is varied and interpreting the response in terms of a regional or global specification of electron density often requires the use of sophisticated environmental models (Schunk et al., 2004;Immel and Mannucci, 2013;Greer et al., 2017).However, these models are mostly focused on the upper ionosphere, E-region and above, and inclusion of the D-region involves separate modules that, in turn, are mostly focused on the prompt D-region to solar transient events, in the forms of flares and related energetic particle events (Bilitza, 1998;Eccles et al., 2005;Sauer and Wilkinson, 2008;Rogers and Honary, 2014;Kulyamin and Dymnikov, 2016).The coupling between the dayside upper and lower ionospheric regions is tenuous as the electron density of the upper ionosphere is largely driven by solar EUV radiation whereas chemistry, mostly involving nitric oxide (NO), controls the quiescent D-region (Siskind et al., 2017).To facilitate the development of improved D-region models the impacts of geomagnetic storms must be considered (Spjeldvik and Thorne, 1975;Dickinson and Bennett, 1978).Particularly germane to this present discussion is the impact of post-storm energetic particle precipitation (EPP) on the chemistry of the D-Region (Seppala et al., 2015;Rodger et al., 2015).Indices used to specify the D-region include the reflection height, H' (in km), and bottom-side profile sharpness, β (in km −1 ), parameters originally developed by James Wait (Wait ans Spies, 1964) and commonly used in research applications (Thomson, 1993;Thomas et al., 2007;Nina et al., 2021) as well as operations involving VLF propagation within the EIWG (Nunn et al., 2004).Another pair of indices refer to the sunset D-Layer Disappearance Time (DLDT)) and the sunrise D-Layer Preparation Time (DLPT) associated with anomalous ionospheric behavior during geomagnetic storms (Choudhury et al., 2015) and earthquakes (Sasmal and Chakrabarti, 2009;Chakrabarti et al., 2010).Within this manuscript we choose to adopt a modified set of indices originally introduced by Nwankwo et al. (2016) which correspond to the mid-day signal amplitude peak (MDP), the mean signal amplitude before sunrise (MBSR) and mean signal amplitude after sunset (MASS).The use of these indices within the effort by Nwankwo et al. (2016) revealed trends, albeit inconsistent, in the signal strength of VLF radio waves propagating within the EIWG in response to several geomagnetic storms of moderate intensity.The intent of the current effort is to expand on the findings of Nwankwo et al. (2016) to further elucidate the effects of geomagnetic storms on the physics of the mid-latitude D-region.The following paragraphs review the status of geomagnetic storm related impacts on the high-, middle-and lowlatitude D-region ionosphere (Lastovicka, 1996).
It was previously noted that solar Lyman alpha radiation is primarily responsible for maintaining the quiescent dayside Dregion (Nicolet and Aikin, 1960) whereas flare-associated X-rays are the dominant D-region ionization source during solar flares (Thomson and Clilverd, 2001;Quan et al., 2021).Solar-flares are transient events which actively emit ionizing X-rays lasting for up to several tens of minutes (Veronig et el., 2002).A related class of solar transient is a solar particle event (SPE) resulting from an interplanetary CME-induced shock (Tsurutani et al., 2003;Mittal et al., 2011;Chandra et al., 2013;Dierckxsens et al., 2015;Gopalswamy, 2018), wherein charged particles, mostly protons, are accelerated to high energies and impact the lower atmosphere at the higher latitudes within the open magnetosphere (Zawedde et al., 2018).The precipitation of greater than 10 MeV solar energetic protons (SEPs) affects both the chemistry of the lower mesosphere (Ahrens and Henson , 2021), between 50 to 85 km (Turunen et al., 2009), and acts as a source of ionization that can temporarily increase the D-region electron density (Hunsucker, 1992;Sauer and Wilkinson, 2008;Neal et al., 2013).A sufficiently intense and energetic flux of SEPs can impact radio-wave communications in the form of a Polar Cap Absorption (PCA) event having a delayed onset following a solar flare, assuming the flare has a related CME, and a duration lasting for up to several days (Rose and Ziauddin, 1962;Potemra et al., 1970;Mitra, 1974;Rogers and Honary, 2014;Rogers et al., 2016).The Antarctic-Arctic Radiation-belt (Dynamic) Deposition-VLF Atmospheric Research Konsortium (AARDDVARK) network was established to probe the D-region with extreme sensitivity (Clilverd et al., 2009(Clilverd et al., , 2014;;Neal et al., 2015).An example of an early use of the AARDDVARK network was monitoring changes in the polar D-region from an SPE (Clilverd et al., 2007).Strictly speaking an SPE is quite distinct from a geomagnetic storm although they both have a common originating source.However, as it relates to our objective of better quantifying the D-region response to geomagnetic storms we are unaware of any efforts to separate out the storm-specific responses from other, albeit related, sources.
During and following geomagnetic storms enhanced fluxes of energetic electrons from the outer van Allen radiation belts (van Allen et al., 1958) precipitate into the sub-auroral atmosphere (Peter et al., 2006) contributing to the formation of the storm-time mid-latitude D-region (Pedersen, 1962;Grafe et al., 1980;Horne et al., 2009;Zawedde et al., 2018;George et al., 2020).The nominal L-shell location (McIlwain, 1961) for the outer radiation belt is from L∼3 to L∼10 (George et al., 2020) which, for an idealized earth dipole field, maps to invariant latitudes of 54.80 and 71.60, respectively (Kilfoyle and Jacka , 1968).The processes which regulate the electron populations within the outer belts during storm conditions are complicated and interdependent (Reeves and Daglis, 2016;Baker, 2019).An aspect not yet discussed is the fundamental role that aurora/magnetospheric substorms play in the dynamics of the magnetosphere (Akasofu, 1964;McPherron, 1979;Rostoker et al., 1980;Spence, 1996;Akasofu, 2020) and, germane to the topic at hand, the impacts on the mid-latitude D-region (Guerrero et al., 2017).A substorm is described as a "transient process initiated on the night side of the earth in which a significant amount of energy derived from the solar wind-magnetosphere interaction is deposited in the auroral ionosphere and magnetosphere" (McPherron, 1979).This description is consistent with the concept of an open magnetosphere wherein the IMF and geomagnetic field lines merge at the dayside magnetopause and are then swept tailward with the solar wind into the nightside, or geotail, where the geomagnetic field lines and IMF respectively reconnect (Russell, 1991).When the IMF has a southward component the magnetopause standoff distance at noon is nominally located at about 10 earth radii (Re) (Aubry, 1970;Fairfield, 1971;Shue et al., 1997;Suvorova and Dmitriev, 2015;Bonde et al., 2018;Samsonov et al., 2020).During a geomagnetic storm the magnetopause can be significantly "eroded" (Wiltberger et al., 2003;Le at al., 2016) reducing the location of the standoff distance which may be within the geostationary orbit of 6.7 Re under extreme conditions (Shue et al., 1998).In response to the storm, the polar-cap potential and the cross-tail current increase as energy is continually transferred from the solar wind into the geotail (Angelopoulos et al., 2020).While it is curious that the electron density within the outer radiation belt can either increase or decrease under storm conditions (Reeeves et al., 2020) more relevant to establishing (nightside) or maintaining (dayside) the ionospheric D-region is that the electrons within the outer belt can be pumped up to extremely high, including relativistic, energies by local wave activity and radial diffusion (Baker, 2019;Kanekal and Miyoshi, 2021).A fraction of these energetic electrons can be subsequently scattered into the atmospheric loss cone (Porazik et al., 2014) by naturally occurring electromagnetic waves (Spjeldvik and Thorne, 1975;Gu et al., 2020;Ripoll et al., 2020;Aryan et al., 2021) and precipitate into the lower ionosphere at mid-latitudes where they collisionally ionize the neutral atmospheric constituents (Rodger et al., 2007(Rodger et al., , 2010(Rodger et al., , 2012;;Naidu et al., 2020).The outer radiation belt can relax to its more quiescent state on time scales ranging from minutes (Turner et al., 2013) to many days (Baker, 2019) following a significant geomagnetic storm.It should be noted, in conclusion, that although the evidence provided herein shows a clear association of magnetic storms with enhanced levels of electron precipitation it is likely that even during geomagnetically quiet intervals electron precipitation persists but apparently at a greatly reduced rate (Mironova et al., 2021).
As previously discussed, Appleton and contemporaries (Colwell and Friend, 1936) in the early 1900's used short-distance VLF radio-wave transmissions to probe and study the D-region whereas the efforts of Wait and colleagues (Wait ans Spies, 1964) were to facilitate global VLF communications within the EIWG.The early work to quantify storm-related D-region impacts involved the general technique of HF radio-wave absorption (Eccles et al., 2005;Pederick and Cervera, 2014;Scotto and Settimi, 2014;Siskind et al., 2017).In this regard Lauter and Knuth (1967) found that the aftereffects of geomagnetic storms on the absorption of 245 kHz radio waves in the mid-latitude D-region could persist for more than 10 days.Supporting in-situ rocket data (Dickinson and Bennett, 1978) revealed that the electron density in the days following an "intense" geomagnetic storm could be 4 to 10 times the normal daytime density and that these measurements were well correlated with changes in HF radio-wave absorption.An interesting finding by Satori (1991) was the countering effect of a Forbush decrease on the density of the D-region at mid-latitudes.A Forbush decrease refers to the measured reduction in the galactic cosmic ray background due to an earth-passing CME (Forbush, 1954;Raghav et al., 2020;Janvier et al., 2021).Cosmic rays are an important D-region ionization source at night and a minor contributor during the day (Moler, 1960).According to Satori (1991), the reduction in the galactic cosmic ray flux associated with a CME counters the increased D-region ionization from precipitating radiation belt electrons during a geomagnetic storm.
Again, the focus of this report is on the geomagnetic storm-related impacts to VLF TRGCP radio-wave propagation and the information on the D-region that can be gleaned from this approach.In this regard (Belrose and Thomas, 1968) reported that VLF amplitudes during a geomagnetic storm were unaffected whereas phase measurements showed rapid fluctuations with residual effects lasting several days following the storm.Muraoka (1979) found that the prevalence of these phase anomalies was dependent on the strength of the magnetic storm.Upon further examination, as reported by Rodger et al. (2007), both signal amplitude and phase variations in VLF TRGCP transmissions were found to be sensitive to electron precipitation events during geomagnetic storms.Choudhury et al. (2015) found that receiver position electron density was the main controlling factor in their storm-time metric of the sunrise DLPT depth in VLF TRGCP radio-wave propagation.This metric plus the MDP parameter (after Nwankwo et al., 2016) were used by Naidu et al. (2020) to ascertain, again, that geomagnetic-storm related impacts to the D-region were due to energetic electron precipitation.Recently Kerrache et al. (2021) clarified the role of lightning-induced electron precipitation (LEP) events (Voss et al., 1998;Blake et al., 2001;Inan et al., 2010) in the pitch angle scattering of outer radiation belt electrons during geomagnetic storms and their impact on VLF TRGCP transmissions within the EIWG.
The mechanisms that affect the low-latitude D-region in response to geomagnetic storms have not been extensively studied and are still relatively unknown (Araki, 1974;Kleimenova et al., 2004;Kumar and Kumar, 2014;Kumar et al., 2015;Maurya et al., 2018).It is recognized that the low-latitude E-and F-region ionospheres can be affected by storm-induced prompt penetration electric fields (PPEFs) (Tsurutani et al., 2008;Timocin, 2022) and disturbance dynamo electric fields (DDEFs) (Blanc and Richmond, 1980;Fejer et al., 1983;Scherliess and Fejer, 1997) plus related substorm effects (Sastri, 2006;Chakraborty et al., 2015;Hui et al., 2017).Generally speaking, PPEFs refer to the immediate and sustained fields resulting from the active response of the magnetosphere to an externally imposed forcing function such as a CME-induced shock (Nava et al., 2016), whereas DDEFs result from the delayed equatorward motion of the thermosphere in response to auroral heating and chemistry (Zesta and Oliveira, 2019;Robinson and Zanetti, 2021).Transient substorms affect the dynamics of the low-latitude ionosphere as the magnetotail attempts to accommodate an increased storm-time reconnection rate (Hajra, 2021).However, these storm-time perturbations do not appear to affect the lower ionosphere except under the most extreme situations.For example, in a limited study of 7 moderate (Dst < -50) to intense (Dst < -100) geomagnetic storms Kumar and Kumar (2014) found that only the intense storm of 16 Dec 2006, with a Dst = -145, had a clear measurable effect on VLF TRGCP transmissions at low latitudes.These finding are consistent with early studies (Araki, 1974) on the effects of large storms on trans-equatorial VLF propagation.Other case studies of intense (Kumar et al., 2015;Maurya et al., 2018) to super (Dst < -200, Wu et al., 2016) geomagnetic storms validate this D-region response although the specific mechanisms for coupling the low-latitude D-region ionosphere to the higher latitudes in terms of gravity waves or chemical processes have not yet been confirmed.
We have provided an extensive and, hopefully, well documented background on the use VLF radio waves to probe the ionospheric D-region with an emphasis on the impacts of geomagnetic storms.Monitoring VLF radio-wave propagation within the EIWG along TRGCPs is a convenient and cost-effective technique for determining how space weather affects this lowest traditional ionospheric density layer.While the D-region is an important enabling element for VLF communications, it is also an interface linking the middle atmospheric regions to the upper ionosphere.We have stressed the role of the D-region within an M-I perspective of a driven system during geomagnetic storms perhaps at the expense of expanding on the impacts to and the feedback from the neutral atmosphere.While much of the basic physics and chemistry of the D-region is well understood how geomagnetic storms affect this region of space remains an active area of research.Our goal within this manuscript is to contribute in some small way to the existing body of knowledge concerning the D-region as it pertains to VLF radio-wave propagation.Therefore, we (in this study) combine the observed diurnal VLF amplitude variation in the D-region with standard measurements of the E and F regions to perform a diagnostic investigation of coupled geomagnetic storm effects, in order to understand the observed storm-induced variations in VLF narrowband based on the state and responses of ionosphere.The analysed storm intervals are 16th-31st September and 22 October-5 November 2011, while the geomagnetic storms of interest within the intervals include the events on 17, 26-27 September, 25 October and 1 November 2011.We monitor variation 2-4 hour mean VLF signal amplitude before local sunrise and after sunset (hereafter respectively denoted as MBSR and MASS), and the mid-day signal amplitude peak (MDP).We note that the acronym DTMA (daytime mean amplitude) was used in Nwankwo et al. (2021) instead of the MDP.The difference between the two metrics is that DTMA represent 1-hour mean value of the signal amplitude around midday, while MDP is the value of the signal amplitude at midday.We also identified typical values of the signal at sunrise and sunset, also recognised as sunrise and sunset terminators (hereafter, denoted as SRT and SST).
When propagating in the EIWG VLF signal amplitude and phase responds to dynamic conditions in the ionosphere such as variation in the reflection height and/or width of the EIWG caused by changes in ionization rate.VLF/LF also response to other solar-induced phonomena (e.g., SID, solar eclipse and geomagnetic stormes-induced energetic particle penetration) as well as non-solar phenomena (e.g., gravity waves, lightening) (Silber and Price, 2017) as extensively discussed in the introduction section.In addition to the diurnal variation of the ionosphere, the received signal also reflects the state of the traversed region of the ionosphere between the transmitted and the receiver (Nwankwo et al., 2020b).Therefore, monitoring the trends in variation of all these metrics enabled us study the response of the VLF radio waves to geomagnetic storms as they pertain to the induced ionospheric pertubations in the waveguide and the diurnal signature or variation in amplitude.The diurnal VLF amplitude indicating portion of the characterised metrics (MBSR, MDP, MASS, SRT and SST) are shown in Fig. 2a.The signals were analysed in conjunction with geomagnetic indices, to describe storm-induced magnetosphere-ionosphere coupling in mid-latitude D region.We thus study the trend in variations of these key metrics under varying geomagnetic storm conditions using the signal propagation characteristics, to understand behaviours attributable to geomagnetic storm-induced variations in the lower ionosphere (besides the visible response of the signal's amplitude and/or phase to solar flare induced X-ray flux).presumably driven by the significant increase in V sw and P D on 17th and 26th (Fig. 3a-f).However, the main reference storms are those of 17th and 26th.The variation of the AE (especially between 26th and 29th) appear to be consistent with storm related high-intensity, long-duration continuous AE activity events (HILDCAAs).Hence, 'fresh energy was injected' into the magnetosphere in the process (Tsurutani et al., 2011).We observed a notable drop in DHO-A118 VLF signal level on 26th around midday following the relatively intense storm condition with Dst up to -101 (Fig. 3a).This scenario (signal strength decrease) have been associated with storm-induced variations in energetic electron precipitation flux (Kikuchi and Evans, 1983;Peter et al., 2006).During a geomagnetic storm, the current system in the ionosphere, and the energetic particles precipitate   4a), we observed a dipping of the MDP on 17th (extending to 20th), as well as dipping of the MASS on 17 Sept., but an increase of the MBSR, SRT and SST.Following the recurrent storms between 26 and 28 Sept., we observed dipping of the MDP on 26 Sept (extending to 29th).The slight increase of the signal (MDP) on 28th appear to be due to the significant flare activity (3 C-class and M-class), suggesting increase in both the instantaneous and background X-ray 340 flux output that usually results to spike in signal amplitude (as depicted in figure 2b).High flare activity can 'overshadow' the signal's response to geomagnetic storms when the event coincide with storm time (Nwankwo et al., 2016).There is also a significant dipping of all the signal metrics (MDP, MBSR, MASS, SRT and SST) on 27 Sept.We note dipping of the MBSR on the days following the main (reference) storms on 18 and 27 Sept. Since the events occurred after dawn (around midday),  November (Dst=-57), presumably driven by the highly variable V sw and P D (Fig. 5d-e).It has been shown that the capability of a given value of the solar wind electric field (SWEF) to create a Dst disturbance or geo-efficiency is enhanced by high solar wind density (Weigel, 2010;Tsurutani et al., 2011).Variation of the AE between 30 Oct. and 3rd Nov. also appear to be consistent with HILDCAAs (Fig. 5h).The DHO-A118 VLF signal level on 25 October around midday also showed a visible reduction following the intense storm condition with Dst up to -132 (Fig. 5a).VLF signal data for GQD-A118 propagation path are not available during 12:00 noon, 25 Oct. to 06:00 pm on 26th October (Fig. 5b).Similar to the first case (Figs. 4 and 5), we note the high flare events on 2nd Nov (up to 7 C-class and M-class), that may have induced a spike in the MDP on the day in both GQD-A118 and DHO-A118 propagation paths.Although dipping of the MDP signal (following storm events) has shown a considerable consistency across the cases presented so far, the MBSR and MASS (in particular) appear to be influenced by storms occurrence time and the high variability or fluctuation of the duskto-dawn ionosphere (and signal) (Nwankwo et al., 2016).However, presenting a consistency across a substantial number of cases is vital to the conclusion of this work.We, therefore, statistically analyze 15 more storm cases between September 2011 and October 2012 in order to obtain the statistical significance of the observations.The 15 storm cases are presented in Table 1, which excluded some cases that were previously analysed in Nwankwo et al. (2016) but also with addition of new cases.
In Figure 7, we show Dst deviations (σ Dst ) and trend in variation of the MDP, MBSR, MASS, SRT and SST signals on the day before the storm (blue bar), the storm day (red bar) and after the storm day (brown bar) for the 15 selected storm cases in (a) GQD-A118 and (b) DHO-A118 propagation paths.σ Dst is the measure or extent of daily fluctuation in measured values of Dst.We recognised the 3 consecutive days as day before event (BE), during event (DE) and after event (AEv).A '0' indicate absence of data.It should be noted however (for this analysis) that these events are separate events, and not continuous events.In GQD-A118 propagation path, about 8 of 12 MDP, 10 of 13 MBSR, 7 of 12 MASS, 3 of 12 SRT and 5 of 12 SST showed dipping features, while 12 of 15 MDP, 9 of 15 MBSR, 10 of 15 MASS, 5 of 15 SRT and 7 of 15 SST showed dipping in DHO-A118 propagation path.These values correspond to 67%, 77%, 58%, 25% and 42% dipping in GQD-A118 propagation path and 80%, 60%, 67%, 33% and 47% dipping in DHO-A118 ptopagation path.The signal levels, along with the percentage dip are presented in Table 2.The MDP signals (in both propagation paths) have generally shown remarkable evidence of dipping following geomagnetic storm conditions.However, we did also observe few cases of increase of the MDP during separate events in both propagation paths (e.g., events 4 and 7 in GQD-A118 and 9 in DHO-A118), as well as increase occurring in both propagation paths during same event (e.g., events 3 and 12).While the probable in amplitude (e.g., significant X-ray output during event 3 and 7 in fig 8c), further investigation into why this characteristic exist will be pursued.In the mean time, we analyze variations in X-ray flux output and geomagnetic indices during events 3 and 12 to better interpret the prevailing ionospheric conditions at the time.In Figure 8 Time (day) background X-ray flux output (as stated earlier), that may have caused increase (or, spike) in the signal level.Thus probably overshadowed geomagnetic effects on the signal.While this explanation may be argued for events 1 (25-27 Sept. 2011) and 400 6 (6-8 Mar.2012), it should be noted that the flare events started well before the storms, and continued until the storms time (in each case), suggesting an established increase in the overall background X-ray before the storms.Hence, it is possible for a storm-induced dipping to manifest under such condition.However, further investigation is encouraged, which is beyond the scope of this work.For event 12 (during 15-17 July 2012), we observed that the peak of the storm (that commenced by midnight on 16th) was on 17th (recognised as AEv).Therefore, any geomagnetic influence on the signal (e.g., dipping) is expected 405 on 17th (or, after) and not 16th, hence the dipping of the AEv signal (on 17 Sept.) instead in DHO-A118 propagation path.
Figure 9 shows Dst deviation (fluctuation) and 2-day mean variations of MDP, MBSR, MASS, SRT and SST signals before, during and after each event for (a) GQD-A118 and (b) DHO-A118 propagation paths.This analysis is important for corroborating the result presented in Figure 7, because the data selection criteria differ from those of Figure 7 in some ways.While BE, in Fig. 7), each acronym (BE, DE or AEv) represent relatively quiescent 2-day mean amplitude before and after DE (but not necessarily in succession to/after DE).However, it should be noted that due to the data averaging (2-day), a 'pronounced' increase or dipping in the signals (comparable to those in the former analysis (fig 7)) are not expected.Another important data selection criterion for this analysis is a relative geomagnetic quiet day BE and AEv with respect to DE.
In GQD-A118 propagation path, 7 of 12 MDP, 7 of 13 MBSR, 7 of 12 MASS, 6 of 12 SRT and 3 of 12 SST showed dipping following the storms, while 10 of 15 MDP, 11 of 15 MBSR, 11 of 15 MASS, 6 of 14 SRT and 6 of 15 SST showed dipping in DHO-A118 propagation path.These values correspond to respective 58%, 54%, 58%, 50% and 25% dipping in GQD-A118 propagation path and 67%, 73%, 73%, 43% and 40% dipping in DHO-A118 propagation path.The signal levels, along with the percentage dip of the signals are presented in Table 3.In general, the trend of variation of the signal metrics considerably reflected the prevailing space weather-coupled effects in the lower ionosphere.The MDP signal appears to be more responsive to geomagnetic perturbations (58% and 67% (67% and 80% for 1-day mean analysis) in respective popragation paths shown in figs 7 and 9) than other signal metrics.However, the 2-day mean analysis showed improvement in MBSR and MASS (73%) the DHO-A118 propagation path, thereby reenforcing the responsiveness of this propagation path to geomagnetic storm impacts.While the mechanism of VLF amplitude (and/or MDP) response to flare-induced SIDs is well understood (see, section 1.3 of (Nwankwo et al., 2016)), the mechanism of the transient response of the signal to geomagnetic storm is not well developed.We speculate that the dipping response of the MDP may be related to (i) positive storm effect, which affects, albeit small, the attenuation of the VLF radio waves (Fagundes et al., 2016) (ii) adjustment of the D layer to storm driven energy input (McPherron, 1979) (iii) precipitation of energetic electrons (Rodger et al., 2010(Rodger et al., , 2012;;Naidu et al., 2020) and/or (iv) charge exchanged between surrounding ionospheric regions (Tatsuta et al., 2015).Since the processes are gradual (when compared to SID scenario) and originating from the magnetosphere, it is not likely that a remote spike occur in the diurnal signal (as observed during flare condition).One other likely reason for the observed dipping characteristic is the modification of the chemistry of the D region by storm pricipitation of SEPs (Turunen et al., 2009), since chemistry (mainly NO) controls the quiescent D-region (Siskind et al., 2017).Nwankwo et al. (2016) noted the existence of pseudo-SRT and SST exhibited by diurnal VLF signal (see, Fig. 2c) as drawback in SRT and SST analysis.This anomaly is due to secondary destructive interference pattern in signals and occurrence of solar flares during sunrise/sunset (Chakrabarti, personal com., 2016).Authors concluded in their study that the post-storm SRT and SST variations do not appear to have a well-defined trend associated with storm effect based on the approach used in the analysis (Nwankwo et al., 2016).In this work, we considered the 'first' SRT and SST values (in the event of a pseudoterminator) during analysis of the signal metrics.A rise in SRT and SST amplitude under geomagnetic storm conditions appear to occur more than otherwise in both propagation paths.We found a respective dipping of 50% and 25% (43% and 40%) of the SRT and SST.Storm-induced disturbances may not have significant influence on these metrics (SRT and SST), and since the sunrise and sunset signatures relates to mode conversion in the VLF propagation path it might imply that the D-region density is not a significant contributor to this effect.It is important to note that of the two propagation paths used in this study, the DHO-A118 signal appears to be more sensitive to geomagnetic storm-induced magnetosphere-ionospheric dynamics.We do not expect a 'perfect' consistency in signal trend and variations across all cases, because the individual effects of solar and other forcing mechanisms (including those of lithospheric and atmospheric sources) on the ionosphere are difficult to estimate (Kutiev, 2013;Nwankwo et al., 2016).This scenario can also cause non-linear coupling processes and consequent significant fluctuations in radio signals.3.2 Investigating the state of the ionosphere over the propagation paths of the VLF signals Here, we study the state of the ionosphere over the two VLF propagation paths using the virtual heights (h E, h F1 and h F2) and critical frequencies (f oE, f oF1, and f oF2) of the E and F regions obtained from two ionosonde stations near the GQD and DHO transmitters.Although we made effort to obtain data from stations near each transmitter/receiver and at the mid-point, we found no ionosonde station at the mid-point, and the nearest station to the receiver (Tortosa) has no data for the period/intervals under study.However, to make up for this dearth of data, we will complement the analysis with the results in the extended study that utilised the GNSS data in the region (e.g., (Nwankwo et al., 2021)).Details of the ionosonde stations used in this study are provided in Table 4.We treat Chilton station as nearest to GQD transmitter and Juliusruh station nearest to DHO transmitter.Tortosa station is closest to the A118 transmitter but has no data for the intervals.We obtained and calculated the daytime (8:00 am -3:00 pm) mean values and standard deviations (σ) of the parameters, and analyzed for the storms of interest (on 17 and 26 September and 1 November) within the intervals 16-19, 25-28 September and 29 October to 2 November 2011.
We exclude analysis of the 25 October storm because the data for this interval are inadequate.
Figure 10 shows the daily mean and standard deviation (SD or σ) of foF2, foF1, foEs, foE, h'F2, h'F, h'Es and h'E during 16-17 September 2011 for Chilton and Juliusruh Stations.We compare the pre-storm day (blue broken line) values with the storm day (red broken line) values.At Chilton station (near the GQD transmitter) result show significant increase and/or fluctuation (increase in SD) of the foF2, and a decrease (with significant fluctuation) of foF1 on the storm day, 17 September.
The height of the E and F regions (h'F2, h'F, h'Es and h'E) significantly increased following the storm.Similar pattern of variations were observed at Juliusruh station.The foF2 and foF1 increased (and/or fluctuated) significantly, as well as h'F2, increased on the storm day, 26 September.Near the DHO transmitter (Juliusruh station) there is an anti-correlated variation in the critical frequencies of the E and F regions; a depression of the foF2 and foF1, but increase in foEs and foE (when compared with the scenario at Chilton station).The height of the F2, F and E regions increased by 47.89 km, 16.08 km and 9.14 km, respectively (which are so far the largest increase of the parameters), while the height of the Es region decreased by 0.16 km (see, Table 5).
Figure 12 show the daytime mean variation and SD of foF2, foF1, foEs, foE, h'F2, h'F, h'Es and h'E during 29 October -02 November 2011 for Chilton and Juliusruh Stations.This interval is of interest because of the fluctuation in geophysical parameters during the days preceding the storm.We include this interval to investigate the couple effect of this extended period of (30-31 Oct.) of geomagnetic disturbances preceding the storm on 1 November.It appears that energy began building up in the magnetosphere-ionosphere system after the first significant spike in V sw around 7:00 am on 29th (and subsequent increase on 30 Oct.) and PD around 10:00 am on 30 October until around 10:00 am on 1 November when the storm was triggered following sudden increased in V sw and southward turning of the B z (Nwankwo et al. 2021).Here, we compare the parameters' level on the relatively quiet day (29 Oct.) with those of the storm day (on 1 Nov.), since the two days preceding the storm were signicantly disturbed.The result show significant increase of foF2 and foF1 at Chilton station.Like the 17 September storm scenario, values of foEs decreased, while the foE remained unaffected for this station.The h'F2 decreased, while the h'F, h'Es and h'E showed significant increase.At Juliusruh station only the critical frequency of the F2 region increase, while those of the F1, Es and E decreased.However, the increase and/or fluctuation of the parameters were significant (in most cases) during the disturbed days (30 and 31 Oct.) preceding the storm, suggesting responses of the E and F ionosphere regions (coupled to the D region) before the storm commencement (as a result of increased geomagnetic activity on the days).We present summary     of the storms day variations in h'F2, h'F, h'Es and h'E for the two stations in Table 5.

500
In summary, foF2, foF1, h'F2, h'F, h'Es and h'E generally showed significant increase and/or fluctuation near both transmitters (GQD and DHO) during the geomagnetic storms, whereas foEs and foE either increased (slightly) or unaffected.It appears that the observed storm-induced increase and fluctuation are largely sustained or further enhanced on the day (or days) following the event (post storm day), suggesting a continuous driving of the ionosphere by the storms and/or a substorm effect.Although the analysis for 1 November storm scenario showed weak correlation, variations of the parameters reflected 505 the coupled responses of the ionosphere to energy build-up ahead of storm commencement.Nwankwo and Chakrabarti (2018) reported significant depression and fluctuations of foF2 following significant geomagnetic disturbances and/or storms in highand mid-latitude, and distortion in the quasi-periodic pattern of the parameter.Their inference was, however, based on a pre-    liminary analysis from the result of single ionosode station.From this comparatively detailed analysis, it is clear that the reported depression of foF2 may occur during some (isolated) storms and locations, and should, therefore, not be treated as 510 global response.Negative storm effects (in which foF2 assumes a negative values) has also been reported (e.g., (Blanch et al., 2013;Kane, 2005)).In this analysis, the largest increase of the h'F2, h'F, h'Es and h'E occured in Juluisruh, near the DHO transmitter (see Table 5).The ionosonde observation show that fluctuations in the reference heights appear to be the dominant responses of the E and F regions to geomagnetic storms, whereas dipping of the VLF radio waves reflects storm-induced dynamics in the ionospheric D region.This observation is instructive in that the observed large ionosode 515 increase and the large amplitude decreases in the DHO-A118 propagation path signal may be related to coupled effects between the ionospheric regions, but also suggestive of strong storm responses (more intense) around/near the DHO receiver or DHO-A118 propagation path.This result is in agreement with the recent findings reported in Nwankwo et al. (2021)   feature is in agreement with the findings of (Choudhury et al., 2015), who reported that the receiver position electron density is the main factor influencing VLF signal at ionospheric sunrise time during long-duration geomagnetic storms.
It is also worth noting that the ancillary information of the timing, classification and location of associated solar flares, CMEs, SPEs, and the timings for the SSCs showed that the strong storm intervals during which large dipping or decrease in DHO signal level occured were associated with SPEs (see, Table 5 in (Nwankwo et al., 2021)).The results of this effort that combined the diagnostics of the D, E and F regions (to probe geo-storm effects in the lower ionosphere) demonstrates that despite the tenuousness of the coupling between the dayside upper and lower ionospheric regions   into five metrics (i.e MBSR, MDP, MASS, SRT and SST), and monitored the trend in variations of the signal metrics for up to 20 storms between September 2011 and October 2012.The goal of the analysis is to understand deviations in the signal that 545 are attributable to the storms.Up to five (5) storms and their effects on the signals were studied in detail, followed by statistical analysis of 15 other cases.Our results showed that the MDP exhibited characteristic dipping in about 67% and 80% of the cases in GQD-A118 and DHO-A118 propagation paths, respectively.The MBSR showed respective dipping of about 77% and 60%, while the MASS dipped by 58% and 67%.Conversely, the SRT and SST showed respective dipping of 25% and 33%, and 42% and 47%, favouring rise of the signals following storms.The MDP consistently showed strong responses to the 550 storms that ohter metrics (followed by the MBSR and the MASS).Among other possible reasons outlined in this paper, we speculate that the responses are related to positive storm effects, the attenuation of the VLF radio waves.Of the two propagation paths examined in this study, we observed stronger dipping of the VLF amplitude of DHO-A118 propagation path during the storms.To understand the state of the ionosphere over the propagation paths and how it affects the VLF responses, we further analysed virtual heights (h E, h F1 and h F2) and critical frequencies (f oE, f oF1, and f oF2) of the E and F regions 555 (from ionosonde stations near the GQD and DHO transmitters).The results of this analysis showed a significant increase and/or fluctuation in the height of the E and F regions (h'F2, h'F, h'Es and h'E) near both transmitters during the geomagnetic storms, with the largest increase occurring in Juluisruh (Germany) station, near the DHO transmitter.This scenario suggest a strong storm responses over the region, possibly leading to the large dipping of VLF amplitude for DHO-A118 propagation path.
The ionosonde observation show that fluctuations in the reference heights appear to be the dominant responses of the E and F regions to geomagnetic storms, whereas dipping of the VLF radio waves reflects storm-induced dynamics in the ionospheric D region.Our findings demonstrates that ionospheric E and F regions plays significant role in driving the storm-induced dynamics of the D region and the associated observed responses of VLF radio waves (within context of solar-terrestrial coupling arena) despite the tenuousness of the coupling between the dayside upper and lower ionospheric regions.

Figure 4
Figure 4 shows daily mean fluctuation of Dst and AE, and variations in the VLF midday signal amplitude peak (MDP), mean signal amplitude before local sunrise (MBSR), mean signal amplitude after sunset (MASS), sunrise terminator (SRT) 335

Figure 5
Figure 5 shows diurnal VLF amplitude for (a) DHO-A118 and (b) GQD-A118 propagation paths, daily variation in (c) X-ray flux output (d) V sw (e) P D (f) Dst (g) A p and (h) AE indices during 22 October -5 November 2011.This period is associated with a severe storm with main phase on 25th October (Dst=-132) and consecutive storms on 1 November (Dst=-71) and 2nd

Figure 6
Figure 6 shows daily deviations of Dst and AE, and variations in the MDP, MBSR, MASS, SRT and SST for (a) DHO-A118 and (b) GQD-A118 propagation paths during 22 October -5 November 2011.Although data for GQD-A118 propagation path during 25 and 26 October is inadequate for the present analysis, we did observe a dipping of the MBSR on the main storm day (25 Oct.).Dipping of the MDP, MASS and SST occurred on 1 Nov., and those of MBSR, MASS, and SRT on 2 Nov., following the consecutive storms.In DHO-A118 propagation path, we observed dipping of the MDP, MBSR, MASS, and SRT on 25 Oct., dipping of the MDP, MBSR, MASS, and SST on 1st Nov., and dipping of the MBSR and SRT on 2 Nov.

Figure 7 .
Figure 7. Dst deviation (or fluctuation), and variations in MDP, MBSR, MASS, SRT and SST signals 1-day before, during and after each of the 15 events for GQD-A118 and DHO-A118 propagation paths.Note that the each Dst bar represent the deviation (σ) corresponding to the VLF amplitude before, during and after the selected events (storm) as listed in Table 1 and shown in figure 8.The events are selected (for the intervals shown in fig 8) and not continous

Figure 8 .
Figure 8.Diurnal VLF amplitude for (a) DHO-A118 and (b) GQD-A118 propagation paths, daily variation in (c) X-ray flux output (d) Vsw, (e) P D and (f) Dst indices for a day before and after each of the 15 storms

Figure 9 .
Figure 9. Dst deviation and 2-day mean variations of MDP, MBSR, MASS, SRT and SST signals before, during and after each event for GQD-A118 and DHO-A118 propagation paths.

Figure 11 .
Figure 11.Daytime mean variation and SD of foF2, foF1, foEs, foE, h'F2, h'F, h'Es and h'E during 25-28 September 2011 for Chilton and Juliusruh Stations One can clearly observed (from fig 13) the strong dipping (or reduction) of the daytime VLF amplitude and the simultaneous increase in VTEC values on the storm days in DHO-A118 propagation path.We note the large increase of VTEC values for ESCO station located near the Receiver, especially during 17 September and 25 October storms.This the adjoining regions of E and F plays significant role in driving the storm-induced dynamics of the D region and the associated observed responses of VLF radio waves within context of solar-terrestrial coupling arena.4ConclusionsIn this work, we performed a diagnostic study of geomagnetic storm-induced disturbances that are coupled to the lower ionosphere in mid-latitude D-region using propagation characteristics of VLF radio signals.We characterised the diurnal signal

Table 1 .
Summary of analysed 15 geomagnetic storm events reason for the observed scenario is suggestive of factors such as propagation characteristics and/or X-ray flux induced spike 390

Table 2 .
Summary of trend in dipping of the signals' metrics during 15 geomagnetic storm case in (a) DHO-A118 and GQD-A118 propaga-

Table 4 .
Ionosode stations near the VLF transmitters, receiver and/or propagation paths Es and h'E.Values of foEs decreased, while the foE remained unaffected in both stations.Also, there appear to be a sustained post-storm increase and/or fluctuations of the parameters on on 18 September, suggesting a continuous driving of the ionosphere by the storm.Figure11show the daytime mean variations and SD of foF2, foF1, foEs, foE, h'F2, h'F, h'Es and h'E during 25-28 September 2011 for Chilton and Juliusruh Stations.The storm during this interval (on 26 September) is well developed (with Dst up to -101 nT) and larger than the 17 September event.The result of this analysis show a slight increase of foF2 and foF1 but a decrease in foEs and foE for Chilton station.The height of the F2 (h'F2) decreased (by 6.90 km) while those of the F, Es and E Figure 10.Daytime mean variations and standard deviation (SD) of foF2, foF1, foEs, foE, h'F2, h'F, h'Es and h'E during 16-17 September 2011 for Chilton and Juliusruh Stations.The blue broken line represent the pre-storm day values, while the red broken line represent the storm day values of the ionospheric parameters.
. Their study combined observed VLF amplitude variations with TEC/VTEC data obtained from multiple GNSS stations including Euskirchen in Germany (EUSK), Hailsham in UK (HERT), Paris in France (OPMT) and Naut Aran in Spain (ESCO), to investigate ionospheric response to storms over some signal propagation paths during the same events.They showed and reported simultaneous increase of VLF amplitude and enhancement of electron density profiles near the DHO transmitter.In figure13we show the daytime variation in VLF amplitude (red line plot) for DHO-A118 propagation path, together with VTEC values obtained from HERT (black line), EUSK (blue line), OPMT (green line) and ESCO (brown line) stations across some locations in Europe during 16-19 and 25-28 September, 24-27 October and 29 October-1 November 2011.HERT is closest to the GQD transmitter (about 508.12 km), EUSK is closest to the DHO transmitter (about 279.99 km), while ESCO is the nearest to the Receiver (about 90.47 km).

Table 5 .
Observed increase (or decrease) of the h'F2, h'F, h'Es and h'E during the storms on 17 and 25 September and 1 November 2011