The Role of Tectonic Plate Thickness and Mantle Conductance in Determining Regional Vulnerability to Extreme Space Weather Events: Possible Enhancement of Magnetic Source Fields by Secondary Induction in the Asthenosphere

During magnetic storms, solar‐magnetosphere‐ionosphere‐Earth interactions give rise to geomagnetically induced currents (GICs) in man‐made technological conductors such as power grids, gas pipelines, and railway networks with potentially damaging outcomes. Generally, electrically conductive regions of the Earth are assumed to be less at risk to GICs than resistive ones, since induced electric fields associated with GICs are linearly related to given magnetic source fields via Earth's impedance. Here, we show that magnetic source fields associated with storms can be enhanced by secondary electromagnetic (EM) induction in Earth's electrically conductive asthenosphere and that this previously neglected effect can give rise to larger electric fields close to the lithosphere‐asthenosphere boundary in regions where the conductance of the asthenosphere is higher. Our analysis of data from the 30 October 2003 “Halloween” and 8 September 2017 storms shows that the magnitudes of electric fields from both storms are affected by lithospheric plate thickness and asthenosphere conductance (conductivity‐thickness product) and that they are 5 times larger in southern Sweden (>5 V/km for the 30 October 2003 “Halloween” storm) than in central Scotland. Our results provide insight into why Sweden experienced a storm‐related power outage in 2003, whereas Scotland did not.

distribution networks (e.g., Bolduc, 2002;Cannon et al., 2013;Radasky, 2011;Wik et al., 2009). Whereas geomagnetic field variations have been continuously recorded at observatories around the world for over a century, there is a paucity of measurements of electric fields induced during magnetic storms in many regions of the industrialized world. Therefore, our direct knowledge of these enhanced fields, their interaction with Earth's conductivity structure, and our ability to model the hazard they pose is limited.
Magnetotellurics (MT) is a passive electromagnetic technique that harnesses natural fluctuations in the electric and magnetic fields induced in Earth due to solar-magnetosphere-ionosphere interactions to study the electrical conductivity of Earth's crust and mantle (e.g., Simpson & Bahr, 2005). By sampling these electric and magnetic fields simultaneously at a fixed sample rate, a quasi-continuous time series is recorded, which can be decomposed into its constituent electric and magnetic spectra at discrete frequencies via Fourier transform. Cross correlation of these spectra for each frequency then allows computation of a frequencyand direction-dependent impedance tensor, (Z, Cagniard, 1953;Simpson & Bahr, 2005): where E is electric field, H = B/μ 0 is magnetic field, subscripts x and y denote north-south and east-west orientated fields, respectively, and μ 0 is the permeability of free space. For stable tectonic environments these impedances can be considered stationary (e.g., Hanekop & Simpson, 2006). Impedances can also be obtained from electrical conductivity models. With perfectly sampled data that are well fitted by a model, measured and modeled impedances should converge. However, perfectly sampled data are rarely available. In the real world there are always differences between modeled and measured impedances. Bahr (2020a, 2020b) point out that for space weather applications the conductivity structure under an MT site is better described by a measured impedance tensor than by any conductivity model that was derived from such measured impedances, where some form of error propagation and data misfit criteria are applied.
Equation 1 suggests a way of approximating the electric fields at a site where only the geomagnetic fields have been recorded during a magnetic storm if an MT impedance tensor is available for that site (Kelbert et al., 2017;Simpson & Bahr, 2020a). Equation 1 is only strictly valid for the case of plane wave (i.e., uniform) exciting fields far above Earth's surface. The primary driver of the main phase of magnetic storms is a ring current lying in the equatorial plane, 3-8 Earth radii above Earth's surface (Daglis et al., 1999), and this can be treated as a quasi-uniform source. However, departures from the plane wave assumption occur at high latitudes during storms due to ionospheric currents associated with the auroral electrojet located 100-150 km above Earth's surface. For ring current models of magnetic storms, we expect the B x and E y components to evince the largest variations. Time derivatives of magnetic time series recorded during the 30 October 2003 "Halloween" magnetic storm in Sweden are strongly north-south polarized, supporting the assumption of approximately east-west oriented induced electric fields for this storm (Watermann & Gleisner, 2009). During times of strong westward electrojets, large GICs are common in Scandinavia (Viljanen et al., 2001) and the 30 October 2003 "Halloween" magnetic storm caused a blackout in Malmö, Sweden. Wik et al. (2009) have calculated that GICs of up to 330 A may have flowed in the Swedish power network a few minutes before the Malmö blackout, while GICs of 42 A were recorded in the power grid at Strathaven, Scotland (Thomson et al., 2005) during the same event. These point estimates depend on the transmission line voltages associated with induced electric fields encompassing the entire network.
Electromagnetic fields are dispersive. Therefore, the magnitudes of GIC-driving electric fields witnessed at the surface of the Earth depend not only on characteristics of the external storm time magnetic field but also on the electrical conductivity structure deep within the Earth, where component frequencies of these surface electric fields are induced. In August 2017, we deployed seven MT instruments in Scotland to (i) monitor electric and magnetic fields and (ii) investigate regional conductivity structure. These instruments registered the 8 September 2017 magnetic storm (Simpson & Bahr, 2020a, 2020b, 2020c, the largest magnetic storm since the 30 October 2003 "Halloween" storm. Here, we compare electromagnetic data from the 8 September 2017 and 30 October "Halloween" magnetic storms in Scotland and Sweden, differences in the region-dependent thickness of the Earth's quasi-rigid outer, electrically resistive shell known as the "tectonic plate" or "lithosphere," and the region-dependent conductance of the underlying electrically conductive layer known as the "asthenosphere" to demonstrate the importance of coupling between electromagnetic fields external and internal to the Earth in generating GICs.

Internal and External Parts of the Geomagnetic Field
The horizontal magnetic field B h and internally induced magnetic field B i are related to the external source field B e via conductance, τ, according to (cf. A2.19 in Simpson & Bahr, 2005): where τ 0 is the mantle conductance for the extreme case in which the internal part of the magnetic field is as large as the external source field. Hence, for τ = τ 0 , electromagnetic induction in the Earth increases the horizontal magnetic field observed at the surface by a factor of 2 relative to the external source field, whereas τ > τ 0 leads to horizontal magnetic fields that are more than twice the external magnetic source field and vice versa for τ < τ 0 . Therefore, for regional studies of the hazard associated with geomagnetic storms, convolution of the geomagnetic fields from distant observatories with local impedance tensors may not provide correct estimates of electric field amplitudes across a region.

Frequency Dependence of Inductance Scale Length
MT impedances (Equation 1) are directly related to subterranean electrical conductivities. The tensor in Equation 1 is frequency dependent. Therefore, it contains information about the depth to which electromagnetic fields penetrate. This penetration depth can be expressed in terms of the complex inductance scale length tensor, C, which for the simplistic case of a homogeneous half-space is defined as (Schmucker, 1973;Weidelt, 1972): where p ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi 2=μ 0 σω p is the electromagnetic skin depth, μ 0 is magnetic permeability of free space, σ is the conductivity of the medium penetrated, and ω is angular frequency. A qualitative interpretation of Equation 3 is that penetration depth increases as electromagnetic wave period increases (i.e., as frequency decreases) and that the depths imaged by different sounding periods will depend on the conductivity structure, with electromagnetic waves of a given period being less attenuated and therefore reaching greater depths in resistive environments than in more conductive ones. Hence, the Earth acts as a filter that modifies the electromagnetic fields induced within it by external magnetovariational fields.

Dependence of Electric Fields on Conductance and Inductance Scale Length
For a three-layer medium comprised of a resistive lithosphere of thickness h, a conductive asthenosphere of conductance τ such that ωμ 0 τp ≫ 1 and an underlying moderately conductive (100 Ωm) upper mantle layer in which electromagnetic fields with angular frequency ω have a penetration depth p, the inductance scale length given in Equation 3 is modified (Bahr et al., 2000): where j = ωμ 0 τ. Therefore, from Equations 1-4, the horizontal electric field, E h , is given by: And the modulus of E h is given by:

10.1029/2020SW002587
Space Weather This expression for the magnitude of the horizontal electric field contains competing terms with regard to the effect of conductance, with larger conductance promoting larger magnetic fields that increase |E h | on the one hand and smaller inductance scale lengths that decrease |E h | on the other. For a given lithospheric thickness, for all reasonable values of τ 0 , the τ term in the second bracket has a larger effect on |E h | than the τ term in the first bracket. For the case of plane wave electromagnetic source fields, Equation 1 embodies the principle that as the thickness of the resistive lithosphere increases, the electric fields induced in it also increase as well as the principle that as the conductance of the asthenosphere increases, induced electric fields with sufficiently long periods to penetrate into the asthenosphere are attenuated. However, storm time magnetic source fields are not plane wave. In this case, Equation 6 suggests that there is a period range corresponding to penetration depths immediately below the lithosphere-asthenosphere boundary (LAB) in which enhancement of magnetic fields within the conductive asthenosphere could cause electric fields to be increased rather than decreased.

Estimation of Asthenosphere Conductance
For a three-layer medium comprised of a resistive lithosphere of thickness h, a conductive asthenosphere of conductance τ, and an underlying moderately conductive (100 Ωm) upper mantle layer in which electromagnetic fields with angular frequency ω have a penetration depth p, Wait's recursion formula (Wait, 1954) for the Schmucker-Weidelt transfer function, C(ω) (Schmucker, 1973;Weidelt, 1972), reduces to (cf. Simpson & Bahr, 2005, Equation 2.44): It has further been shown by Bahr et al. (2000) that the real (Re) and imaginary (Im) parts of C(ω) can be expressed as polynomials in ωμ 0 τp and that for large asthenosphere conductance, τ, such that ωμ 0 τp ≫ 1 and long periods such that Re(C) > h, the MT phase, ϕ, tends to 90°and is related to τ via:

Comparison of Measured Magnetic Fields in Scotland and Sweden During Magnetic Storms
For the 8 September 2017 storm, the amplitude of B x , which induces orthogonally oriented (E y ) electric fields, is 1.75 times as large at UPS geomagnetic observatory, Sweden, compared to RAN, an MT site in central Scotland (Figure 1a). A similar enhancement occurs for the 30 October 2003 "Halloween" storm, comparing data from UPS (Sweden) and ESK (Scotland) geomagnetic observatories ( Figure 1b). As the latitudes of these two sites differ by only 3°, we would not expect such large differences in magnetic field amplitudes to be generated by the external geometry of the storm source field alone. Further arguments against the observed differences being attributable to source field geometry and heterogeneity are provided in section 3.3.

Estimated Electric Fields for the 8 September 2017 and 30 October 2003 "Halloween" Magnetic Storms in Scotland and Sweden
A good agreement between the measured electric fields and those calculated by multiplying measured magnetic field spectra with frequency-dependent impedance tensors computed from time series obtained before and after the 8 September 2017 storm has been demonstrated previously (Simpson & Bahr, 2020a). This agreement between measured and estimated electric fields-typified by site RAN, Scotland ( Figure 2)-justifies more general application of this technique for nowcasting electric fields to investigate storm-related hazards.

Space Weather
SIMPSON AND BAHR Figure 2 also shows electric fields computed using geomagnetic observatory data from UPS, Sweden, for the same 7-8 September 2017 time interval and a local impedance tensor obtained from the BEAR array (Bahr & Simpson, 2002). E y fields are 5 times bigger for the Swedish site compared to the Scottish site, reaching peakto-peak amplitudes of 1.25 V/km at UPS compared to 0.25 V/km at RAN. We similarly computed electric fields for the 30 October 2003 "Halloween" storm near UPS and ESK observatories (Figure 3). At UPS, peak-to-peak electric field variations of 5.6 V/km may have occurred, whereas at ESK, induced voltages were likely less than those witnessed at UPS during the smaller 8 September 2017 storm ( Figure 2). Hence, the region around UPS, Sweden, appears to be more vulnerable to extreme space weather events than Scotland. This is consistent with the observation that large GICs are common in Scandinavia (Viljanen et al., 2001).

Source Field Geometry
Whereas for the estimation of the electrical impedance from MT data, a plane wave source (with zero wavenumber, k) is assumed, a magnetic storm is not a homogenous source and therefore the diffusion equation that describes the induction process applicable to MT is replaced by: The source heterogeneity is best understood for the solar quiet magnetic daily variation (Sq) generated by current vortices within the ionosphere, with a wavenumber for the m'th Sq harmonic (Bahr & Filloux, 1989): where θ is the colatitude and R is the Earth's radius. For the first and second harmonics, the corresponding wavelength λ = 2π/k = 9,000 km,  which is roughly the diameter of the Sq vortex. This wavelength is greater than a factor 10 larger than the penetration depth at the associated frequency but is nevertheless sufficiently small to justify the need for a description of the source field geometry for the case of Sq variations. In contrast, a geomagnetic storm modeled with a ring current having a radius of between 3 and 8 Earth radii (Daglis et al., 1999) is described by wavelengths significantly larger than the penetration depths reached in MT studies. Indeed, Osipova et al. (1989) have shown that for Scandinavia, electromagnetic source field heterogeneities only play a more significant role than conductivity structure at periods longer than 10,000 s. Therefore, for magnetic storms (in contrast to Sq variations), the source geometry should not play a major role at the MT sounding periods considered in our study. A caveat to this is that we cannot entirely exclude the possibility that substorms associated with much smaller current systems in the ionosphere occur during any particular geomagnetic storm. However, for the 8 September 2017 storm, source field heterogeneities associated with substorms cannot be significant for two reasons: (i) Other sites running simultaneously with RAN during the 8 September 2017 storm >100 km away show very similar magnetovariations (Simpson & Bahr, 2020a), and the remaining differences are partly due to horizontal gradients in mantle conductivity (Simpson & Bahr, 2020b) as indicated by small but nonvanishing magnetic perturbation tensors, W (Schmucker, 1970;Simpson & Bahr, 2005). (ii) Our comparison between the electric fields measured during the 8 September storm and those predicted by use of the MT impedances ( Figure 2) offers a way of quantifying the errors introduced by any source field heterogeneities: Since the effect of the 3-D conductivity distribution on the electric field at a site is completely described by the frequency-dependent impedance tensors and since these impedance tensors are derived assuming a plane wave source, any discrepancies between the measured and estimated electric fields at a particular site can only be attributable to departures from the plane wave source ensuing from either the magnetic storm or other sources of noise that are filtered out during MT data processing (Egbert & Booker, 1986;Simpson & Bahr, 2005). Figure 2 demonstrates, however, that our technique of estimating the electric fields during the storm using MT impedances generates field amplitudes similar to those measured during the storm; that is, source field heterogeneities are negligible and the plane wave assumption appears approximately valid. Scaling between observed and estimated electric fields for the 8 September 2017 storm has been rigorously investigated by Simpson and Bahr (2020a), who showed that discrepancies are likely to be attributable to source field polarization.

Mantle Conductance
Amplitude spectra for the 8 September 2017 (Figures 4a and 4b) and 30 October 2003 "Halloween" storms (Figures 4c and 4d) are largest for frequencies <0.001 Hz or equivalently periods longer than 1,000 s. These long-period signals have penetration depths (see Equation 3, Methods, section 2.2) deeper than the electrical LAB (Simpson, 2013) under Sweden (180 km; Bahr & Simpson, 2002) and Scotland (80 km). Cherevatova et al. (2014) have estimated from MT data with periods <1,000 s that the electrical lithosphere under the mountainous region of southern Norway is thicker than under Sweden, as one would expect from isostatic principles. The premise that deeply penetrating electromagnetic fields have a major influence on GICs measured at Earth's surface is further supported by large amplitude GICs in the frequency band 10 −2 to 10 −4 Hz corresponding to penetration depths from the lower crust to the asthenosphere measured directly in the electric power grid in Guangdong, China (Liu et al., 2008).
To date, studies of storm-related surface electric fields for the UK (Beamish et al., 2002;Ivannikova et al., 2018) have been limited to thin sheet conductance models that only take into account the galvanic effects of ocean bathymetry and near-surface (uppermost 5 km) conductance contrasts inferred from lithology, which is known to be an unreliable indicator of conductivity. These models do not consider first-order

10.1029/2020SW002587
Space Weather effects of induction in the mantle (e.g., Simpson, 2002). Rosenqvist and Hall (2019) have presented a pseudo-3-D conductance model for the upper crustal electrical conductivity structure of Sweden obtained by amalgamating and extrapolating 1-D and 2-D models (Ádám et al., 2012;Engels et al., 2002;Korja et al., 2002;Rasmussen et al., 1987). They excited their model with a spatially uniform magnetic source field and used it to infer a first-order influence of the geomagnetic coast effect (Lines & Jones, 1973;Parkinson & Jones, 1979) on the induced electric fields that drive GICs. However, the long-period electromagnetic response of surficial conductance sheets is strongly influenced by the choice of underlying electrical resistivity (Simpson, 2002) and sensitive to model discretization parameters . Below the crust, Rosenqvist and Hall's (2019) model contains a 1,000 Ωm half-space extending to infinity. However, MT studies (e.g., Bahr & Simpson, 2002;Simpson, 2002) have demonstrated that the electrical resistivity of the continental mantle below Europe and Fennoscandia is significantly less than 1,000 Ωm. The assumption of an unrealistically resistive mantle exaggerates both the coast effect and the magnitudes of the induced electric fields. Both Sweden and Scotland are surrounded by shallow, marginal seas that lie mainly on the European continental shelf rather than by deep oceans. Simpson and Bahr (2020a) have used 3-D modeling (Siripunvaraporn et al., 2005) to demonstrate that the coast effect is minimal for sites in central Scotland, whereas the electrical resistivity of the underlying mantle exerts a first-order influence on the induced electric fields. This point can be further understood when we consider that MT resolves conductanceconductivity of a layer multiplied by thickness of a layer. MT studies (e.g., Haak & Hutton, 1986;Kelbert et al., 2012;Sakkas et al., 2002;Simpson, 1999Simpson, , 2000Simpson et al., 1997) have revealed variations in the electrical conductivity of the continental crust and mantle spanning 8 orders of magnitude that in some cases affect electromagnetic transfer functions more strongly than the coast effect. A 1 Ωm, lower-crustal conductor extending throughout a 15 km thick lower crust gives a conductance contrast of 15,000 S compared to a resistive (1,000 Ωm) lower-crustal layer. This is equivalent to the conductance of a 3,750 m deep ocean. A 10 Ωm sublithospheric upper mantle extending from a depth of 180 km to the midmantle transition zone at 410 km (where the electrical conductivity again increases (e.g., Bahr et al., 1993)) gives a conductance contrast of approximately 23,000 S, which will have a significantly greater effect on the long-period electromagnetic response than the shallow seas surrounding the UK and Sweden. We further note that long-period (>1,000 s; Figure 4) induced electric fields would have penetration depths extending into the lower mantle according to Rosenqvist and Hall's (2019) model, and this is known to be incorrect (Simpson, 2002;Simpson & Bahr, 2005).
We employ a generic Earth model with two subsurface parameters that affect MT impedances through their effects on electromagnetic fields-lithosphere thickness and asthenosphere conductance-to explain how regional differences to space weather vulnerability occur. We note that tectonic boundaries also affect induced electric fields that drive GICs in technological conductors (e.g., Dong et al., 2015). Therefore, we do not advocate using this model for estimating electric fields but only for elucidating the processes involved. We have already demonstrated (Figure 2) that electric fields can be satisfactorily estimated by multiplying MT impedances with measured magnetic field spectra in the frequency domain. The information content of the impedances that are estimated from direct measurements of electric and magnetic fields is always greater than that of any secondary model-even one with a greater complexity than our generic modelor than impedances generated from such a model. While 3-D conductivity models are important in the context of investigating tectonic structure (e.g., Simpson, 2000;Simpson & Warner, 1998), they become obsolete for the purpose of estimating storm time electric fields where MT impedances-to which all information about the effect of Earth's heterogeneous conductivity structure on electric fields are intrinsic-are available for "Models can be confirmed by the demonstration of agreement between observation and prediction, but confirmation is inherently partial" (Oreskes et al., 1994).
The asthenosphere below Scandinavia has been inferred to be electrically anisotropic (Bahr & Simpson, 2002). Therefore, three variables are required to describe its conductance (e.g., Simpson, 2013)the magnitude and strike direction of the higher conductance and the anisotropy ratio. However, these information are entirely contained within the MT impedances used to estimate the electric fields.
At sites in Finland, the asthenosphere is electrically anisotropic with conductances of 32,000 S in the SSW-NNE strike direction (Bahr, 1988) compared to 7,000 S in the cross-strike direction (Bahr & Simpson, 2002). We repeated the conductance estimation technique (Bahr & Simpson, 2002;Bahr et al., 2000;Methods, section 2.4) and strike determination (Bahr, 1988) for three MT Sites B09, B11, and B13 (Bahr & Simpson, 2002) close to UPS and for RAN, ESK, and SEB, MT sites in central Scotland (Simpson & Bahr, 2020b). Our analysis reveals that the along-strike (defined as the direction of high conductivity in an anisotropic or 2-D environment) conductance of the asthenosphere under central Sweden-13,000 S-is 3-4 times higher than under central Scotland-4,000 S striking NW-SE at ESK to approximately E-W at RAN, while both regions have similar but less well-constrained, crossstrike conductances of approximately 200 S. These strike directions are not consistent with the alignment of the coastlines, indicating that Earth's deep electrical conductivity structure rather than the coast effect dominates the long-period response functions. From Equation 1, larger electric fields correspond to higher impedances (i.e., more resistive environments) and larger magnetic fields. From Equation 1, we might therefore expect the electric fields to be smaller for sites in Sweden compared to those in Scotland. However, this is only the case for E x -polarized fields (Figures 2 and 3). Examining the time series and spectra shown in Figures 1-3, we find that the B x , B y , and E y fields for Sweden are larger than those for Scotland. We will next demonstrate that amplification of the magnetic fields occurs due to electromagnetic induction in Earth's asthenosphere.

Coupling Between Magnetic Source Fields and Mantle Conductivity
As well as a dominant external part, the time-varying component of Earth's magnetic field has a significant internal part attributable to electromagnetic induction in the Earth (Schuster, 1889). The Gauss separation (Backus et al., 1996;Gauss, 1839) enables geomagnetic fields to be decomposed into components that arise due to processes internal and external to the Earth, respectively. This decomposition is facilitated by the fact that whereas internal and external parts of B x and B y have the same sign, internal and external parts of B z have different signs (Simpson & Bahr, 2005). Therefore, the measured B z , which is the sum of external and internal vertical fields, is generally smaller than the measured horizontal components.
For plane waves impinging on a 1-D Earth, B z would be zero, but magnetic storms do not have a plane wave source field geometry. During the 2003 and 2017 storms, B z is significantly larger at UPS than at ESK (Figure 1). We, therefore, infer that the high mantle conductance under Sweden gives rise to a larger internal part of the magnetic field at UPS compared to ESK and RAN (see also section 2.1, Methods). For horizontal magnetic fields at UPS, this amplification effect is significantly stronger in B x than B y due to the SSW-NNE orientation of the maximum conductance. According to Equation 2 (section 2.1, Methods), these inductive effects, which are also manifest in the regional variability of horizontal magnetic transfer functions for Sweden (Bahr & Simpson, 2002), lead to the approximate doubling of the magnitude of the B x component for Sweden compared to Scotland seen in Figure 1 for both the 8 September 2017 and 30 October 2003 "Halloween" magnetic storms.

Effect of LAB Depth and Conductivity on Vulnerability to Storms
We substituted values for lithospheric thickness and asthenospheric conductance for Sweden and Scotland, into an expression for electric field magnitude that includes amplification of magnetic fields by internal induction (Equation 6 and section 2.3, Methods) to investigate how the LAB region affects the magnitude of electric fields observed at Earth's surface. At periods of~200 s, we find that electric fields are enhanced by both increased asthenospheric conductance and increased lithospheric thickness (Figure 5a), indicating that the effects of secondary inductance are important. At periods >1,000 s, electric fields decrease as asthenospheric conductance increases and increase as lithospheric thickness increases, indicating that lithospheric thickness exerts a primary control on electric field magnitudes ( Figure 5b). Disregarding, other factors such as source field heterogeneity, our 1-D model predicts that E y electric fields during geomagnetic storms should be approximately 2-3 times larger at UPS compared to ESK. This is in qualitatively good agreement with the average trend evident in the E y electric field spectra (Figure 4), given differences in the directions of maximum conductance in the electrically anisotropic asthenosphere.

Discussion and Conclusions
We have used MT data recorded in Scotland during the 8 September 2017 storm (Simpson & Bahr, 2020a, 2020b, 2020c and geomagnetic observatory data from Sweden and Scotland recorded during the 8 September 2017 and 30 October 2003 "Halloween" storms to demonstrate that the impact of extreme space weather events cannot be adequately modeled without considering electromagnetic induction in Earth's mantle. The MT impedances that we have used to estimate electric fields are more complete than any model as they contain within them all information regarding the combined effects of ocean bathymetry, mantle heterogeneity, galvanic effects, and electrical anisotropy. Our results show that the amplitude of E y was approximately 5 times larger for UPS, Sweden, compared to RAN and ESK, central Scotland for both storms. We attribute this result to the combined effects of (i) the greater thickness of the Swedish lithosphere compared to the Scottish lithosphere and (ii) enhancement of the magnetic source fields as a consequence of secondary electromagnetic induction related to the higher conductance of the asthenosphere below Fennoscandia compared to central Scotland. In March 1989, a geomagnetic storm caused a blackout in Québec, Canada (Bolduc, 2002). Lying on the North American craton, Québec, like Sweden, is an area of thickened lithosphere (Jaupart et al., 1998).
Although our model is simplistic and used only to elucidate the processes involved, our results are significant because the effects of secondary induction in Earth's asthenosphere and lithospheric plate thickness have not previously been considered when compiling hazard maps (e.g., Love et al., 2016;Lucas et al., 2020) showing the electric fields likely to be associated with magnetic storms. We have demonstrated that for periods <1,000 s, corresponding to depths just below the electrical LAB (e.g., Simpson, 2013), which is at least 100 km deeper below Sweden than Scotland, a trade-off occurs between the amplification effect of high mantle conductance on the magnetic fields measured at Earth's surface and the attenuation effect of high mantle conductance on a given magnetic field. Therefore, secondary induction could be a source of error in space weather forecasting and GIC studies that concentrate on short period electromagnetic fields with shallow penetration depths that do not adequately constrain the problem.
The electrical properties of the lithosphere and asthenosphere are not currently well constrained for many parts of the world. Our paper demonstrates the existence of a practical application for this knowledge and the need for more MT data acquisition programs similar to Earthscope in the United States (e.g., Schultz, 2009;Yang et al., 2015). Upper crustal electrical conductivity models underlain by unrealistic deep crustal and mantle resistivities are inadequate for the purpose of calculating storm time electric fields that drive GICs.

Data Availability Statement
Electromagnetic data generated by the authors are availabe in Simpson and Bahr (2020c) and Simpson and Bahr (2020d). Geomagnetic observatory data are available from INTERMAGNET.