The GIC and Geomagnetic Response Over Fennoscandia to the 7–8 September 2017 Geomagnetic Storm
Abstract
Between 7 and 8 September 2017, Earth experienced extreme space weather events. We have combined measurements made by the IMAGE magnetometer array, ionospheric equivalent currents, geomagnetically induced current (GIC) recordings in the Finnish natural gas pipeline, and multiple ground conductivity models to study the Fennoscandia ground effects. This unique analysis has revealed multiple interesting physical and technical insights. We show that although the 7–8 September event was significant by global indices (Dst∼150 nT), it produced an unexpectedly large peak GIC. It is intriguing that our peak GIC did not occur during the intervals of largest geomagnetic depressions, nor was there any clear upstream trigger. Another important insight into this event is that unusually large and rare GIC amplitudes (>10 A) occurred in multiple Magnetic Local Time (MLT) sectors and could be associated with westward and eastward electrojets. We were also successfully able to model the geoelectric field and GIC using multiple models, thus providing a further important validation of these models for an extreme event. A key result from our multiple conductivity model comparison was the good agreement between the temporal features of 1-D and 3-D model results. This provides an important justification for past and future uses of 1-D models at Mäntsälä which is highly relevant to additional uses of this data set. Although the temporal agreement (after scaling) was good, we found a large (factor of 4) difference in the amplitudes between local and global ground models due to the difference in model conductivities. Thus, going forward, obtaining accurate ground conductivity values are key for GIC modeling.
Key Points
- The maximum peak GIC did not coincide with the sudden storm commencement
- Largest GIC peak did not occur during particularly strong driving conditions
- New comparison of 1-D and 3-D models justify the past and future use of 1-D models at Mäntsälä
1 Introduction
Space weather is the study and prediction of near-Earth plasma conditions and the ensuing effects on human technology, environment, and infrastructure. Although there are numerous consequences of space weather, geomagnetically induced currents (GICs) are a particularly hazardous effect (Boteler, 2003). They are an end link in the extensive space weather chain, and records of their effects go back over 150 years. Historically, GICs have caused physical damage to ground infrastructure (Bolduc, 2002; Pulkkinen et al., 2005; Rosenqvist et al., 2005), resulting in substantial monetary losses. GICs are set up by a geoelectric field which arises from (1) rapid variations in the geomagnetic field (dB/dt) caused by ionospheric and magnetospheric currents and (2) the conducting properties of the ground. Thus, the combination of (1) and (2) are crucial to understanding the geoelectric field and GICs since it is the geoelectric field coupled with the physical properties of the ground-based network which causes GICs to flow in ground-based conducting systems. Examples of these are power grids, railways, telecommunication cables, and oil and gas pipelines. It has been discussed recently that although enhanced peak geoelectric fields arise from the large-scale (>1,000 km) enhancement of ionospheric electrojet currents, smaller (100–500 km) spatial-scale current structures embedded within large systems are highly important (Ngwira et al., 2015) as well. Understanding the nature of small-scale ionospheric structures such as their magnetospheric driving mechanisms is crucial to advancing the understanding of GICs, improving forecasting capabilities, and the development of mitigation strategies (Pulkkinen, 2015).
An early example of a localized GIC effect was reported by Anderson et al. (1974) on the outage of the Piano (Illinois) to Cascade (Iowa) link of the L4 coaxial cable. This occurred due to voltages that exceeded the critical shutdown level. This event was later revisited by Boteler and van Beek (1999) to provide a more up-to-date interpretation of the physical magnetospheric-ionospheric (M-I) mechanisms responsible. The conclusion was that the large dB/dt was driven by a localized eastward electrojet (EE) source current at approximately 100-km altitude. It is postulated that the rapid inward movement of the magnetopause caused sudden variations in the convection current. This manifested as EE variations which drove sufficiently large dB/dt and then GICs, raising the voltage in the power cable.
Pulkkinen et al. (2003) performed a detailed case study of GICs during the 6–7 April 2000 storm. In this event, the GICs resulted from an EE intensification and the penetration of westward electrojet (WE) currents into the Fennoscandia region. The authors noticed that the irregular behavior of rapid geomagnetic field variations was likely indicative of smaller-scale (order of 100 km) current systems superimposed on the larger electrojet structure. They report that although large GICs correlated to enhanced electrojets, it was in fact the smaller scale current features which were the main drivers of GICs during this event. It is also worth noting that the largest GICs were associated with substorm activity, but geomagnetic pulsations can also be a driver of large GICs.
In a recent study performed by Ngwira et al. (2015), the authors reported that when the geoelectric field is enhanced during geomagnetic storms, the peak maximum can manifest at localized and regional spatial scales. To complicate matters, the results of Ngwira et al. (2015) show that localized geoelectric field peaks can manifest at different latitudes and during different phases of a geomagnetic storm. One M-I coupling process that they put forward to explain this phenomenon is a localized substorm. Additionally, complex ground conductivity gradients may also play a significant role in the regional occurrence of peak geoelectric fields.
Similarly, Pulkkinen et al. (2015) studied the localized nature of geoelectric fields, but from a statistical standpoint. IMAGE data were used to conduct this study in which they compared station-specific geoelectric fields to those which were spatially averaged (over 500 km). The results are applicable to latitudes between 55° and 75°. The authors reported that the geoelectric field is spatially and temporally localized and that spatially averaged and station-specific geoelectric field values can be significantly different over distances of around 250 km. These effects were observed during extreme storm conditions and can be driven by magnetospheric processes such as substorms and geomagnetic pulsations. However, this study did not account for the local full 3-D ground conductivity profile, and thus the results are valid in the absence of significant local geophysical anomalies.
According to the literature (Ngwira et al., 2015, 2018; Pulkkinen et al., 2003, 2015; Viljanen et al., 2006), substorms are a likely driver of localized peak geoelectric fields. Additional evidence of this was recently presented by Ngwira et al. (2018) who showed that for two geomagnetic storms, large and localized dB/dt were associated with substorm activity. They report that although the magnetospheric dynamics of substorms can be global (based on Time History of Events and Macroscale Interactions during Substorms data), the ground dB/dt response can manifest in the form of localized extreme peaks. The authors also propose that intense dB/dt may be driven by the poleward edge of the poleward expanding discrete aurora. Nevertheless, the authors acknowledge that more work is required to resolve this, and many open questions still remain.
The physical nature of substorms (Akasofu, 1964), particularly their triggering mechanisms, have been heatedly debated for decades (see Kan et al., 1991, and references therein). Although the purpose of this study is not to join this debate, it is worth noting that substorm triggering and the inherent energy loading-unloading processes (Tanskanen et al., 2005) are crucial to understanding the complex localized ground responses experienced during geomagnetic storms. Viljanen, Tanskanen, and Pulkkinen (2006) studied the relation between substorms and rapid temporal variations of the geomagnetic field. Their data implied that the largest dB/dt are typically observed during the midnight sector, which is consistent with substorm activity. In addition, storm-time substorms resulted in dB/dt values which were generally twice as large as nonstorm substorms, posing a greater GIC hazard. The majority of large dB/dt were related to WE currents, but there was notable scatter, suggesting that peak dB/dt values can arise from smaller-scale ionospheric current structures such as vortices. This is consistent with the findings of Pulkkinen et al. (2003).
Important to consider is what classification of disturbed solar wind conditions drive large GICs. In Huttunen et al. (2008), it was shown that the largest GIC amplitudes are driven by interplanetary coronal mass ejections (ICMEs) compared to corotating interaction regions and that the ICME sheath component was responsible for the largest GIC values. Boundary layers (separating ejecta and ambient solar wind) and ejecta portions of ICMEs were also capable of generating large GICs, but their amplitudes were smaller. However, when computing correlations between GICs and ICME plasma and field properties, there was significant spread in their data suggesting additional factors should be considered. According to Kappenman (2003), magnetospheric shocks from large interplanetary pressure pulses can also drive significant geoelectric fields which are hazardous from a GIC perspective. More importantly, such events can do so at any latitude and thus demonstrate that GICs even pose a threat at equatorial latitudes. This was also recently shown by Carter et al. (2015).
Drawing upon the existing literature (Boteler & van Beek, 1999; Ngwira et al., 2015, 2018; Pulkkinen et al., 2003, 2015; Viljanen et al., 2006), it is clear that peak geoelectric fields differ over spatial scales smaller than or comparable to ground conducting networks (<500 km). These localized effects are attributed to both localized ionospheric current sources (which drive geomagnetic activity) and/or complex ground conductivity features such as coastlines and mineral deposits. What is clear is that the M-I mechanisms responsible require further study. In addition, the three-dimensional ground conductivity profile also plays a large role in dictating the properties of the geoelectric field (Bedrosian & Love, 2015). In regions where geophysical anomalies or gradients are present, such as Fennoscandia (Engels et al., 2002; Korja et al., 2002), this element has to be taken into account. The motivation for the present paper is to analyze the complex and localized geomagnetic and geoelectric ground effects over Fennoscandia for the recent 7–8 September 2017 storm. We study the ICME properties, ground magnetic response, ionospheric equivalent currents, geoelectric field, and GICs which flow in the Finnish natural gas pipeline at the Mäntsälä compressor station. We also implement a three-dimensional ground conductivity model (Engels et al., 2002; Korja et al., 2002) to determine the geoelectric field and model the largest interval of GICs.
The manuscript is structured as follows. Section 2 provides details of the experimental data, data-derived products, and the models. Section 3 provides a brief overview of the 7–8 September ICME according to OMNI data. The results are presented in section 4, before discussing their physical implications and context with the existing literature in section 5.
2 Experimental Data and Models
2.1 Solar Wind
For solar wind measurements, we extracted plasma and field parameters from NASA/Goddard Space Flight Center's OMNI data set (60-s resolution) via the OMNIWeb (http://omniweb.gsfc.nasa.gov) service. These data represent the solar wind conditions at the bow shock nose according to the bow shock model by Farris and Russell (1994). It should be noted that the OMNI data set is not measured in situ at the bow shock nose by a physical spacecraft. It is composed of multiple spacecraft observations which have been propagated (King & Papitashvili, 2005) from further upstream using parameters measured in situ close to the L1 point.
2.2 Ground Measurements and Ionospheric Equivalent Currents
To assess the ground response over Fennoscandia, we use data provided by the IMAGE magnetometer network (Tanskanen, 2009). The network consists of 42 stations distributed over latitudes from 51° to 79° . Their main purpose is to investigate the ground response over Fennoscandia due to the auroral electrojet currents. The cadence of the IMAGE magnetometer data is 10 s, meaning it is also suitable to measure the rate of change of magnetic field (dB/dt) which is required for GIC studies. Figure 1a shows a map of the IMAGE stations as of 2017.
As demonstrated by Figure 1, the station coverage is relatively dense between latitudes of 65° and 70°, and in general across Finland. This provides the capacity to resolve spatially localized ground responses down to 100 km in some regions. These complicated effects are driven by localized ionospheric current sources. In addition, the broad coverage of the network is well suited to deriving ionospheric equivalent currents (IEQ).
We use the IEQ system developed by Amm and Viljanen (1999) which uses the spherical elementary current method (Amm, 1997). Here, ground magnetic measurements by IMAGE are used to derive the equivalent currents in the ionosphere at 100-km altitude. Note that the IEQ are horizontal ionospheric currents which would reproduce the observed ground magnetic field. In reality, the true ionospheric currents are a 3-D combination of horizontal and field-aligned currents. Using only ground magnetometer data, it is not possible to distinguish between these.
The GIC data used in this study are from the Finnish natural gas pipeline at the Mäntsälä (MAN) compressor station in which measurements have been recorded regularly since November 1998. The MAN station is located in southern Finland (60.6°N, 25.2°E) and is approximately 40 km from the reference magnetometer measurements at Nurmijärvi (NUR); Pulkkinen, Pirjola, et al. (2001) discuss the uncertainties associated with this distance. Figure 1b shows a map of the pipeline in which the main lines and some minor branches are included (black trace). To the eastward direction, the pipeline crosses into Russia as shown by the dashed black line. The GIC is determined by measuring the magnetic field above the pipeline that is created by natural variations and the GICs. The reference signal from the nearby Nurmijärvi (NUR) magnetometer station is used to eliminate the non-GIC contribution. The current along the pipeline can then be determined using Biot-Savart law. A more complete description of how these measurements are conducted is provided by Pulkkinen, Viljanen, et al. (2001).
It is worth noting that pipelines usually have a cathodic corrosion protection system that tries to keep them at a lower potential than the surrounding soil. As recognized by Pulkkinen, Pirjola, et al. (2001), the related current is inevitably mixed with GIC and cannot be eliminated. So the corrosion protection system causes some uncertainty. However, Viljanen, Tanskanen, and Pulkkinen (2006) showed that the measured current behaves like a “true” GIC since it is clearly related to dB/dt and the modeled geoelectric field. Thus, the measurement of the current flowing along the pipeline is justified. As observations show, if the geomagnetic field is quiet, the measured current is small despite some noise typically less than 1 A. From the corrosion viewpoint, it is the current that entered the pipeline which is important, whether it is GIC or some unknown source; the current must exit at some point and thus may cause corrosion.
2.3 Ground Conductivity Models
We utilize the crustal conductivity maps developed by Korja et al. (2002), which covers Fennoscandia and the surrounding sea water (SMAP). A full 3-D conductivity model is defined based on the SMAP conductance in three sheets with fixed thicknesses. Figure 2 shows the conductance profiles for each sheet and their corresponding depths. Figure 2d shows an illustration of the 3-D conductivity model. Figure 2a covers depths from 0 to 10 km and includes the surface, sea water, sediment, and the upper crust; strong coastal gradients due to the highly conducting sea water are present in this layer. Figure 2b corresponds to 10–30 km and is the middle crust. Figure 2c is the lower crust for depths of 30–60 km. A schematic of the 3-D conductivity model is shown in Figure 2d in which the lateral conductivity in each layer is given by the distribution in Figures 2a–2c. The underlying deep crust/mantle conductivity is chosen to be σ∞=10−3 S/m in accordance with the conductivity between 60 and 100 km in the standard normal model utilized by Engels et al. (2002). The conductance of the water was “estimated from bathymetric data by converting depths to conductances and taking into account the salinity variations in the Baltic Sea” (Rosenqvist & Hall, 2019). For the bedrock, the conductances were estimated from 1-D and 2-D Magnetotelluric surveys. Please see Rosenqvist and Hall (2019) for further details of the model and its validation.
3 The 6–9 September 2017 Event
Figure 3 shows OMNI data between 00:00 on 6 September and 18:15 on 9 September 2017. Figures 3a–3f show |B|, Bx,y,z (geocentric solar magnetospheric), |V|, ni, Ti, and SYM-H. The interplanetary shock observed at L1 at ∼ 23:50 UTC on 6 September was produced by two Earth-directed coronal mass ejections (CMEs) launched in short succession on 4 September, which likely interacted close to the Sun. Another interplanetary shock arrived at 23:00 UTC on 7 September, which was caused by a fast CME that erupted on 6 September (Shen et al., 2018; Werner et al., 2019).
This CME reached the Earth relatively soon, implying that, despite being fast, the interplanetary space was already preconditioned by the previous CMEs. It caught up with the preceding structure and propagated through its magnetic cloud, resulting in a strong shock followed by a highly compressed sheath region. The first decrease in the SYM-H index (Figure 3f) indicates the onset of the geomagnetic storm and is associated with the arrival of the ejecta at ∼ 23:00 UTC with southward Bz∼−10 nT. With the arrival of the southward orientated interplanetary shock (∼−30 nT), SYM-H sharply decreased again to ∼−100 nT, reaching a minimum of ∼−150 nT. Finally, the magnetic ejecta of the third CME arrives at 11:55 UTC with Bz also southward oriented. This was responsible for another SYM-H decrease from ∼−50 to ∼−110 nT.
The biggest SYM-H minimum was associated with the passage of the second shock and the following CME sheath region. The combination of the shock-inside-CME structure, strong southward Bz, and the strongly compressed sheath (characterized by large amplitude and sharp magnetic field gradients and directional changes) is likely the cause of the strong geomagnetic response. In recent work (Lugaz et al., 2015), it has been shown that similar shock-inside-CME structures account for 20% of all geomagnetic storms in cycle 23. Also, about 15% of the intense storms in the same cycle are caused by the sheath region behind shocks (Myllys et al., 2016; Zhang et al., 2007).
3.1 Historical Significance
Listed in Table 1 are the top 20 daily maximum GIC values for the Finnish natural gas pipeline at MAN from 1998 to 2017. We should note that due to occasional problems with the measurements, some large events are excluded from this list. Nevertheless, the current event is listed as number 10 in these statistics, suggesting that from a GIC perspective, the 7–8 September 2017 storm was extreme. It should be mentioned that the configuration of the pipeline has been slightly modified during the collection of these statistics; however, the values are comparable. The daily maximum GIC was 30.1 A and is approximately half of the daily maximum measured during the 2003 Halloween storm. Similarly, in Table 2, the daily maximum WE currents derived from IMAGE data are listed for the period 1994–2017. As a side note, the largest EE within the same interval was 2.95 MA on 24 November 2001. The 7 September 2017 storm is listed at 8 and 9 showing that this storm produced substantial and extreme WE currents over middle to high latitudes. In fact, by comparing Tables 1 and 2, one can see many similarities, suggesting that large GIC events are likely to occur during extreme WE currents at these locations.
| # | Date | Maximum GIC (A) |
|---|---|---|
| 1 | 2003/10/29 | 57.0 |
| 2 | 2003/10/30 | 48.8 |
| 3 | 2004/11/09 | 42.8 |
| 4 | 2012/03/15 | 39.1 |
| 5 | 2004/11/07 | 34.8 |
| 6 | 2001/11/24 | 32.0 |
| 7 | 2013/03/17 | 31.6 |
| 8 | 2001/11/06 | 31.6 |
| 9 | 2003/10/31 | 30.3 |
| 10 | 2017/09/08 | 30.1 |
| 11 | 2000/07/15 | 30.1 |
| 12 | 2004/11/08 | 29.1 |
| 13 | 2003/10/14 | 28.7 |
| 14 | 2002/10/01 | 28.1 |
| 15 | 2005/01/21 | 27.5 |
| 16 | 2002/09/07 | 26.4 |
| 17 | 2002/09/08 | 24.5 |
| 18 | 2003/11/20 | 23.8 |
| 19 | 2000/04/06 | 22.7 |
| 20 | 2001/04/11 | 21.8 |
- Note. The current event is #10 (in bold). Dates are formatted as year/month/day. GIC = geomagnetically induced current.
| # | Date | Maximum WE (MA) |
|---|---|---|
| 1 | 2003/10/30 | −8.41 |
| 2 | 2001/11/06 | −5.25 |
| 3 | 2003/10/29 | −5.05 |
| 4 | 2004/11/08 | −4.82 |
| 5 | 2000/09/17 | −4.69 |
| 6 | 2003/11/20 | −4.47 |
| 7 | 2004/11/09 | −4.20 |
| 8 | 2017/09/08 | −4.13 |
| 9 | 2017/09/07 | −4.06 |
| 10 | 2005/01/07 | −3.96 |
| 11 | 2003/05/29 | −3.95 |
| 12 | 2004/11/07 | −3.73 |
| 13 | 1999/09/22 | −3.70 |
| 14 | 2006/12/15 | −3.66 |
| 15 | 2006/12/14 | −3.56 |
| 16 | 1998/05/04 | −3.47 |
| 17 | 2015/03/17 | −3.43 |
| 18 | 2000/07/15 | −3.09 |
| 19 | 1998/02/18 | −3.05 |
| 20 | 2011/08/05 | −3.04 |
- Note. The current event is #8 and #9 (in bold). Dates are formatted as year/month/day. WE = Westward Electrojet; GIC = geomagnetically induced current.
4 Results
4.1 GICs in the Finnish Natural Gas Pipeline at the Mäntsälä Compressor Station During 7–8 September 2017
Plotted in Figure 4 are solar wind and ground measurements during the period 7 September 2017 at 18:00 to 8 September 2017 at 23:59 UTC, which was during the merged ICME interval. For reference, Figures 4a and 4b show the following solar wind parameters: |B|, Bz, |V|, and ni. The magnetic field measurements recorded at NUR can be found in Figure 4c along with the corresponding dBx/dt in Figure 4d below. The quantity dBx/dt is computed from the difference between the samples divided by sampling time dBx/dt=(Bx(n+1)−Bx(n))/Δt, where Δt is 10 s; note that this is then scaled from nanotesla per second to nanotesla per minute for easier comparison with the existing literature. The GICs measured in the Finnish natural gas pipeline at the MAN compressor station is plotted in Figure 4e. The magnetogram below in Figure 4f is constructed from station measurements which are similar in longitude, but with increasing latitude (top to bottom). Figure 4g shows the PCN index which can be thought of as a proxy describing the global energy deposited into the polar cap region. The blue highlighted regions mark intervals which are of specific interest to this study due to the enhancement of GICs shown in Figure 4e. From here on, we refer to these periods as intervals [1]–[3]. We note that there is, unfortunately, occasional noise up to several amperes in the measured GICs. However, the most active intervals can be easily identified, and they occur consistently with enhanced dBx/dt values.
4.1.1 Interval [1]
Around 20:30 on 7 September 2017, the upstream Bz changes orientation to southward as shown in Figure 4a. In concert, the ground measurements shown in Figure 4f exhibit gradual magnetic field decreases (e.g., Oulujärvi, Ranua, Sodankylä, and Ivalo), suggesting an increase in magnetospheric convection. Approximately 2.5 hr later, the interplanetary magnetic field (IMF) Bz further decreases from −10 to around −30 nT due to the southward orientated ICME shock. This immediately caused substantial magnetic depressions over a large range of latitudes; although it seems RAN-SOD recorded the largest depressions first. This interval is marked as [1] and represents the first region of interest due to the sudden commencement of GIC activity and dBx/dt in Figures 4d and 4e). This is also accompanied by strong WE currents, as expected since the IMAGE stations were in the midnight sector. The GICs during this interval exceeded 20 A, and it is worth noting that there is a delay of around 1 hr from the intensification of Bz (23:20) and the peak GIC at 00:31 on 8 September. It is worth pointing out that at approximately this time, shown in Figure 4b, there was a local peak in ion density and a small rotation of the IMF. Almost in parallel, the NUR station records a strong (−200 to −800 nT) magnetic depression and large dBx/dt at a similar time. It is difficult to determine if this is the cause or related to the GIC enhancement since the upstream and ground timings are very close, and any propagation error from L1 may also contribute to difficulties correlating features on such small scales. Therefore, even though direct confirmation is difficult, it is important to highlight the timeliness of the features in the OMNI and NUR data. Finally, we should note that at 20:00 and 23:00, there are sharp increases in the PCN index which correspond to the enhancements in Bz, thus indicating a global deposit of energy into the polar cap region at these times. This is expected due to the enhanced magnetospheric convection, driven by the initiation of dayside magnetic reconnection.
An expanded view of interval [1] is plotted in Figure 5. Figures 5a–5e show |B|, Bx,y,z, |V|, ni, and Pdyn=mpniV2 according to OMNI. Plotted in Figure 5f is the magnetic field measurements from the NUR station in which the black, blue, and red traces correspond to the Bx, By, and dBx/dt quantities, respectively. Figure 5g is the GICs measurement at MAN. In this panel, the threshold of 5 A has been marked since it was shown by Pulkkinen et al. (2003) that GICs above this value at MAN can be considered as rare. It is clear that large GICs levels are measured at the beginning of this interval due to the southward directed IMF. However, the extreme GIC values occur after about 90 min into this interval and the peak occurs at 00:31 UT. At 00:29, a sudden increase in ion density was measured, resulting in an associated spike in Pdyn as shown in Figures 5d and 5e. In concert, a rotation in the IMF is observed according to the By and Bz components shown by the blue and red traces in Figure 5b, respectively. Approximately 1 to 2 min afterward, the NUR station records a large and sudden increase in the WE current, which is responsible for correspondingly large dBx/dt values, thus setting up a large GICs in the nearby MAN pipeline in excess of 25 A. It is our interpretation that the physical mechanism causing this sudden enhancement of a WE is associated with a substorm which occurred during the existing geomagnetic storm. This is founded in the following evidence: (1) NUR is located in the midnight sector and observed a strong WE enhancement, consistent with the substorm electrojet, (2) the IMF had been directed southward for approximately 4.5 hr providing ample tail loading, (3) the GIC enhancement is correlated with a notable increase in dynamic pressure suggesting a possible triggering mechanism, and (4) the IMAGE-derived AL index (IL) shows a sudden depression of around −2,000 nT. It should be noted that as mentioned before, the short timings make (3) difficult to confirm, and thus although the features of the event are consistent with a substorm, these upstream features cannot be confirmed as the trigger. It is also noteworthy that in addition to the peak GIC, large GICs were also observed during this interval which on two occasions reach almost 20 A, and on many occasions exceeded 5 A (although a part of the values above 5 A are noisy).
4.1.2 Interval [2]
Similar to interval [1], the second interval [2] of enhanced GICs is also accompanied by a southward IMF which turns from north-south at 11:00 on 8 September (see Figure 4). This interval corresponds to the beginning of an ICME ejecta evidenced by the smooth magnetic field profile. At this time, the IMAGE stations are in the dawn postnoon sectors which is reflected by the EE (increase of Bx) seen in Figure 4c. The GICs during this time are substantial (∼ 15 A) and prolonged (∼ 5 hr). Although the dBx/dt in interval [2] does increase to levels comparable to interval [1], there are no notable peak GICs similar to the previous interval but more of a prolonged enhancement. The PCN index does increase significantly in Figure 4g, suggesting that the energy input to the auroral region is enhanced during this time. The decreased (but still significant) GICs compared to interval [1] are likely due to the fact that they are associated with an EE which is generally weaker than their westward counterpart (Kappenman, 2006). Interestingly, there are large fluctuations in Bx between 14:00 and 15:00 suggesting rapid variations in the EE.
A more detailed plot of interval [2] is provided in Figure 6, and the format is identical to that of Figure 5. The immediate contrast is that the GICs here are driven by an EE evidenced by the increase of the horizontal magnetic field to almost 1,000 nT at NUR. The EE is expected due to the dawn-noon sectors in which the IMAGE stations occupy during this time. The IMF turns initially southward around 11:10 UT, and then again at 11:30, shortly after, GICs in excess of 5 A are observed due to dBx/dt induced by changes in the EE. The data at 14:00 is particularly interesting since sudden and rapid variations in Bx occur, suggesting rapid temporal variations in the EE currents. Since the general trend of Bx is preserved between 14:00 and 15:00 in Figure 6f, it appears as if smaller spatiotemporal features are superimposed with the large structure between 11:30 and 15:00. Unsurprisingly, the sharp changes in Bx between 14:30 and 15:00 are accompanied by elevated dBx/dt and GICs; the GICs reach values of approximately 15 A. Although it is not a one-to-one correlation, it is noteworthy that the temporal variations in the GIC measurements seem to correspond to those observed in Bx (although noise disturbs the GIC time series). Later in the interval (∼15:15), there are also changes in Bx which induce GICs above 5 A. Regarding the external driving, at 14:00 when the Bx changes begin, there is also a steady increase in ni and Pdyn as shown in Figures 6d and 6f. At 14:45 close to the peak of this Pdyn increase, there are two sudden south-north-south rotations in the IMF in quick succession, as shown by the red trace in Figure 6b. This interval is a clear demonstration that it is not the steady increase (as we see at the beginning of the interval) of an ionospheric current which drive GICs but more the smaller-scale spatiotemporal variations.
4.1.3 Interval [3]
There is a substantial maximum peak in GICs during interval [3] around 18:00 on 8 September (see Figure 4e). In fact, in terms of amplitude and not duration, the GIC was largest during this interval slightly exceeding 30 A (now there is no obvious noise any more). Surprisingly, this value is even greater than that measured during interval [1] during arguably greater geomagnetic activity. What is particularly noteworthy about this observation is the upstream conditions, or lack thereof. The largest GIC measured in the MAN pipeline during the 7–8 September 2017 storm does not correlate with any clear upstream transient signature. There is a moderate increase in the PCN index, but nothing comparable to the preceding intervals. Nevertheless, the IMF was southward for ∼ 3 hr, but the magnitude of the southward component was relatively weak (∼10–15 nT). It should be mentioned that it is possible that interval [3] corresponds to the edge of the ejecta region, as suggested by the small change in magnetic field. However, no significant evidence of this was provided by the plasma parameters shown in Figure 4b which remain unchanged for several hours afterward. Therefore, we suggest that this is likely not the reason.
A more detailed plot of interval [3] is provided in Figure 7 in which the format is identical to Figures 5 and 6. This figure covers the interval 17:30–19:00 on 8 September. The peak GIC at 17:54 UT is clear in Figure 7g and is associated with a sudden enhancement in dBx/dt as shown in Figure 7f above. Interestingly, there are variations in the ground Bx and By, indicative of a local vortex ionospheric current flow pattern, or another small spatiotemporal feature. The time after the peak value at 18:00, the GICs also regularly exceed the 5-A thresholds (red dashed trace). Although the geomagnetic variations (and dBx/dt) are smaller compared to the previous 10 min, the GICs still reached significant values. Thus, in this event, the peak GIC at MAN did not occur in the most geomagnetically active interval.
4.2 Ionospheric Equivalent Currents During the Three Enhanced GIC Intervals
In this section we present maps of IEQ during intervals [1]–[3]. We also direct readers to the supporting information to view video clips of IEQ over each interval, and the entire 7–8 September period.
4.2.1 GIC Interval [1]
Presented in Figure 8 are IEQ and ground measurements during interval [1]. Figures 8a–8e show maps of the IEQ ±4 min around the peak GIC at 00:31 plotted in Figure 8c. An expanded version of Figure 8c can be found in Figure 8f. To highlight the differences between magnetometer stations, several time series of magnetic field measurements are shown in Figures 8g–8k, which cover a 4-hr period. We have indicated in these time series the GIC peak and also the start time of interval [1], which was highlighted in Figure 4 and presented in greater detail in Figure 5. The color scale is shared between each panel for the sake of comparison.
What is immediately obvious from Figure 8 is that the spatiotemporal behavior of the IEQ is highly dynamic during this time. Please also see the supporting information for the video of IEQ during this period (Movie S1). Four minutes prior to the GIC peak, there is a large enhancement of the currents covering a latitude range of 60–68° consistent with the equatorward expansion of the auroral oval during a geomagnetic storm. Two minutes later, this band appears to separate, and more localized structures can be found. Thus, the same large-scale WE current is being observed during this time, but it appears that smaller spatiotemporal structures begin to develop which are superimposed or embedded within the larger current system. From Figures 8b–8e one can see that the spatiotemporal evolution of the currents continues, resulting in many possible smaller-scale current sources from both temporal and spatial standpoints. This dynamic behavior creates correspondingly complex and localized ground signatures. The dBx/dt profiles are particularly erratic, and most stations which are separated by hundreds of kilometers show markedly different values. For example, Figures 8h and 8i show notably large magnetic depressions and correspondingly large dBx/dt. The magnetic field profiles of these stations differ greatly, for example, during 23:00–23:12 on 7 September, and 00:12–00:36 on 8 September. These smaller-scale current structures reach latitudes down to 60° and thus drives the GICs at MAN. From these observations, it appears that the occurrence of these smaller-scale structures coincides with the enhanced GIC and dBx/dt.
4.2.2 GIC Interval [2]
The IEQ during interval [2] are mapped in Figure 9 at four time instances separated by 60 min over 12:00–15:00 UT. The movie of IEQ during interval [2] is available in the supporting information (Movie S2). The common feature between each panel is the presence of an EE current, at latitudes from 58° to 70°. At 13:00 UT shown in Figure 9b, the large EE has reached the latitudes of MAN which drives GICs of approximately 5 A in the pipeline. During the interval between 14:00 and 15:00 UT, the elevated EE remains at similar latitudes, but the behavior becomes more erratic (see the supporting information Movie S2 01:12–01:48). In general, there is a pulsation-like behavior of the EE in which enhancements are shortly followed by reductions. There are three instances of this over the 1-hr period. This effect is also apparent from Figure 6f by the variations of Bx observed between 14:00 and 15:00. These fluctuations have time periods of approximately 18 min. It is worth mentioning that these short-time variations appear superimposed on the larger EE structure. This is evidenced by the fact that the larger structure (duration of 180 min) of the magnetic field profile in Figure 6 appears to be preserved. The IEQ structures differ from interval [1] since they do not possess vortex-like properties but are pulsations within the larger EE system. Nevertheless, it was capable of setting up large GICs.
4.2.3 GIC Interval [3]
Plotted in Figure 10 are the IEQ and selected ground measurements during interval [3]. The format of this figure is the same as that of Figure 8. A movie of the IEQ during this interval is available in the supporting information (Movie S3). From Figures 10a–10e, it is clear that the IEQ are highly complex, localized, and dynamic during this interval. At the time of the peak GIC (30 A) there is a complex rotation of the IEQ which extends over UPS (Uppsala, Sweden) and NUR stations. This is also realized in the ground signatures since both Bx and By exhibit rotational behavior. This rotation is accompanied by significant dBx/dt and then a sudden peak GIC. In higher latitudes (65–70°), the IEQ and geomagnetic variations are even larger, and additional rotational source current structures are observed. In general, the IEQ behaves quite erratically over this interval in which multiple spatial and temporal source currents are observed, creating localized ground signatures. For example, in Figures 10g, 10h, and 10j, large dBx/dt are measured at 18:15, but the dBx/dt in Figures 10i and 10k are comparatively lower. What is also apparent is that the largest magnetic depressions do not necessarily result in the largest dBx/dt values, which appears much more complex. Comparing this interval to the previous one, the ionospheric structures here are more vortex-like (Apatenkov et al., 2004; Henderson, 2013; Henderson et al., 2001; Viljanen et al., 2001) and rotational by nature, whereas the behavior during interval [2] manifested as pulsation-like variations in the EE. It is clear that these observed complex geomagnetic and ionospheric disturbances correlate very well with the dBx/dt and GIC recordings. This is indicative that the geoelectric field variations at MAN were primarily driven by external sources. This may be evidence that the local ground conductivity is rather uniform, and a 1-D approach may be sufficient. This point will be expanded upon and verified in the following section using model results.
4.3 Model Results
From an operational space weather perspective, it is useful to have the capability to model the surface geoelectric field and reconstruct the observed GICs. Here we focus explicitly on interval [3] since it was the largest peak GIC that occurred during this event and appeared to contain the least noise. We adopt 3-D modeling using the conductivity model shown in Figure 2. To model the induced electric field, the current distributions in the ground are solved in the frequency domain by the finite element method (FEM) using the COMSOL Multiphysics® software package. A description of the FEM tool is available on the COMSOL Multiphysics® webpage (https://www.comsol.eu/multiphysics/finite-element-method). The technical setup of the run is made available in the supporting information (Text S1) in the form of a COMSOL report. The geoelectric field is first modeled using a uniform source current which is then scaled by the magnetic field from NUR. We also repeated these results based on the interpolated magnetic field at MAN. Since we observed no notable difference, we used the measured (at NUR) rather than interpolated (at MAN) magnetic field because interpolation also adds some uncertainty. The geoelectric field is modeled in the frequency domain at discrete frequencies; therefore, we interpolate the amplitude and phase response at frequencies corresponding to the magnetic field spectrum. An inverse Fourier transform is then applied in order to transform into the time domain. It should be noted that our model results do not account for spatial variations of the source. However, over a few hundred kilometers around MAN, this is a reasonable simplifying assumption (Viljanen et al., 2006). Presented in Figure 11 is the induced surface geoelectric field over Fennoscandia due to a unit amplitude incident northward (Figure 11a) and eastward (Figure 11b) magnetic field (for period t=1,000 s) from a spatially uniform current source. For the geoelectric field modeling, the computational load per one frequency and one source excitation is about 15 min on a normal desktop computer (32-GB memory); however, this can increase to 1.5 hr for finer-grid resolutions. It becomes clear that some regions are prone to stronger surface geoelectric fields, such as the Norwegian coastline. This coast effect arises due to the enhancement of the geoelectric field due to the large conductivity gradient between the ground and sea water (Gilbert, 2005), although the Gilbert (2005) cases may overestimate the enhancement in a case like this since our conductivities on the coast side are not as extreme as in their examples. It can be pointed out that there are also irregularities and structures associated with the costal effect in Figure 11. This is likely caused by the complex 3-D ground conductivity gradients as demonstrated in Figure 2. Nevertheless, this is beyond the scope of the current study but warrants future investigation. An important point is that this amplification of the geoelectric field is not restricted to coastline proximity, and similar effects can be seen in other regions which contain large contrasts in ground conductivity. Another important thing to be noted is that 1-D models do not account for lateral conductivity features, and therefore in regions with significant lateral conductivity gradients, 3-D models will be required.
In addition to the SMAP 3-D (uniform source) model, we also employ the 1-D model by Viljanen, Pulkkinen, et al. (2006), as well as the 1-D conductivity profile at the MAN location from the SMAP model. The conductance of each layer adopted in each model is shown in Table 3 together with the layer thickness. From Pulkkinen, Viljanen, et al. (2001), the GICs can be modeled using the following expression GIC=aEx+bEy, where the coefficients a and b are dependent on the spatial configuration and resistance of the pipeline. At MAN, a and b are −70 and 88 Akm/V, respectively. This assumes a spatially uniform geoelectric field over the area of the pipeline, or at least in a large enough region around Mäntsälä. The modeled GIC is therefore obtained by evaluating this expression. It should be noted that the GIC could also be obtained by integrating along the pipeline. Although the above expression is a simplification, it works reasonably well as shown by Viljanen, Tanskanen, and Pulkkinen (2006).
| Z Vil (S/m) | SMAP original (S/m) | SMAP adjusted (S/m) |
|---|---|---|
| 0.02567 (150 km) | 7.553×10−5 (10 km) | 7.553×10−5 (10 km) |
| 2.5974 (∞ km) | 1.20×10−3 (20 km) | 1.20×10−3 (20 km) |
| 1.40×10−3 (30 km) | 1.40×10−3 (30 km) | |
| 1.00×10−3 (∞ km) | 1.6 (∞ km) |
- Note. The value in brackets corresponds to the layer thickness. Z Vil = Viljanen model.
Figure 12 shows the comparison between the modeled geoelectric field (Figures 12a and 12b) and the reconstructed GICs (Figure 12c). An important thing to note is that the Viljanen, Pulkkinen, et al. (2006) model adopts a 1-D model with significantly larger conductivities than SMAP. Thus, the surface impedances between the models are significantly different by about a factor of 4 for the frequency range of interest. As a result, the amplitude of the geoelectric field based on the SMAP model (both 1-D and 3-D) is predicted to be about a factor of 4 higher than that of the Viljanen, Pulkkinen, et al. (2006) model. In order to make the predictions comparable to the Viljanen, Pulkkinen, et al. (2006) model and to the observed GICs, it is necessary to increase the conductance in the SMAP model. To investigate this, the conductivity in the underlying layer is systematically increased and the least squares error between the measured and modeled GICs was computed in order to find the conductivity value which provided the best fit to the data. We then produced an adjusted SMAP model by selecting this conductivity for the underlying deep crust layer. As shown in Table 3, the adjusted SMAP has a conductance of 1.6 S/m in the bottom layer which has been significantly adjusted from 0.001 S/m. However, it is now more in line to that used by Viljanen, Pulkkinen, et al. (2006). For that reason, the adjusted SMAP 1-D amplitudes (blue trace) are comparable to the Viljanen, Pulkkinen, et al. (2006) result (green trace). It is worth noting that the main difference between these two model results is the amplitude of the higher frequencies. This is because the conductance in the upper layers (10–60 km) of the SMAP model is smaller than the Viljanen, Pulkkinen, et al. (2006) model. The difference in geoelectric field amplitudes as a function of depth is a direct consequence of the skin depth of the earth and is a function of frequency. Lower frequencies correspond to deeper layers, which is why the higher frequencies are affected by the upper layers. This is a clear demonstration of why models with a realistic conductivity in each layer is crucial.
The SMAP 3-D model plotted in Figure 12 has been scaled by a factor of 4. This is required since the 3-D model uses the original SMAP conductivities in each layer, meaning that the geoelectric field amplitudes are significantly larger. What is important is that the temporal variations of the scaled SMAP 3-D model are in particularly good agreement with the remaining models. In fact, one could argue that the result is an improvement of the adjusted 1-D model. There is one caveat to mention: The discrepancies at the higher frequencies are not clear in the scaled SMAP 3-D model, and this is likely due to the fact that the entire signal is scaled. Therefore, the scaling has an influence on the entire frequency range rather than just dampening the lower frequencies—which is the case in the adjusted SMAP 1-D model. It is important to remember that the 3-D model in this case does not account for spatial variations of the current source but lateral variations in ground conductivity. Moving forward, the equivalent currents will replace the uniform current in this procedure; however, this will be addressed in future studies in which the requirements for 1-D and 3-D models are more comprehensively investigated.
5 Discussion
Our aim is to study the geomagnetic and geoelectric response over Fennoscandia to the 7–8 September 2017 space weather events, particularly their relation to GICs and subsequent regional effects. This study utilizes a synergy of observations (OMNI and IMAGE), data-derived products (equivalent currents), and models (ground conductivity and GICs) to comprehensively study the geomagnetic and geoelectric response over Fennoscandia to the 7–8 September 2017 event. A motivation for this study was to investigate the occurrence of GIC events over 30 A and the potential driving mechanisms. In general, three periods of enhanced GICs were observed, all in concert with southward IMF intervals. The largest GIC did not occur during the intervals when the geomagnetic depressions and enhancements were of their highest amplitude, nor were there any clear upstream trigger. In each case, the largest dBx/dt and GIC amplitudes occurred during rapid spatiotemporal ionospheric current variations which appeared to be relatively localized (several hundred kilometers). In general, GICs were driven by both WEs and EEs, and GIC enhancements occurred during both sheath and ejecta components of the ICME.
5.1 Observations
What seems particularly clear regarding the September 7–8 interval and the Mäntsälä GIC data set is that enhancements in GIC appeared to be driven by a multitude of processes such as southward IMF turnings, embedded small-scale ionospheric current structures, EE and WE enhancements, substorms, and global magnetospheric convection. Thus, the complexity of the ground response appeared to reflect the complexity of the upstream measurements.
The first interval of GICs occurred when the IMAGE stations were located in the midnight sector (11:00 on 7 September to 02:30 on 8 September), beginning shortly after the southward orientated ICME shock (the sudden storm commencement). This was shown in Figure 4 and in more detail in Figure 5. Substorm activity likely played a role in the large midnight WE shown in Figures 4 and 8; according to Viljanen, Tanskanen, and Pulkkinen (2006), large dBx/dt are likely to occur in the midnight MLT sectors due to substorm activity. This result is also consistent with a similar case study performed by Pulkkinen et al. (2003) for the 6–7 April 2000 geomagnetic storm, although their GIC readings in the same pipeline were smaller than in this case. Further evidence of a substorm during this time was reported by Clilverd et al. (2018) based on the Wp index (Nosé et al., 2012). Similar to this study and around the same UT time, Clilverd et al. (2018) showed enhanced GICs of almost 35 A at the Halfway Bush substation in Dunedin, South Island, New Zealand. Thus, this interval was part of a large-scale response to the southward orientated ICME shock since multiple global locations experienced unusually large values of GIC.
The second interval of enhanced GICs (Figures 6 and 9) between 11:00 and 16:00 on 8 September was the longest, but the peak amplitude did not reach the same magnitudes at the other intervals. This interval coincides with the intensification of the EE and also the penetration of the WE electrojet to Fennoscandia (see Figure 9). This is consistent with the results reported by Pulkkinen et al. (2003). The large GICs did not result from the gradual increase of the EE but small features embedded within the larger structure. An unanswered question is the source of the small spatial scale EE features (Figure 6, 14:00–15:00). The three successive enhancements drive the large amplitude dBx/dt and >15-A GIC. This feature tends to agree with Pulkkinen et al. (2003) in that significant contribution to GICs originates from smaller-scale spatiotemporal structures superimposed on the large-scale electrojet. As previous studies have shown (Anderson & Fuselier, 1994), this reiterates that the EE is also important to driving large GICs. It is particularly worth noting that this interval also appeared to be part of a larger global response (Clilverd et al., 2018).
According to the Wp index shown in Figure 1 of Clilverd et al. (2018), a substorm also occurred on 8 September at approximately 18:00 UT, but those authors did not record particularly enhanced GICs. In our data (see Figure 10), the response to this substorm was quite localized and the effects varied both in latitude and longitude over Fennoscandia. We measured the largest peak GIC during this interval, although the GIC enhancement period was shorter. If a substorm was responsible for this GIC interval, then the trigger is unclear since the upstream period exhibited little transient signatures. Thus, this may have been internally driven. What was apparent from the IEQ shown in Figure 9 was that we observed rotational and localized features during this time. Such dynamic behavior could shed some light on the recent work by Ngwira et al. (2018) who reported that substorms can drive localized dB/dt. It should also be noted that substantial geomagnetic variations can occur during any part of the geomagnetic storm but are more commonly associated with the sudden commencement as opposed to the minimum of the main phase. What was notable here was that the largest GIC was not associated with the sudden commencement when the largest amplitude geomagnetic perturbations took place. This highlights the need for a complete understanding of the entire cycle of complex events and their various storm phases.
5.2 Models
We also modeled the surface geoelectric field over Fennoscandia for a uniform source current to highlight the enhancements over the Fennoscandia region resulting from the complex ground conductivity profiles. We can conclude from this that Fennoscandia contains many localized geological features which yield a strong influence on the amplitudes of the surface geoelectric field. Honkonen et al. (2018) pointed out that in areas of significant lateral conductivity gradients, differences between 1-D and 3-D models can be large and that coarse global conductivity models do not necessarily contain the required resolution to capture such features. However, it is important to remember that this influence is strongly dependent on locations where there are large conductivity gradients. It was interesting that in our case, the temporal features of the 1-D and 3-D models agreed very well. This means that both 1-D and 3-D models were able to capture the complicated transient features of the geoelectric field and GICs. It is logical to infer from this that the source of these dynamics were the ionosphere currents and not due to local inhomogeneous ground conductivity features. Based on this evidence, it appears that a 1-D model at MAN is justified and yields adequate performance for GIC modeling purposes. It must be stressed that this is only applicable at this location, and similar efforts would need to be undertaken in other key regions. We should also address the difference in amplitudes between SMAP and the Viljanen, Pulkkinen, et al. (2006) model. We deduced that this was entirely due to the difference in conductivities used, and not the result of any 1-D and 3-D effects. In spite of this, there is sufficient evidence to suggest that separate 1-D and 3-D models will likely provide similar results in this region, provided that the conductivity values between the models are consistent. This latter point was tested by adjusting the SMAP conductivity values in the lower layer.
It will be important to assess in what locations 3-D models are required. This will undoubtedly be significant in some regions of Fennoscandia where there are numerous complex conductivity features. In addition to this, the localized picture is further complicated by the dynamic nature of the source currents driven by magnetosphere-ionosphere coupling mechanisms such as substorms. In future work, we will combine the equivalent currents and SMAP model to determine the geoelectric field resulting from spatially varying current sources. Although all of the applied models agreed from a temporal standpoint, we did observe a large amplitude difference between SMAP and the Viljanen, Pulkkinen, et al. (2006) model which was entirely due to the values of ground conductivity used. In the context of model validation, it will be important to perform additional validation of the various models in this region due to these discrepancies in conductivity values. Regardless, we will devote attention to this in the near future. What is clear is that if one can accurately predict the ground magnetic field (or logically the ionospheric currents), then we can feasibly approach a reasonable approximation to the geoelectric field and GIC, provided that accurate models of the ground and relevant ground networks are available.
An important aspect in these studies is to consider the orientation of the geoelectric field with respect to the pipeline (Ingham & Rodger, 2018). In our case, only the parallel component of the geoelectric field drives GIC along the pipeline. In a general case, one would spatially integrate along the pipeline and then the parallel contributions would drive the GIC (Pulkkinen et al., 2001). Therefore, based purely on observations at MAN, over a long system it is difficult to determine precisely what contributions drive large GIC in the case where very localized (100 km) source currents are present. However, in this study, the sources were comparable to the effective scale of the pipeline of a few hundred kilometers. For this reason, we assumed a spatially uniform electric field of a fixed amplitude. For this simplified case, there are two opposite directions of the field that maximize the modulus of GIC and are defined by the coefficients a and b in the expressions
and
, as shown by ArajäRvi et al. (2011). In our case, α is approximately 51° or 231° when measured clockwise from the north direction. The geometry of the pipeline in Figure 1b shows a long section close to eastward whose GIC is driven by Ey, whereas the long section to the northwest is driven by Ex. Thus, as expected, a and b are approximately equal in magnitude. It is also worth noting that there are shorter branches to the south and southwest which are affected by Ex.
5.3 Future Implications
In this event, it is logical to conclude that it was not one single magnetospheric process which resulted in such large amplitude GICs, but multiple processes (e.g., global storm-time convection and substorms) capable of impacting multiple MLT sectors (midnight and afternoon). Substorms are a well-established magnetospheric energy release mechanism (Akasofu, 1964) and have been shown to coincide with enhanced dB/dt (Viljanen et al., 2006) and thus are important to GICs. Having said that, it is not clear which processes were dominant at different times—this will require further study. One definite conclusion is that each interval was associated with a clear southward IMF and subsequent energy loading into the magnetosphere, but as we observed from interval [3], the trigger for energy release is not always obvious from upstream data. However, in the context of this event, the strength and duration of −Bz does not explain the absolute GIC amplitudes, and therefore other factors need to be considered, particularly in terms of key magnetosphere-ionosphere coupling processes. What is particularly fascinating is that the largest peak GIC was not during the largest amplitude magnetic depressions/enhancements; in the sense that substantially larger geomagnetic depressions were recorded in interval [1] and in fact, the EE was also stronger in interval [2] at NUR. Thus, one would not have expected the large GIC to occur in interval [3]. It is a valuable insight into this event and data set that extreme GICs do not always have to accompany the largest ionospheric currents and magnetic disturbances. Nevertheless, activity must be high in terms of dB/dt, but the relationship between dB/dt and geoelectric field in terms of amplitude, duration, and the combination of the individual components is far from straightforward. Thus, the fine structures of geomagnetic variations which drive GICs prove to be extremely difficult to predict, and more work is needed before individual events can be predicted with reasonable accuracy. This point also demonstrates further that global geomagnetic indices are not ideal metrics to determine the occurrence or amplitude of GICs, and improved metrics will be required.
It is important to interpret scientific results in context with the engineering topology of the ground-based technological infrastructure. From our data, there were clear instances when the geoelectric field demonstrated preferential orientations with respect to the pipeline that subsequently drove large GIC. For example, large northward and eastward directed geoelectric fields are seen in Figure 12 which are closely aligned with long sections of the pipeline as shown in Figure 1b. Therefore, although the magnitude of the ground disturbance is important, it is also necessary to determine the orientation of the subsequent geoelectric field. It will be vital when assessing the vulnerability of ground-based systems to GIC to determine the key physical processes that enhance geoelectric fields parallel to long running sections.
Localized high-amplitude dB/dt and geoelectric field peaks are crucial to GICs (Ngwira et al., 2015, 2018), and we see some evidence of similar disturbances here. However, in our case it is logical to conclude that any localized variations were of a spatial scale comparable to the pipeline. This is based on the result in which the agreement between the measured GIC and that derived using a spatially uniform geoelectric field were favorable. Nevertheless, the difficulty of predicting and understanding the fine structure of GICs are intrinsically linked to their regional and localized nature. The subsequent impact on technological infrastructure from localized effects remains unclear, particularly whether highly structured events such as this present a greater risk. Unfortunately, it is unlikely that operational models based on MHD physics will perform well for these complex events, and therefore it will be important to perform rigorous model-data comparisons of the magnetospheric and ground dynamics during ICMEs of varying complexity.
6 Conclusion and Summary
- The peak amplitude GIC occurred during a surprising interval in which the absolute amplitude of geomagnetic disturbances were weaker than previous intervals, no upstream trigger was visible, and the preceding upstream driving criteria were not particularly strong.
- Uncommonly large GICs (>10 A) were distributed across multiple MLT sectors and were driven by small-scale spatiotemporal structures superimposed onto the EE and WE.
- Although noise disturbs the GIC, in some cases, temporal variations in the GIC measurements appear to correspond to those observed in Bx.
- The comparable results of 1-D and 3-D models demonstrates the lack of significant lateral ground conductivity gradients at Mäntsälä and therefore justifies the past, present, and future uses of 1-D models at that location.
- Provided that accurate ground disturbances are obtained at Mäntsälä, the available models are able to reproduce the GIC during complex and strong events such as the 7–8 September event and thus is applicable to future events.
- Although the temporal behaviors were in good agreement, we measured large differences in amplitudes between a local 1-D and the SMAP 3-D models at Mäntsälä which demonstrates the need from the community for accurate and high-resolution 3-D ground conductivity models.
To summarize, many studies have shown the complexity of extreme events (Kappenman, 2003; Pulkkinen et al., 2003; Viljanen et al., 2006) in the context of GICs and the high degree of difficulty in its prediction. To add to this, our analysis demonstrates that the largest GIC during an event does not necessarily occur during the periods when the largest geomagnetic disturbances are observed and during the most expected upstream driving conditions. For complex events such as this, more work is needed to understand the precursor of large sudden GIC peaks since conditions other than the duration and absolute value of −Bz likely play a strong role, as was the case here. Identifying the key magnetosphere-ionosphere coupling mechanisms to GICs responsible for their localized nature will also be critical to the improvement of operational services and the development of mitigation strategies. The September 2017 space weather events were the result of a highly structured ICME transient, and it will be important to study additional events of varying complexity to fully understand the nature of the key M-I coupling mechanisms and their associated ground effects.
Acknowledgments
This study was initiated at a workshop hosted by the Finnish Meteorological Institute in March 2018. We acknowledge all who participated in that workshop for their fruitful discussions. We thank the institutes who maintain the IMAGE Magnetometer Array: Tromsø Geophysical Observatory of UiT the Arctic University of Norway (Norway), Finnish Meteorological Institute (Finland), Institute of Geophysics Polish Academy of Sciences (Poland), GFZ German Research Centre for Geosciences (Germany), Geological Survey of Sweden (Sweden), Swedish Institute of Space Physics (Sweden), Sodankylä Geophysical Observatory of the University of Oulu (Finland), and Polar Geophysical Institute (Russia). FMI acknowledges the Gasum company for a long-term collaboration in GIC studies of the Finnish natural gas pipeline. We acknowledge use of NASA/GSFC's Space Physics Data Facility's OMNIWeb service, and OMNI data. IRFU acknowledges support from the Swedish Civil Contingencies Agency grant 2016-2102. FMI authors acknowledge support from the Academy of Finland grant 314670. The IMAGE data can be obtained free of charge at the website (http://space.fmi.fi/image). OMNI data are available via the NASA OMNIWeb service at the website (https://omniweb.gsfc.nasa.gov/). Equivalent current plots can be found at http://www.space.fmi.fi/MIRACLE/iono.html along with details of the derivation. GIC data can be viewed at the following address http://space.fmi.fi/gic/ and dowloaded from the website (http://space.fmi.fi/gic/man_ascii/man.php). The SMAP model conductivity profiles of Fennoscandia and their derivation can be found in the following manuscripts: Korja et al. (2002) and Engels et al. (2002).
