the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Benchmarking the Swedish Power Grid Against a 1-in-100-Year Geoelectric Field Scenario
Abstract. Sweden's communication and power systems have been impacted by extreme space weather events in the past. For instance, the May 1921 storm caused a fire at the telegraph and telephone station in Karlstad, and the 2003 Halloween storm led to a blackout in Malmö. In this study, we present the first comprehensive assessment of the potential impacts of a 1-in-100year event on the entire Swedish power grid. Using magnetic field observations from the 30 October 2003 event as a baseline, we constructed two extreme scenarios. In Case 1, we used the observed magnetic field across Fennoscandia. In Case 2, we assume a spatially uniform ionospheric current system, producing identical magnetic waveforms across the country. Then the estimated 3D electric field was scaled using region-specific scaling factors derived from recent statistical analyses of electric field extremes in Sweden. The scaled geoelectric field and power lines voltages are computed using the recently developed RAISE model, which includes realistic ground conductivity and power line topology. Our results show that the strongest horizontal electric fields, around 12 V/km, occur within the 55° and 58° MLAT band, particularly in regions with sharp lateral conductivity gradients. East–west-oriented power lines are especially vulnerable, as they align with the dominant orientation of the induced electric field. Overall, during the peak of a 1-in-100-year geomagnetic storm, more than 100 transmission lines are expected to experience voltages above 50 V multiple times over the course of the substorm. At the peak of the strongest disturbance, triggered by a sudden weakening of the westward electrojet, around 100 lines are expected to exceed 100 V. These results provide critical insights into infrastructure vulnerability under extreme space weather.
- Preprint
(17049 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 29 Jan 2026)
- RC1: 'Comment on egusphere-2025-4900', Anonymous Referee #1, 03 Dec 2025 reply
-
RC2: 'Comment on egusphere-2025-4900', Anonymous Referee #2, 08 Jan 2026
reply
This is an interesting paper that studies the effects of two possible 1-in-100 years geoelectric field events on the Swedish power grid using the most advanced tools available in Sweden. However, the approach to create the 1-in-100 years geoelectric fields is not convincing at present and needs some reconsideration. The topic is worthwhile, and with some more work this should make a valuable addition to the series of Swedish publications that characterize geomagnetic activity and the power grid response in Sweden.
Major comment:
You create two possible 1-in-100 years events, Case 1 and Case 2, based on magnetic field observations during the Halloween storm. In Case 1, you calculate the geoelectric field based on the magnetic field observations and then scale the resulting values upward in southern Sweden. In Case 2, you assume spatially uniform external magnetic field with an amplitude and time development based on one magnetometer station during storm. The resulting geoelectric field is again scaled similar to Case 1.
I find this approach problematic. Clearly, the reason for the characteristic latitude of the geoelectric field maximum, on which the scaling is based, is due to the external driver. In Case 2, you first create a spatially uniform external driver, but then change the latitude profile of the resulting geoelectric field such that it no longer corresponds to the driver. Nonetheless, you describe Case 2 to represent "an idealized scenario in which the magnetic field time series is spatially uniform" (lines 113-114). This is inconsistent. I suggest that either you drop the scaling or scale the external driver instead of the resulting geoelectric field. The latter also requires changing the way you describe the case.
In Case 1, you note that your analysis fails to explain the observed Malmö blackout. This you explain to be due to the missing magnetometer observations in the key area (lines 237-239). You also note that in northern Sweden the Halloween event is classified as a 1-in-100 years event whereas in southern Sweden it only reached 1-in-10 or 50 years (lines 108-110). Combining these, it rather sounds to me that the Halloween storm might well have been a 1-in-100 years event across Sweden, but the insufficient magnetometer coverage in southern Sweden did not allow full observation of its effects in this area. Thus, the role of the scaling in this case would not be to increase the classification of the storm but to rather correct for missing data. Such a correction should be applied to the driver rather than to the resulting geoelectric field. As this may not be straightforward, I suggest at least discussing the point in the text.
Minor comments:
Lines 11-12: "East–west-oriented power lines are especially vulnerable, as they align with the dominant orientation of the induced electric field." Do you mean this statement applies generally? Is the geoelectric field in Sweden typically oriented in the east-west direction? If not, please modify the statement.
Line 29: "most commonly" On what is this claim based? Kp is also very widely used.
Line 43: "(NERC, 2016)" What is NERC?
Line 63: "the paper of this paper" Delete "the paper of"?
Lines 67-68: "a once in 100 year event" Shouldn't "year" be "years"? Please check throughout the manuscript.
Lines 71-80: This section does not describe either data or methods. It could be moved to section 3.
Lines 75-76: What about differences due to 1D and 3D modelling? 3D modelling is expected to give higher geoelectric field amplitudes than 1D due to the effect of lateral conductivity gradients.
Line 86: "external (Bext) produced by", "internal (Bint ) produced by" Something is missing here. Do you mean external component and internal component?
Line 90: More detailed descriptions of GIC-SMAP and RAISE should be included. For example: What are the inputs and outputs of each model? What are the key principles and assumptions of these models? What are the key sources of uncertainty?
Line 102: Please check the style of references throughout the text.
Lines 104-105: "amplitude of the EH reaches an amplitude" Maybe write "amplitude of EH reaches a level" to avoid repetition?
Table 1: Should you add Lanabere et al. (2023, 2024) and Rosenqvist et al. (2025) as references to step 1 (cf. lines 71-80 and 108-110)?
Lines 118-120: Please explain this in more detail. Does GIC-SMAP expect Btot or Bext as input? I would have expected a first-principles model to take Bext as input and return Btot as output, but apparently this is not the case?
Figure 1: What are the arrows in panel (a)?
Figure 2: "specific sites" Are these the same locations as the stars in Fig. 1a?
Figure 3: "Magnetic field perturbations (dBH,ext /dt) at the magnetic stations along 25.42oE (solid line) and in Sweden (dashed line)." Please clarify this sentence. The stations in Finland and Estonia do not have a constant longitude. What are the arrows in panel (a)?
Line 193: "the induced electric field response is greater at HAN." Do you really mean that the induced electric field was calculated at HAN? (I understand this to mean the true location of HAN.) Or do you mean that the induced electric fields calculated using Bext from HAN at the locations marked with stars in Fig. 3a were greater than those calculated using Bext from OUJ?
Lines 194-197: I do not understand this reference. Dimmock et al. (2024, Figure 4) concerns a different event (23 Apr 2023) and they do not show the geoelectric field.
Line 211: "and the similarly response at the OUJ station" Please explain what this means.
Figure 5: What are the arrows?
Line 223: "3 V/km"
Line 225: "1 V/km" Why were these threshold values selected?Section 4: Due to the limitations of the power grid model you use, the analysis is limited to estimating voltages in separate transmission lines. Such an approach does not consider the power grid as an entity as proper GIC estimation would do, and hence does not provide a full picture of the response. The limitations of the chosen approach in this respect should be discussed. For example, the most intense GIC are typically observed at substations located at the edges of the power grid (https://doi.org/10.1051/swsc/2012017). According to your power grid map, both Malmö and Karlshamn appear to be located at the southern edge of the Swedish power grid. Even without exact values for the resistances, more useful results would probably be gained by estimating the GIC instead of analyzing the separate line voltages. Of course, in order to explain actual impacts, such as the Malmö blackout, you would also need to know the true grid geometry at the time of the event.
Lines 263-265: Is the number of power lines with the 100 V threshold exceeded really a relevant problem? It does not mean that there would be large GIC at the ends of all these power lines. Large GIC would be expected at the substations located at the edges of the power grid (see the point above).
Lines 345-346: "How do small-scale ionospheric currents during extreme storms influence the spatial variability of geoelectric fields and GICs?" What do you mean by small-scale? Due to the ~100 km distance between the ground and the ionosphere, magnetic field perturbations due to ionospheric current structures smaller than about 100 km in scale size are significantly attenuated before reaching the ground.
Lines 354-355: "The largest scaled EH are found within the 55◦ –58◦ MLAT band" This is a circular conclusion since you have artificially scaled the geoelectric field amplitude up particularly in this latitude band.
Line 356-357: "due to the prevailing orientation of the horizontal electric field" This conclusion sounds far more general than is probably intended or can be drawn based on you results. Please modify. (See also the similar comment on the abstract.)
Lines 365-366: "limited magnetometer coverage in southern Sweden" How accurate is the conductivity model in this area? Inaccuracies in the conductivity model can be a significant source of error for the modeled geoelectric field (https://doi.org/10.1029/2022SW003370).
Lines 368-370: This sentence is very long and difficult to follow. Please divide it into two or more shorter sentences. What is "sudden weakening of worst-case scenarios"?
Citation: https://doi.org/10.5194/egusphere-2025-4900-RC2
Video supplement
Benchmarking the Swedish Power Grid Against a 1-in-100-Year Geoelectric Field Scenario Vanina Lanabere https://doi.org/10.5446/71703
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 295 | 117 | 17 | 429 | 18 | 20 |
- HTML: 295
- PDF: 117
- XML: 17
- Total: 429
- BibTeX: 18
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors describe their work to complete the end to end modelling of GIC flow in the Swedish high voltage power grid using magnetic field measurements, the SMAP conductivity model and the RAISE representation of the power network. They investigate the October 2003 storm and then extend this to a 1-in-100 year storm using magnetic latitudinal bands of different scaling based on a recent statistical analysis of geoelectric fields.
They find the peak geoelectric field to lie between 10-30 V/km in some regions of Sweden. They use the model to identify lines that are expected to be more vulnerable to high current flow, though represented as total integrated voltage along particular lines. They discuss the implications and compare to regions where known outages have occurred before. They have generated two informative movies to illustrate the dynamic nature of the GIC flow.
This is an extremely well written paper and I could not find any issues with the logic, clarity or flow of the manuscript. The figures and tables are excellent and the results are useful for the operator. The authors are to be commended.
My only minor comments are:
1. why are the values of the line and transformer resistance under national security restrictions? They are available in other countries for example, so are probably similar in Sweden. Average values of resistance per km are given in e.g. Viljanen et al (2012) doi: 10.1051/swsc/2012017if you wanted to model GIC and use a standard 0.5 Ω for the grounding resistance.
2. Why not use Belsk, Brorfelde and Wingst observatories in the analysis to capture variations outside and to the south of Sweden in October 2003?