Accepted author manuscript, 6.99 MB, PDF document
Available under license: CC BY: Creative Commons Attribution 4.0 International License
Final published version
Licence: CC BY: Creative Commons Attribution 4.0 International License
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Article number | e2021SW002903 |
---|---|
<mark>Journal publication date</mark> | 31/01/2022 |
<mark>Journal</mark> | Space Weather |
Issue number | 1 |
Volume | 20 |
Publication Status | Published |
Early online date | 11/11/21 |
<mark>Original language</mark> | English |
Geomagnetically induced currents (GICs) are an impact of space weather that can occur during periods of enhanced geomagnetic activity. GICs can enter into electrical power grids through earthed conductors, potentially causing network collapse through voltage instability or damaging transformers. It would be beneficial for power grid operators to have a forecast of GICs that could inform decision making on mitigating action. Long lead-time GIC forecasting requires magnetospheric models as drivers of geoelectric field models. However, estimation of the geoelectric field is sensitive to high-frequency geomagnetic field variations, which operational global magneto-hydrodynamic models do not fully capture. Furthermore, an assessment of GIC forecast uncertainty would require a large ensemble of magnetospheric runs, which is computationally expensive. One solution that is widely used in climate science is “downscaling,” wherein sub-grid variations are added to model outputs on a statistical basis. We present proof-of-concept results for a method that temporally downscales low-resolution magnetic field data on a 1-hr timescale to 1-min resolution, with the hope of improving subsequent geoelectric field magnitude estimates. An analog ensemble (AnEn) approach is used to select similar hourly averages in a historical data set, from which we separate the high-resolution perturbations to add to the hourly average values. We find that AnEn outperforms the benchmark linear-interpolation approach in its ability to accurately drive an impacts model, suggesting GIC forecasting would be improved. We evaluated the ability of AnEn to predict extreme events using the FSS, HSS, cost/loss analysis and BSS, finding that AnEn outperforms the “do-nothing” approach.