Reconstruction of Storm-Time Total Electron Content Using Ionospheric Tomography and Artificial Neural Networks: A Comparative Study Over the African Region

Abstract

The work presented here aims to evaluate the capabilities of Multi-Instrument Data Analysis System (MIDAS) compared with artificial neural networks (ANNs) to reconstruct storm-time total electron content (TEC) over the African low-latitude and midlatitude regions. For MIDAS, the inversion was done based on the Global Positioning System (GPS) measurements from receiver stations extending from −30∘ to 36∘ in latitude and 30∘ to 44∘ in longitude while for ANNs, individual storm-time models based on historical GPS data from receivers within the same region covered by MIDAS were used. Based on the minimum Dst index reached during the storm period, moderate (−50 nT ⩾ Dst > −100 nT), strong (−100 nT ⩾ Dst > −200 nT), and severe (−200 nT ⩾ Dst > −350 nT) storms were used for validation. MIDAS and ANNs results were compared with IRI-2016 predictions and validated with real GPS TEC observations. A statistical analysis revealed that MIDAS and ANNs provide comparable results in storm-time TEC reconstruction with average mean absolute errors of 4.81 and 4.18 TECU respectively. However, MIDAS performed better compared to ANNs in following TEC enhancements and depletions as well as short-term features observed during the selected storm periods. In terms of latitude, it was found that on average, MIDAS performs 13% better than ANNs in the African midlatitude, while ANN model performs 24% better than MIDAS in low latitudes. Furthermore, comparisons with IRI predictions showed that both MIDAS and ANNs produce more accurate estimations of the storm-time TEC than IRI model.