Advancing Ionospheric Predictions in North Africa: A Deep Learning Approach Integrating Ground and Satellite GNSS observations

The North Africa region faces significant challenges due to the need for ionospheric ground observational data, complicating efforts to achieve accurate predictions. Our study strategically integrates ground and satellite-derived ionospheric data from Global Navigation Satellite System (GNSS) observations to address this pivotal gap. Leveraging this extensive dataset, we implemented a sophisticated Deep Neural Network (DNN) methodology, resulting in a notable advancement in predicting ionospheric irregularities stemming from space weather phenomena. Our approach meticulously utilized two years of Vertical Total Electron Content (VTEC) data sourced from 12 strategically located GNSS ground stations that span a geographical range from 0°N to 40°N latitudes and 25°W to 50°E longitudes. This focused dataset aimed to enhance the accuracy and reliability of predictions tailored explicitly for the North African context. Moreover, we seamlessly integrated critical geomagnetic indices—including Dst, F10.7, and KP— to bolster our predictive capabilities with the GNSS-derived VTEC data. This intricate fusion facilitated the training of our DNN-LSTM model on robust time-series datasets, empowering it to deliver forecasts characterized by spatial precision and temporal relevance. Notably, when comparing DNN prediction to the IRI2020 model, our DNN-based approach showcased higher accuracy, evidenced by reduced RMSE values. Our research illuminates the immense potential of integrating ground and space-based data with machine learning paradigms. Such integrative methodologies hold promise for advancing ionospheric studies, especially in regions characterized by sparse ground station coverage, thereby enriching our comprehension of ionospheric dynamics.