Using AI in Space Weather Forecast

In the last few years, we have witnessed the booming developments of artificial intelligence in all sectors. A White House Executive Order on “Maintaining American Leadership in Artificial Intelligence” requires that the Federal Government to “promote sustained investment in AI R&D in collaboration with industry, academia, international partners and allies, and other non-Federal entities to generate technological breakthroughs in AI and related technologies and to rapidly transition those breakthroughs into capabilities that contribute to our economic and national security.” ML is a revolutionary new methodology that has already delivered transformative advances in many fields of science and industry and will do the same in space sciences [e.g., Camporeale, 2019]. 

We have used AI to improve the ionospheric total electron content (TEC) specification and forecast [e.g., Liu et al. 2020; Sun et al. 2022; Sun et al., 2023; Wang et al., 2023]. In Liu et al., 2020, an LSTM model was constructed to forecast the spherical harmonics coefficients and then reconstruct the TEC map using those forecasted coefficients. Later, we developed the VISTA algorithm to reconstruct the global TEC maps [Sun et al., 2022] and then made the VISTA TEC dataset publicly available through University of Michigan’s DeepBlue data repository [Sun et al., 2023]. Based on the VISTA TEC dataset, a modified U-Net was developed to forecast the global TEC maps directly [Wang et al., 2023]. Besides ionospheric TEC project, we have also used AL to develop explainable SYM-H forecasting using Gradient Boosting Machines (GBM) method [Iong et al. 2023].