Machine Learning for solving the Bz Problem in Space Weather Forecasting

Our current inability to forecast the Bz magnetic field embedded within interplanetary coronal mass ejections (ICMEs) - the Bz problem - is one of the most critical challenges in predicting severe space weather effects at Earth. Here we shine new light on the Bz problem by investigating the hypothesis that upstream in situ measurements of the ICME sheath region and the first few hours of the magnetic obstacle provide sufficient information for predicting the downstream Bz component. To this end, we use data science methods and train them on 348 ICMEs observed in Wind, STEREO-A, and STEREO-B mission data extracted from the open-source ICMECATv2.0 catalog. Since the application of our approach in an operational setting needs a reliable automated ICME identification functionality, we present an automated ICME identification method and quantitatively assess its robustness. In this contribution, I will show the predictive ability expected from the developed Bz prediction pipeline, discuss the lessons learned from using data science, and outline future strategies for predicting the effects of ICMES on our planet Earth.