Dartmouth Events

Abstract: In recent times, the adoption of data-intensive machine learning methods has increased tremendously in many fields, including space science. Here, we will present and discuss two examples demonstrating the application of deep neural networks in modeling the near-Earth geospace environment. First, we present a deep learning‐based approach to predict the onset of a magnetic substorm, defined as the signature of the auroral electrojets in ground magnetometer measurements. We use a time history of solar wind speed, proton number density, and interplanetary magnetic field (IMF) as inputs to forecast the occurrence probability of an onset over the next 1 hr. In the second application, we present a deep convolutional neural network for modeling the global distribution of the Birkeland field-aligned current system using a 1‐hr time history of IMF, solar wind, and geomagnetic and solar indices. We will analyze the performance of these models under different geomagnetic conditions, and discuss the importance of domain knowledge, and the datasets used for training the model.