Forecasting Geomagnetic Storm Disturbances and Their Uncertainties Using Deep Learning.

Severe space weather produced by disturbed conditions on the Sun results in harmful effects both for humans in space and in high‐latitude flights, and for technological systems such as spacecraft or communications. Also, geomagnetically induced currents (GICs) flowing on long ground‐based conductors, such as power networks, potentially threaten critical infrastructures on Earth. The first step in developing an alarm system against GICs is to forecast them. This is a challenging task given the highly non‐linear dependencies of the response of the magnetosphere to these perturbations. In the last few years, modern machine‐learning models have shown to be very good at predicting magnetic activity indices. However, such complex models are on the one hand difficult to tune, and on the other hand they are known to bring along potentially large prediction uncertainties which are generally difficult to estimate. In this work we aim at predicting the SYM‐H index characterizing geomagnetic storms multiple‐hour ahead, using public interplanetary magnetic field (IMF) data from the Sun‐Earth L1 Lagrange point and SYM‐H data. We implement a type of machine‐learning model called long short‐term memory (LSTM) network. Our scope is to estimate the prediction uncertainties coming from a deep‐learning model in the context of forecasting the SYM‐H index. These uncertainties will be essential to set reliable alarm thresholds. The resulting uncertainties turn out to be sizable at the critical stages of the geomagnetic storms. Our methodology includes as well an efficient optimization of important hyper‐parameters of the LSTM network and robustness tests. Plain Language Summary: Geomagnetic storms are disturbances of the geomagnetic field caused by interactions between the solar wind and particle populations mainly in the Earth's magnetosphere. These time‐varying magnetic fields induce electrical currents on long ground‐based conductors that can damage power transmission grids and other critical infrastructures on Earth. As a first step to forecast the ground magnetic perturbations caused by geomagnetic storms at specific mid‐latitude locations, the objective of this work is to predict the SYM‐H activity index, which is generated from ground observations of the geomagnetic field at low and mid‐latitudes, and which provides a measure of the strength and duration of geomagnetic storms. We use the IMF data measured by the Advanced Composition Explorer spacecraft at the L1 Lagrangian point and SYM‐H values to forecast the behavior and severity of geomagnetic storms multiple‐hour ahead. This forecasting is done using a type of artificial neural network model called long short‐term memory. We also propose robust ways to estimate the uncertainties of these predictions, which help us to better understand machine‐learning models in activity indices prediction and lead to more accurate and reliable forecasting of geomagnetic storms and their ground effects in the near future. Key Points: A long short‐term memory (LSTM) model is built to forecast SYM‐H index multiple‐hour ahead using interplanetary magnetic field (IMF) measurements and SYM‐H valuesPrediction uncertainties from the LSTM model are estimated and turn out to be considerable in the critical phases of geomagnetic stormsThe uncertainty quantification is found to be crucial to achieve a reliable forecasting model and determine the optimal look‐forward [ABSTRACT FROM AUTHOR]