Automatic Detection of Interplanetary Coronal Mass Ejections from In Situ Data: A Deep Learning Approach

Abstract

Decades of studies have suggested several criteria to detect Interplanetary coronal mass ejections (ICME) in time series from in-situ spacecraft measurements. Among them the most common are an enhanced and smoothly rotating magnetic field, a low proton temperature and a low plasma beta. However, these features are not all observed for each ICME due to their strong variability. Visual detection is time-consuming and biased by the observer interpretation leading to non exhaustive, subjective and thus hardly reproducible catalogs. Using convolutional neural networks on sliding windows and peak detection, we provide a fast, automatic and multi-scale detection of ICMEs. The method has been tested on the in-situ data from WIND between 1997 and 2015 and on the 657 ICMEs that were recorded during this period. The method offers an unambiguous visual proxy of ICMEs that gives an interpretation of the data similar to what an expert observer would give. We found at a maximum 197 of the 232 ICMEs of the 2010-2015 period (recall 84 ± 4.5%) including 90% of the ICMEs present in the lists of Nieves-Chinchilla et al. (2018) and Chi et al. (2016). The minimal number of False Positives was 25 out of 158 predicted ICMEs (precision 84 ± 2.6%). Although less accurate, the method also works with one or several missing input parameters. The method has the advantage of improving its performance by just increasing the amount of input data. The generality of the method paves the way for automatic detection of many different event signatures in spacecraft in-situ measurements.

Deep-SWIM: A few-shot learning approach to classify Solar WInd Magnetic field structures

“…For our supervised approach we assign binary labels to each segment specifying whether it contains a discontinuity or not. The labels are provided by expert hand-labeling for one day (11-18-2018) or using a non-machine learning based heuristic approach which compares the rotation angle of neighboring points to identify discontinuities (2006-2021) 2 . Importantly, not all labels in the heuristic catalog are expected to be correct.…”

“…At the present moment, state-of-the-art applications of deep learning to solar wind measurements are primarily confined to supervised classification of large-scale structures like interplanetary coronal mass ejections (ICME) [2,1]. The primary driver behind this is the occurrence rate of these structure.…”

“…While some magnetic events, e.g. interplanetary coronal mass ejections, are high-strength and therefore easy to detect [1,2], smaller-scale events, like discontinuities, are more difficult to spot. Discontinuities describe "discontinuous" spatial changes in the solar magnetic field.…”

“…The data is publicly available on NASA's Coordinated Data Analysis Web at https://cdaweb.gsfc.nasa.gov/2 The hand-labeled and heuristic catalog were provided by Dr. Adam Szabo and Dr. Ayris Narock who are both authors of this work.…”

Identification of Electron Diffusion Regions with a Machine Learning Approach on MMS Data at the Earth's Magnetopause

“…Machine learning has proven to be an incredibly powerful tool to process large amounts of data, such as provided by the MMS. The interest of the scientific community toward the field of big data and the machine learning techniques has been growing for the last few years with applications like automatic detection and the classification of astrophysical events among others (e.g., Argall et al., 2020; Nguyen et al., 2019). Machine Learning algorithms can be split into two different categories of learning: the first one is supervised, when an algorithm learns by iteratively minimizing its prediction error after comparison with existing labels.…”