Automatic detection of the Earth's magnetopause boundary layer from multiple mission data

The Earth’s magnetopause boundary layer (BL) is a dynamic plasma region shaped by the interaction between the solar wind and the magnetosphere. Its properties exhibit strong spatial and temporal variability, making the BL a complex and heterogeneous structure. These characteristics have hindered systematic BL identification and large-scale statistical analysis. In this paper, we present a novel supervised machine learning approach for the automated detection of the magnetopause BL using data from the MMS, THEMIS, and Cluster missions. Our method achieves a BL detection with a precision constrained primarily by the inherent ambiguity in the labeling task. This approach enables, for the first time, the construction of a comprehensive BL dataset comprising over 5 million data points, offering unprecedented spatial coverage compared to previous studies. This repository gives access to two datasets: The BL detected from multiple missions by a HistGradientBoostingClassifier, with the methodology detailed in the paper [in prep]. The cold ions dataset from multiple missions. To detect their presence, we fit a gaussian on the energy flux of the lower 12 channels ($< 100$ eV) of the ion omnidirectional spectrogram. When a good fit is found and the center of the obtained gaussian is strictly in the 0-100 eV energy range, cold ions are considered present. The height, center and width of the fitted gaussian are called max_coldions, center_coldions and std_coldions, and are respectively set to 0, -1 and -1 when no fit is found. These datasets can be read with the pandas=='2.3.3' and numpy=='2.2.6' version. All the details are provided in the paper [in prep].