GOCE ML-calibrated magnetic field data

Cite as:

Styp-Rekowski, Kevin; Michaelis, Ingo; Stolle, Claudia; Baerenzung, Julien; Korte, Monika; Kao, Odej (2022): GOCE ML-calibrated magnetic field data. V. 0204. GFZ Data Services. https://doi.org/10.5880/GFZ.2.3.2022.002

Status

I N R E V I E W : Styp-Rekowski, Kevin; Michaelis, Ingo; Stolle, Claudia; Baerenzung, Julien; Korte, Monika; Kao, Odej (2022): GOCE ML-calibrated magnetic field data. V. 0204. GFZ Data Services. https://doi.org/10.5880/GFZ.2.3.2022.002

Abstract

The Gravity field and steady-state ocean circulation explorer (GOCE) satellite mission carries three platform magnetometers. After careful calibration, the data acquired through these can be used for scientific purposes by removing artificial disturbances from other satellite payload systems. This dataset is based on the dataset provided by Michaelis and Korte (2022) and uses a similar format. The platform magnetometer data has been calibrated against CHAOS7 magnetic field model predic-tions for core, crustal and large-scale magnetospheric field (Finlay et al., 2020) and is provided in the ‘chaos’ folder. The calibration results using a Machine Learning approach are provided in the ‘calcorr’ folder. Michaelis’ dataset can be used as an extension to this dataset for additional infor-mation, as they are connected using the same timestamps to match and relate the same data points. The exact approach based on Machine Learning is described in the referenced publication.

The data is provided in NASA CDF format (https://cdf.gsfc.nasa.gov/) and accessible at: ftp://isdcftp.gfz-potsdam.de/platmag/MAGNETIC_FIELD/GOCE/ML/v0204/ and further de-scribed in a README.

Additional Information

21 September 2022: addition of the key reference (Styp-Rekowski et al., 2022) to the DOI landing page, data description (PDF) and README.

Methods

The data was recorded onboard the GOCE satellite mission with varying time intervals of the differ-ent subsystems measuring. The magnetometer measurements (16s intervals) were aligned to match the closest position measurement (1s intervals) and interpolated accordingly. All other avail-able data of different intervals was interpolated and aligned to the same timestamps.

The data was calibrated using a Machine Learning approach involving Neural Networks, the whole method of calibration is described precisely in the referenced publication. The data was mainly processed for its calibration which yields a lower residual compared to a refer-ence model than the uncalibrated data, more details about the many steps involved can be found in the referenced publication.

Authors

Styp-Rekowski, Kevin;TU Berlin, Berlin, Germany;GFZ German Research Centre for Geosciences, Potsdam. Germany
Michaelis, Ingo;GFZ German Research Centre for Geosciences, Potsdam. Germany
Stolle, Claudia;IAP - Leibniz Institute of Atmospheric Physics at the University of Rostock, Kühlungsborn, Germany
Baerenzung, Julien;GFZ German Research Centre for Geosciences, Potsdam. Germany
Korte, Monika;GFZ German Research Centre for Geosciences, Potsdam. Germany
Kao, Odej;TU Berlin, Berlin, Germany