Global oceanic warming might become observable from space by analyzing tidal magnetic signals with feed-forward artificial neural networks

M2 tidal magnetic field amplitudes at satellite height (430km over sea level).

The Geoforschungszentrum Potsdam (GFZ) has a long history in geomagnetic research. Over the last years, GFZ researchers were able to identify the sensitivity of tidal magnetic fields to oceanic climate change (e.g., Saynisch et al., 2016, 2017). Tidal magnetic fields are generated by the movement of saline water through Earth's background magnetic field. The signals emerge from the oceans and are detectable by satellite magnetometers, e.g. the recent ESA mission Swarm. GFZ scientists found that variations in oceanic salinity and oceanic warming can lead to significant changes in the oceanic tidal signals.

By calculating monthly M2 tidal magnetic maps and global ocean heat content from various oceanographic observation-based state estimates over the period of 1990–2015, a database for teaching an artificial neural network was generated, which includes also the errors of possible observations. In a recently published study, a feed-forward neural network, well trained with this database, was able to estimate global oceanic heat content from tidal magnetic signals only. The neural network successfully estimates even those parts of the database which were omitted during the learning phase. Finally, the network was applied to interpret real satellite observations of the magnetic M2 signal (Sabaka et al. 2018). The corresponding heat content estimates of the neural network match well with oceanographic based estimates over the same time period.

Irrgang, C., Saynisch, J., Thomas, M. (2019): Estimating global ocean heat content from tidal magnetic satellite observations. Scientific reports 9, no. 7893,