Abstract
Ocean tide generated magnetic fields contain information about changes in ocean heat content and transport that can potentially be retrieved from remotely sensed magnetic data. To provide an important baseline towards developing this potential, tidal signals are extracted from 288 land geomagnetic observatory records having observations within the 50-year time span 1965-2015. The extraction method uses robust iteratively reweighted least squares for a range of models using different predictant and predictor assumptions. The predictants are the time series of the three vector components at each observatory, with versional variations in data selection and processing. The predictors fall into two categories: one using time-harmonic bases and the other that directly use lunar and solar ephemerides with gravitational theory to describe the tidal forces. The ephemerides predictors are shown to perform better (fitting more variance with fewer predictors) than do the time-harmonic predictors, which include the traditional 'Chapman-Miller method'. In fitting the oceanic lunar tidal signals, the predictants with the highest signal/noise involve the 'vertical' magnetic vector component following principle-component rotation. The best simple semidiurnal predictor is the ephemeris series of lunar azimuth weighted by the inverse-cubed lunar distance. More variance is fitted with predictors representing the lunar tidal potential and gradients calculated for each location/time.This article is part of the theme issue 'Magnetometric remote sensing of Earth and planetary oceans'.