Item

Abstract

The forecast of the International Geomagnetic Reference Field, `IGRF SV', passed half of the update period of five years, and the prominent model of the Earth's magnetic field, `CHAOS', is updated in somewhat of a routine fashion, meanwhile a few times a year.&nbsp; The `CHAOS' model uses for its short forecasts (delivered separately from product table in the format used by the Swarm satellite mission) just a simple extrapolation. Which simple numerical methods are appropriate or at least sufficient for such forecasts?<br><br>Some basic fast, numeric or statistical approaches (as simple linear or polynomial fits and autoregressive forecasting) are compared by the error they give in the light of the available updates.&nbsp; The retrospective errors for delivered products (i.e. the Swarm core field auxiliary product, 'AUX_COR') and from dedicated numerical experiments are assessed and summarized.&nbsp; This may (or may not) create hints for the required update frequency, if not for the most suitable method. Also, extrapolation tests can be useful to get an idea about the uncertainty of predictions for (non-physics-based) parent models of candidates for the incoming `IGRF' - and to outline an acceptable strategy.