Monthly means of the magnetic field measurements taken by ground observatories are a
useful data source for studying temporal changes of the core magnetic field. However,
the usual way of calculating monthly means as the arithmetic mean of all days
(geomagnetic quiet as well as disturbed) and all local times (day and night) may result in
contributions of external (magnetospheric and ionospheric) origin in the monthly
means. Such a contamination makes monthly means less favorable for core field
studies.
We investigate this problem by calculating modified monthly means from observatory
hourly means using different statistical approaches (arithmetic mean, median, robust
mean, ...) and data selection criteria (all days, quiet days only, local night data only,
...). An assessment of the different approaches is done by means of generalized
cross-validation. |