Time series analysis of the 10 000 yr geomagnetic secular variation record from SE Australia

Summary. A 10 000 yr continuous secular variation record from intensively dated lake sediments in SE Australia has been subjected to periodogram and maximum entropy method analysis. Tests on synthetic data reveal some of the limitations of the latter method, particularly when applied to complex number series. Anticlockwise precession of the magnetic vector at a period of 5000 ± 1000 yr is tentatively ascribed to dipole precession, and clockwise precession at a period of about 2800 yr is probably due to westward drift of features of the non-dipole field.
The effect of calibrating the radiocarbon time-scale is important and results in periodicity shifts of up to 25 per cent. Even for well-dated lacustrine sequences power spectra are poorly constrained: it is thus possible that the geomagnetic secular variation on a time-scale of thousands of years is more uniform than often supposed. Mismatches between declination and inclination spectra can arise as a natural consequence of certain types of source mechanism and should not be simply attributed to figments of the analysis employed.