THE TIME-AVERAGED GEOMAGNETIC-FIELD - GLOBAL AND REGIONAL BIASES FOR 0-5 MA

Palaeodirectional data from lava flows and marine sediments provide in formation about the long-term structure and variability in the geomagn etic held. We present a detailed analysis of the internal consistency and reliability of global compilations of sediment and lava-flow data. Time-averaged field models are constructed for normal and reverse pol arity periods for the past 5 Ma, using the combined data sets. Non-zon al models are required to satisfy the lava-flow data, but not those fr om sediments alone. This is in part because the sediment data are much noisier than those from lavas, but is also a consequence of the site distributions and the way that inclination data sample the geomagnetic field generated in the Earth's core. Different average held configura tions for normal and reverse polarity periods are consistent with the palaeomagnetic directions; however, the differences are insignificant relative to the uncertainty in the average field models. Thus previous inferences of non-antipodal normal and reverse polarity field geometr ies will need to be re-examined using recently collected high-quality palaeomagnetic data. Our new models indicate that current global sedim ent and lava-flow data sets combined do not permit the unambiguous det ection of northern hemisphere flux lobes in the 0-5 Ma time-averaged f ield, highlighting the need for the collection of additional high-lati tude palaeomagnetic data. Anomalous time-averaged held structure is se en in the Pacific hemisphere centred just south of Hawaii. The locatio n of the anomaly coincides with heterogeneities in the lower mantle in ferred from seismological data. The seismic observations can be partly explained by lateral temperature variations; however, they also sugge st the presence of lateral compositional variations and/or the presenc e of partial melt. The role of such heterogeneities in influencing the geomagnetic held observed at the Earth's surface remains an unresolve d issue, requiring higher-resolution time-averaged geomagnetic field m odels, along with the integration of future results from seismology, m ineral physics and numerical simulations.