Paleomagnetic directions and K/Ar dating of 0 to 1 Ma lava flows from La Guadeloupe Island (French West Indias): Implications for time-averaged field models

Twenty-six lava flows spanning the last million years were sampled in La Guadeloupe, French West Indies. Because of the lack of continuous volcano-stratigraphic sections in La Guadeloupe, dating is necessary in order to describe the temporal evolution of the geomagnetic field in this time interval. New K/Ar ages ranging from 50 ka to 1 Ma have been obtained on andesites using the Cassignol-Gillot technique at the Universite Paris Sud Institut de Physique du Globe de Paris Orsay laboratory. Additional flows (with K/Ar ages obtained using the same dating technique) were also sampled for paleomagnetic investigations. More than 200 samples were analyzed using both alternating field AF and thermal stepwise demagnetization techniques. Duplicate samplings of two flows at three different sites demonstrate that within-flow dispersion is negligible for the andesitic lava flows sampled in this study. Direct comparison with an earlier paleomagnetic study performed on the island indicates that, for the three investigated flows, modern demagnetization techniques yield much better defined paleomagnetic directions. The Matuyama-Brunhes transition was recorded in a three-flow section and is dated at 781 +/- 18 ka, in good agreement with other recent radiometric age determinations. The mean paleomagnetic pole calculated from the 23 normal polarity flows is indistinguishable from geographic north, which implies that no significant persistent axial quadrupole term can be identified at this site for the last million years. This result contradicts earlier results and has important implications for models of the time-averaged field (TAF). An Occam algorithm was used to construct a TAF model from the global volcanic database from the last 5 Ma. Substitution of the mean direction calculated from the earlier study for the Lesser Antilles by our new mean value reduces the quadrupole term by more than 30%. This effect, which was produced by changing data from a single site, demonstrates that older paleomagnetic sites may need to be reinvestigated. Furthermore, it also highlights the limitations of TAF models that can be inferred From paleomagnetic databases in their present state.