The first attempts to quantify the width and height of hotspot swells were made more than 30 years ago. Since that time, global bathymetry, ocean-floor age, and sediment thickness datasets have improved considerably. Swell heights and widths have been used to estimate the heat flow from the core-mantle boundary, constrain numerical models of plumes, and as an indicator of the origin of hotspots. In this paper, we repeat the analysis of swell geometry and buoyancy flux for 54 hotspots, including the 37 considered by Sleep (1990) and the 49 considered by Courtillot et al. (2003), using the latest and most accurate data. We are able to calculate swell geometry for a number of hotspots that Sleep was only able to estimate by comparison with other swells. We find that in spite of the increased resolution in global bathymetry models there is significant uncertainty in our calculation of buoyancy fluxes due to differences in our measurement of the swells' width and height, the integration method (volume integration or cross-sectional area), and the variations of the plate velocities between HS2-Nuvel1a (Gripp and Gordon, 1990) and HS3-Nuvel1a (Gripp and Gordon, 2002). We also note that the buoyancy flux for Pacific hotspots is in general larger than for Eurasian, North American, African and Antarctic hotspots. Considering that buoyancy flux is linearly related to plate velocity, we speculate that either the calculation of buoyancy flux using plate velocity over-estimates the actual vertical flow of material from the deep mantle or that convection in the Pacific hemisphere is more vigorous than the Atlantic hemisphere.

No abstract

both parameters by computing volumes along the hotspot tracks. Neither Walvis nor St. Helena show a 'classical' hotspot behavior. We find that two plumes are at the origin of the St. Helena chain. This study also shows a swell associated with the Circe seamount, supporting the existence of a hotspot NW of the St. Helena trail. The variation in swell and volcanic fluxes suggests temporal variability in the plume behavior at time scales of 10-20 m.y. and 5 m.y., which may be related to oscillations and instabilities of the plume conduit, respectively. Cumulative fluxes in the area are largest for Walvis and weakest for Circe, and all are significantly lower than that reported for the Hawai'i hotspot.

[1] We propose a filtering method to characterize large-scale depth anomalies. The MiFil method (for minimization and filtering) requires two stages: a first one to roughly remove the volcano component by minimizing the depth anomaly and a second one to smooth the shape and totally remove the small spatial length scale remaining topography using a median filter. The strength of this method, directly applicable on two-dimensional grids, is that it does not require any assumption on the location, amplitude, or width of the large-scale feature to characterize, except its minimal width. We only consider the spatial length scale of the features to remove. Application to volcanic chains of the south central Pacific is presented, and the results lead to a better understanding of the tectonics and volcanism emplacement of the zone. The Society is the only ''classical'' hot spot that corresponds to the simple interaction of a plume with the lithosphere and for which a buoyancy flux of 1.58 ± 0.15 Mg s À1 is obtained. The Marquesas volcanic chain, although quite comparable, presents a swell morphology that prevents such interpretation and quantification. For the Tuamotu and Cook-Austral volcanic chains, no reliable quantification can be made because the depth and geoid anomalies are caused by several phenomena occurring at different depths that cannot be separated.

The South Pacific Superswell is known as a broad area beneath French Polynesia characterized by numerous volcanic chains and very shallow seafloor compared to the depth predicted for its age by classical seafloor subsidence models. So far, its exact extent has not yet been established. Thanks to better bathymetric coverage and a specially adapted filtering method, we present here a new, complete, and precise mapping of the superswell. We show that the superswell covers a region broader than previously expected. It extends between latitudes 10°N and 30°S and longitudes 130°W and 160°W and has a maximum amplitude of 680 m. It is composed of two branches that display different characteristics when compared to other geophysical observations. Under the southern branch a dip in the geoid is observed which could be linked to upwelling in a convective mantle where the low‐velocity zone is located immediately below the lithosphere as indicated by tomography models. Under the northern branch, there is a 12 m high in the geoid surface anomaly that may be explained by isostatic compensation of a mass deficiency located in the mantle at 300 km depth. The northern branch could also be related to another superplume that is now going down but which was rising up 30–90 m.y. ago, when it initiated the secondary plumes which created the Line Islands.