New global seamount census from altimetry-derived gravity data

Recent revisions to the satellite-derived vertical gravity gradient (VGG) data reveal more detail of the ocean bottom and have allowed us to develop a non-linear inversion method to detect seamounts in VGG data. We approximate VGG anomalies over seamounts as sums of individual, partially overlapping, elliptical polynomial functions, which allows us to form a non-linear inverse problem by fitting the polynomial model to the observations. Model parameters for a potential seamount include geographical location, peak VGG amplitude, major and minor axes of the elliptical base, and the azimuth of the major axis. The non-linear inversion is very sensitive to the initial values for the location and amplitude; hence, they are constrained by the centre and amplitude of the uppermost contours obtained with a 1-Eotvos contour interval. With these initial conditions from contouring, we execute a step-wise and fully automated inversion and obtain optimal model estimates for potential seamounts; these are statistically evaluated for significance using the Akaike Information Criterion and F tests. A logarithmic barrier technique is applied to ensure positivity of all seamount amplitudes. After automatic and manual inspections of the model parameters we estimate actual heights and basal ellipses of the inspected potential seamounts directly from the predicted bathymetry grid. In this study, we find globally 24 643 potential seamounts (h >= 0.1 km) that are located away from continental margins; 8458 potential seamounts are taller than 1 km. Although our global estimate is significantly lower than predictions from previous studies, a first-order reconciliation of the size-frequency statistics obtained from those studies reveals that the previous counts are systematically overestimated. Because of the ambiguity of gravity signals due to small seamounts of h < 1 km and the overlap with abyssal hills, we estimate the global seamount census to lie in the 40 000-55 000 range. The seamount data from this study are accessible from http://www.soest.hawaii.edu/PT/SMTS.