Objective, unbiased, multi-scale regional-residual separation of shipboard bathymetry data

Seafloor topography comprises a multi-scale hierarchy of superimposed geological features. Short-wavelength submarine volcanic features, such as seamounts and oceanic islands, for example, maybe superimposed on long-wavelength mid-plate topographic swells. Traditional techniques of separating seamounts and oceanic islands from the depth of the surrounding seafloor (e.g. Bandpass, Gaussian, Median and Modal filtering) are intrinsically bound to scale criteria. At untargeted scales, the dimensions of seamounts and oceanic islands maybe inaccurately recovered and regional trends in surrounding seafloor poorly tracked. Shipboard bathymetry dat acquired during VEMA cruise 3312, which crossed the Japan and Mariana Trenches and the Magellan Seamount province illustrate these points well. Of all techniques, a 130 km wide median filter best approximates a manually fitted regional depth of the seafloor. This filter correctly delineates 39.2% of the Magellan seamounts and the RMS difference between observed and calculated seafloor is 352 m. We found, however, that the ability of the filter to delineate the seamounts is scale-dependant, with the dimensions (e.g. height, width) of 40-75 km wide features being significantly better delineated than 0-30 km and 100-200 km wide ones. In order to address this scale-dependance, we have developed two objective, multi-scale, techniques. Firstly, a Micro-Macro Interpretation Construct (MiMIC) algorithm has been constructed which simulates manual interpretation. A second algorithm interprets a Spatial Wavelet Transform (SWT). The two algorithms, when applied to the Vema cruise data, reduce the RMS difference from 352 m to 63m and 83m respectively. There is also a significant improvement in the dimensions of the seamounts recovered. In each case, the algorithms out perform a median filter, especially outside of its targeted width. We found that the derived regional bathymetry is unbiased by seamounts, even in deep-sea trenches and some fracture zones where the regional seafloor is unusually deep and in seamount clusters where the regional seafloor is unusually shallow. In this paper, we use multi-scale techniques to delineate and estimate the volumes of seamounts over large areas of the Pacific seafloor. The implications of the work are then examined for studies which require a regional depth of the seafloor, which is not "contaminated" by seamounts, such as those that involve thermal modeling of the relationship between seafloor depth and crustal age and mantle dynamic modeling of mid-plate topographic swells.