Sparse distribution of depth soundings in the ocean make it necessary to infer depth in the gaps using alternate information such as satellite-derived gravity and a mapping from gravity to depth. We design and train a neural network on a collection of 50 million depth soundings to predict bathymetry globally using gravity anomalies. We find the best result is achieved by pre-filtering depth and gravity in accordance with isostatic admittance theory described in previous predicted depth studies. When training the model, if the training and testing split is a random partition at the same resolution as the data, the training and testing sets will not be independent, and model misfit results will be too optimistic. We solve this problem by partitioning the training and testing set with geographic bins. Our final predicted depth model improves on old predicted depth model rms by 16%, from 165 m to 138 m. Among constrained grid cells, 80% of the predicted values are within 128 m of the true value. Improvements to this model will continue with additional depth measurements, but higher resolution predictions, being limited by upward continuation of gravity, shouldn’t be attempted with this method.

A document by Hugh Harper. Click on the document to view its contents.

To date, approximately 20% of the ocean floor has been surveyed by ships at a spatial resolution of 400 m or better. The remaining 80% has depth predicted from satellite altimeter-derived gravity measurements at a relatively low resolution. There are many remote ocean areas in the southern hemisphere that will not be completely mapped at 400 m resolution during this decade. This study is focused on the development of synthetic bathymetry to fill the gaps. There are two types of seafloor features that are not typically well resolved by satellite gravity: abyssal hills and small seamounts (< 2.5 km tall). We generate synthetic realizations of abyssal hills by combining the measured statistical properties of mapped abyssal hills with regional geology including fossil spreading rate/orientation, rms height from satellite gravity, and sediment thickness. With recent improvements in accuracy and resolution, It is now possible to detect all seamounts taller than about 800 m in satellite-derived gravity and their location can be determined to an accuracy of better than 1 km. However, the width of the gravity anomaly is much greater than the actual width of the seamount so the seamount predicted from gravity will underestimate the true seamount height and overestimate its base dimension. In this study we use the amplitude of the vertical gravity gradient (VGG) to estimate the mass of the seamount and then use their characteristic shape, based on well surveyed seamounts, to replace the smooth predicted seamount with a seamount having a more realistic shape.