The oceans are truly Earth’s last great unknown. With about 75% of the world's ocean and 53% of the U.S. ocean, coastal, and Great Lakes waters unmapped, interpolation across large distances is typically required among sparse bathymetric datasets/measurements in order to fill the data gaps and create a definitive model of the seafloor. As with any scientific measurement, the uncertainty that comes with bathymetry is important, but particularly for the application of nautical charting and navigational safety, it is crucial as it can be one of the factors of maritime accidents. The widespread use of digital bathymetric models developed using interpolation methods across numerous disciplines in the academic, government, and private sectors are on the increase with little or no accompanying estimates of the uncertainty inherent in the models. This calls for additional work to bolster the existing ones in this area of research in ocean mapping as the significance of estimating uncertainty in these models ranges from academically and commercially beneficial to potentially lifesaving. Hence, this work will present a novel machine learning approach to estimating and characterizing the uncertainties in bathymetric models given different datasets combinations, qualities, and seafloor morphologies to better quantify the estimate of interpolation uncertainty. Our ongoing investigation is evaluating the performance of deterministic and stochastic interpolation techniques on different testbeds with different datasets in the U.S. This effort is targeted at determining an optimal interpolation method that is fit for purpose and provides the best estimate of uncertainty based on different data combinations, qualities, and seafloor morphologies in operational settings. An optimal interpolation method in the context of this work is one that meets the application requirements (for example preserves shallow depths and does not overestimate depth for nautical charting and safety of navigation; preserves morphology for modeling applications, etc.) while still providing the best estimate of uncertainty. Beyond the broader impact of better filling the bathymetric gaps, the findings of this work are expected to help the hydrospatial community confidently choose a situation-specific – i.e., based on location, depth, data characteristics, morphologies – optimal interpolation method. Consequently, the eventual benefits will be enhancing nautical chart accuracy, hydrographic survey planning, and improving National Oceanic and Atmospheric Administration (NOAA) data-driven projects such as the National Bathymetric Source, Hydrographic Health Modelling, Precision Navigation, etc.