Application of a Topological Metric for Assessing Numerical Ocean Models with Satellite Observations

Abstract:

Satellite-based sensors provide a vast amount of observational data over the world ocean. Active microwave radars measure changes in sea surface height and backscattering from surface waves. Data from passive radiometers sensing emissions in multiple spectral bands can directly measure surface temperature, be combined with other data sources to estimate salinity, or processed to derive estimates of optically significant quantities, such as concentrations of biochemical properties. Estimates of the hydrographic variables can be readily used for assimilation or assessment of hydrodynamic ocean models. Optical data, however, have been underutilized in ocean circulation modeling. Qualitative assessments of oceanic fronts and other features commonly associated with changes in optically significant quantities are often made through visual comparison. This project applies a topological approach, borrowed from the field of computer image recognition, to quantitatively evaluate ocean model simulations of features that are related to quantities inferred from satellite imagery. The Modified Hausdorff Distance (MHD) provides a measure of the similarity of two shapes. Examples of applications of the MHD to assess ocean circulation models are presented. The first application assesses several models’ representation of the freshwater plume structure from the Mississippi River, which is associated with a significant expression of color, using a satellite-derived ocean color index. Even though the variables being compared (salinity and ocean color index) differ, the MHD allows contours of the fields to be compared topologically. The second application assesses simulations of surface oil transport driven by winds and ocean model currents using surface oil maps derived from synthetic aperture radar backscatter data. In this case, maps of time composited oil coverage are compared between the simulations and satellite observations.