Synthetic-aperture radar (SAR) imagery can provide wavenumber and frequency data to generate bathymetry estimates for locations where limited access or extreme ocean conditions can make standard bathymetry survey techniques difficult or impossible. The availability of SAR data could allow for regular bathymetry estimates of navigational channels providing insight into dredging intervals and efficiency. Estimates would also be of value in military applications to gather information about denied areas. Strong currents, extreme ocean conditions, and significant discharge from the river make the Mouth of the Columbia River (MCR) on the US West Coast a challenging location for bathymetry estimation. Successful bathymetry estimations at the Mouth of the Columbia River could provide opportunities to apply this method to other, less challenging locations.
This study assimilates SAR wave number observations with hydrodynamics characteristics generated from numerical models. A Regional Ocean Modeling System (ROMS) one-way coupled to a Simulated Waves Nearshore model (SWAN) is used to model the wave-current dynamics and provide wave characteristics and current velocities for an estimated bathymetry. Solving the dispersion relationship for the change in wavenumber with respect to water depth allows us to implement weighted least squares method with a Bayesian estimation to minimize the error between the model predicted and observed wavenumbers using covariance matrices. This framework allows for a best guess bathymetry of the Mouth of the Columbia River to be updated by assimilating solely SAR wavenumber observations.
Our goals are to determine if SAR observations from four individual images can be assimilated to estimate a bathymetry that qualitatively reproduces the location and depths of the bathymetric features in the actual bathymetry, explore sensitivity of variations in assimilated frequency, and to establish limitations of this method and SAR observations used for bathymetry inversion. The framework was tested using model-derived observations that mimicked those collected via SAR. The assimilation using the model derived observations illustrated that high quality observations, a highly-skilled model, and ample spatial coverage can result in estimates of the pronounced features of the MCR bathymetry. To utilize the SAR observations for assimilation, it was necessary to assign a frequency to the observed wavenumbers. Assigned frequencies were determined by comparing wave directions between the observations and an offshore buoy. Under certain conditions the framework provided improvements to the initial bathymetry when assimilating the SAR observed wavenumbers and their assigned frequencies. The method could occasionally predict the location and extent of features including the MCR navigational channel, Peacock Spit, Clatsop Spit, and a dredge disposal site. Sensitivity to assigned frequencies in the assimilation results was explored and it was determined that the frequencies observed by the buoy (or very similar to those) provided the best bathymetric results. The results of this method were limited by the presence of breaking waves in the SAR images and spatial coverage of observations.