Accuracy of Nearshore Bathymetry Inverted From X-Band Radar and Optical Video Data

Shore-based remote sensing platforms are increasingly used to frequently (similar to daily) obtain bathymetric information of large (similar to km(2)) nearshore regions over many years. With recorded wave frequency Omega and wavenumber k (and hence wave phase speed c = Omega/k), bed elevation z(b) can be derived using a model that relates Omega and k to water depth. However, the accuracy of z(b) as a function of the sensor and the method of Omega - k retrieval is not well known, especially not under low-period waves. Here, we assess the accuracy of z(b), based on two sensors with their own method of phase speed retrieval, in a dynamic, kilometer-scale environment (Sand Engine, The Netherlands). Bias in z(b) is systematic. A fast Fourier transform (FFT) method on X-band radar imagery produced z(b) too shallow by 1.0 m for -15 m <= z(b) <= -9 m, and too deep by 2.3 m for z(b) >= -6 m. A cross-spectral method on optical video imagery produced z(b) too shallow by 0.59 m for -10 m <= z(b) <= -5 m, and too deep by 0.92 m for z(b) >= -1 m. Intermediate depths had negligible bias, -0.02 m for the radar-FFT approach and -0.01 m for the video-CS approach. The collapse of the FFT method in shallow water may be explained by the inhomogeneity of the wave field in the 960 m x 960 m analysis windows. A shoreward limit of the FFT method is proposed that depends on z(b) in the analysis windows.