Making satellite-derived empirical bathymetry independent of high-quality in-situ depth data: An assessment of four possible model calibration data.

The empirical approach of satellite-derived bathymetry provides a straightforward, easily-implemented, and very effective way of estimating shallow-water depths from high-spatial-resolution satellite images. However, this approach has a great challenge that it requires high-quality in-situ depth data, which often do not exist or are not available due to financial and/or technical reasons, as the depth calibration data. To address this problem, this article assesses the feasibility of replacing high-quality in-situ depth data with four kinds of possible sourced depth data in satellite-derived empirical bathymetry. These four kinds of depth data are satellite altimeter depth data from laser altimeter systems on some earth observing satellites; satellite photogrammetric depth data, which are derived from satellite stereo-images with the photogrammetric method; nautical chart depth data, which come from existing electronic or paper charts; and radiative transfer model-based depth data, which are simulated in accordance with a bio-optical radiative transfer model (RTM). Correspondingly, four kinds of empirical bathymetry's model calibration data are referred to as altimeter data, photogrammetric data, chart data, and RTM data, respectively. The objective of this article is to assess the similarities, differences, limitations, and applicability of these four kinds of possible model calibration data, in order to know which of them is most suitable for a certain empirical bathymetry task. WorldView-2 and Sentinel-2A multi-spectral images covering four typical shallow water areas in Hawaii and Xisha Islands and Stumpf's logarithmic band ratio model were used to compare and contrast the bathymetric performance of these four kinds of possible model calibration data. The result showed that all four kinds of possible model calibration data can be used for satellite-derived empirical bathymetry when the high-quality in-situ depth data are not provided, but they will provide different bathymetric accuracy values. According to five evaluation indexes (i.e., bathymetric performance, data availability, data quality, complexity of data generation, and data adaptability), the altimeter data perform best, the chart data perform second best, the RTM data follow with the third performance, and the photogrammetric data perform markedly worse than other three kinds of model calibration data. Such factors as the in-situ property, spatial coverage, co-registration and independence of model calibration data must be taken into consideration as well during a model calibration data selection. [ABSTRACT FROM AUTHOR]